# Git Signal — llmgram.app > Curated catalog of AI/ML GitHub repos. Each one analyzed by Grok: what it does, why it matters, audience, maturity, alternatives. > **Last updated:** 2026-04-22 21:35 UTC > **Web:** > **Raw JSON:** > **This file:** --- ## Overview | Metric | Value | |---|---| | Total repos | 2461 | | Total stars (cumulative) | 19,351,719 | | Categories | 10 | | Languages | 50 | ### By category - **library**: 316 - **tool**: 161 - **model**: 114 - **framework**: 70 - **demo**: 64 - **sdk**: 58 - **agent**: 45 - **eval**: 42 - **infra**: 29 - **data**: 26 ### By language (top 10) - **Python**: 1296 - **TypeScript**: 262 - **Jupyter Notebook**: 228 - **C++**: 98 - **JavaScript**: 77 - **Rust**: 65 - **Go**: 54 - **HTML**: 44 - **C#**: 43 - **C**: 30 ### By maturity - **beta**: 352 - **production**: 300 - **experimental**: 236 - **archived**: 37 --- ## Leaderboards ### Hot now - **langchain-ai/langchain** · hot=0.96 · trend=up · ★134,515 - **Significant-Gravitas/AutoGPT** · hot=0.94 · trend=up · ★183,670 - **run-llama/llama_index** · hot=0.93 · trend=up · ★48,813 - **vllm-project/vllm** · hot=0.92 · trend=up · ★77,729 - **infiniflow/ragflow** · hot=0.92 · trend=up · ★78,763 - **langchain-ai/langgraph** · hot=0.92 · trend=up · ★30,036 - **huggingface/transformers** · hot=0.91 · trend=up · ★159,759 - **OpenHands/OpenHands** · hot=0.91 · trend=up · ★71,771 - **affaan-m/everything-claude-code** · hot=0.90 · trend=up · ★161,026 - **langgenius/dify** · hot=0.90 · trend=up · ★138,311 - **mudler/LocalAI** · hot=0.90 · trend=up · ★45,704 - **crewAIInc/crewAI** · hot=0.90 · trend=up · ★49,559 ### Rising - **langchain-ai/langchain** · rising=0.95 · trend=up · ★134,515 - **ollama/ollama** · rising=0.94 · trend=up · ★169,714 - **ggml-org/llama.cpp** · rising=0.94 · trend=up · ★105,751 - **Significant-Gravitas/AutoGPT** · rising=0.94 · trend=up · ★183,670 - **huggingface/transformers** · rising=0.94 · trend=up · ★159,759 - **crewAIInc/crewAI** · rising=0.93 · trend=up · ★49,559 - **run-llama/llama_index** · rising=0.93 · trend=up · ★48,813 - **openclaw/openclaw** · rising=0.92 · trend=up · ★362,363 - **cline/cline** · rising=0.92 · trend=up · ★60,586 - **OpenHands/OpenHands** · rising=0.92 · trend=up · ★71,771 - **mudler/LocalAI** · rising=0.92 · trend=up · ★45,704 - **openai/openai-python** · rising=0.92 · trend=up · ★30,559 ### Durable - **langchain-ai/langchain** · durable=0.92 · trend=up · ★134,515 - **affaan-m/everything-claude-code** · durable=0.91 · trend=up · ★161,026 - **deepseek-ai/DeepSeek-V3** · durable=0.89 · trend=stable · ★102,727 - **openai/whisper** · durable=0.88 · trend=up · ★98,030 - **run-llama/llama_index** · durable=0.88 · trend=up · ★48,813 - **open-webui/open-webui** · durable=0.87 · trend=up · ★132,658 - **langgenius/dify** · durable=0.87 · trend=up · ★138,311 - **huggingface/transformers** · durable=0.87 · trend=up · ★159,759 - **Significant-Gravitas/AutoGPT** · durable=0.86 · trend=up · ★183,670 - **mudler/LocalAI** · durable=0.86 · trend=up · ★45,704 - **hiyouga/LlamaFactory** · durable=0.86 · trend=up · ★70,476 - **Mintplex-Labs/anything-llm** · durable=0.86 · trend=up · ★58,596 --- ## Repos ### agent (45) #### [Significant-Gravitas/AutoGPT](https://github.com/Significant-Gravitas/AutoGPT) *Python · ★183,670 · NOASSERTION · beta · score:0.85 · hot:0.94 · rising:0.94 · durable:0.86 · board:rising · trend:up* AutoGPT is the vision of accessible AI for everyone, to use and to build on. Our mission is to provide the tools, so that you can focus on what matters. **Why it matters.** AutoGPT is an open-source platform that enables users to build, deploy, and manage autonomous AI agents using LLMs like GPT or Claude to automate complex workflows. It matters now because it democratizes AI agent development amid growing interest in agentic systems, though its beta status and technical setup requirements mean it's more relevant for those ready to experiment rather than immediate production use. _themes: agents · autonomous-agents · llm · ai-platform_ #### [OpenHands/OpenHands](https://github.com/OpenHands/OpenHands) *Python · ★71,771 · NOASSERTION · production · score:0.85 · hot:0.91 · rising:0.92 · durable:0.85 · board:rising · trend:up* 🙌 OpenHands: AI-Driven Development **Why it matters.** OpenHands is a toolkit for AI-driven development, offering a Python SDK, CLI, and GUI to build, run, and scale AI agents using LLMs like GPT or Claude for tasks such as code generation and automation. It matters now because the growing adoption of LLMs in development workflows demands accessible tools for creating agent-based systems, helping developers integrate AI efficiently amid rapid advancements in the field. _themes: agents · llm · cli · developer-tools_ #### [affaan-m/everything-claude-code](https://github.com/affaan-m/everything-claude-code) *JavaScript · ★161,026 · MIT · production · score:0.80 · hot:0.90 · rising:0.92 · durable:0.91 · board:rising · trend:up* The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond. **Why it matters.** This repository offers a performance optimization system for AI agent harnesses like Claude Code, including features for skills management, memory persistence, security scanning, and continuous learning to enhance efficiency and productivity. It matters now because the growing adoption of AI agents in development workflows demands robust tools for real-world optimization, and this project provides battle-tested configurations evolved from extensive use. However, its focus on specific tools like Claude may limit broader applicability amid the rapidly evolving AI landscape. _themes: agents · llm · optimization · productivity_ #### [zhayujie/CowAgent](https://github.com/zhayujie/CowAgent) *Python · ★43,516 · MIT · production · score:0.80 · hot:0.90 · rising:0.89 · durable:0.85 · board:hot · trend:up* CowAgent (chatgpt-on-wechat) 是基于大模型的超级AI助理，能主动思考和任务规划、访问操作系统和外部资源、创造和执行Skills、通过长期记忆和知识库不断成长，比OpenClaw更轻量和便捷。同时支持微信、飞书、钉钉、企微、QQ、公众号、网页等接入，可选择OpenAI/Claude/Gemini/DeepSeek/ Qwen/GLM/Kimi/LinkAI，能处理文本、语音、图片和文件，可快速搭建个人AI助理和企业数字员工。 **Why it matters.** CowAgent is an open-source AI agent framework that builds autonomous assistants capable of task planning, memory management, skill execution, and multi-modal interactions across platforms like WeChat and QQ, using various LLMs. It matters now due to the growing demand for customizable, lightweight AI agents in enterprise and personal settings, especially in regions with strong messaging app integration, offering a practical alternative to heavier frameworks amid the AI agent boom. Its high adoption, with over 43k stars, highlights community interest in accessible tools for real-world automation. _themes: agents · llm · multi-modal · task-planning_ #### [lobehub/lobehub](https://github.com/lobehub/lobehub) *TypeScript · ★75,355 · NOASSERTION · production · score:0.80 · hot:0.89 · rising:0.88 · durable:0.83 · board:hot · trend:up* The ultimate space for work and life — to find, build, and collaborate with agent teammates that grow with you. We are taking agent harness to the next level — enabling multi-agent collaboration, effortless agent team design, and introducing agents as the unit of work interaction. **Why it matters.** LobeHub is a platform for building, collaborating with, and managing AI agents, focusing on multi-agent systems and integration with models like GPT and Claude to facilitate team-based workflows. It matters now because AI agents are increasingly used for automated tasks in professional settings, but its hype-driven popularity with 75k stars may overshadow potential limitations in originality or scalability compared to established tools. The 'NOASSERTION' license adds uncertainty for enterprise adoption. _themes: agents · collaboration · multi-agent · ai-integration_ #### [cline/cline](https://github.com/cline/cline) *TypeScript · ★60,586 · Apache-2.0 · production · score:0.80 · hot:0.88 · rising:0.92 · durable:0.83 · board:rising · trend:up* Autonomous coding agent right in your IDE, capable of creating/editing files, executing commands, using the browser, and more with your permission every step of the way. **Why it matters.** Cline is an AI-powered agent that integrates into your IDE to perform tasks like creating or editing files, executing commands, and browsing the web, but only with explicit user permission to ensure safety. It matters now because it represents a practical step towards autonomous coding assistance in an era of advancing AI agents, potentially boosting developer productivity while addressing security concerns in software development workflows. _themes: agents · ai-assistant · coding · automation_ #### [NousResearch/hermes-agent](https://github.com/NousResearch/hermes-agent) *Python · ★101,542 · MIT · production · score:0.80 · hot:0.88 · rising:0.88 · durable:0.82 · board:hot · trend:up* The agent that grows with you **Why it matters.** Hermes Agent is a self-improving AI agent that learns from user interactions, supports various language models, and runs on low-cost infrastructure, enabling persistent and personalized AI assistants. It matters now because it addresses the growing need for adaptable, cost-effective agents in AI development, potentially reducing vendor lock-in and enhancing efficiency for real-world applications. _themes: agents · llm · self-improving · multi-model_ #### [openclaw/openclaw](https://github.com/openclaw/openclaw) *TypeScript · ★362,363 · MIT · production · score:0.70 · hot:0.88 · rising:0.92 · durable:0.82 · board:rising · trend:up* Your own personal AI assistant. Any OS. Any Platform. The lobster way. 🦞 **Why it matters.** OpenClaw is an open-source personal AI assistant that runs locally on your devices, integrating with numerous messaging platforms like WhatsApp and Telegram to provide a private, always-on conversational experience. It matters now because it addresses growing privacy concerns by allowing users to own their data and avoid cloud dependencies, offering a viable alternative in an era of increasing AI surveillance and data breaches. _themes: agents · privacy · local-ai · integration_ #### [HKUDS/nanobot](https://github.com/HKUDS/nanobot) *Python · ★40,081 · MIT · beta · score:0.80 · hot:0.88 · rising:0.86 · durable:0.79 · board:hot · trend:up* "🐈 nanobot: The Ultra-Lightweight Personal AI Agent" **Why it matters.** Nanobot is an ultra-lightweight personal AI agent built in Python that provides core functionalities for interacting with LLMs like OpenAI and Anthropic, emphasizing minimal code and integrations for tasks such as web search and real-time updates. It matters now because it lowers the barrier to AI agent development in a crowded field, offering efficiency and rapid feature additions, which could appeal to developers seeking scalable alternatives amid ongoing AI advancements. _themes: agents · llm · lightweight · integration_ #### [openai/codex](https://github.com/openai/codex) *Rust · ★77,020 · Apache-2.0 · production · score:0.85 · hot:0.87 · rising:0.92 · durable:0.81 · board:rising · trend:up* Lightweight coding agent that runs in your terminal **Why it matters.** Codex is a CLI-based coding agent from OpenAI that provides AI-assisted coding help, such as code generation and suggestions, running locally in the terminal. It matters now because it integrates with OpenAI's ecosystem and ChatGPT plans, offering practical productivity tools for developers amid the growing adoption of AI in coding workflows, though its reliance on proprietary models limits full open-source flexibility. _themes: agents · inference · cli · ai-coding_ #### [aaif-goose/goose](https://github.com/aaif-goose/goose) *Rust · ★42,719 · Apache-2.0 · production · score:0.80 · hot:0.87 · rising:0.89 · durable:0.83 · board:rising · trend:up* an open source, extensible AI agent that goes beyond code suggestions - install, execute, edit, and test with any LLM **Why it matters.** Goose is an open-source AI agent built in Rust that allows users to install, execute, edit, and test code or workflows using various LLMs, offering a desktop app, CLI, and API for broader automation tasks. It matters now because AI agents are increasingly demanded for practical, everyday applications amid LLM proliferation, but its reliance on external providers and ongoing transition to a new foundation raises questions about stability and true independence. While extensible, it may not fully differentiate from similar tools in a crowded market. _themes: agents · ai-agents · llm · workflows_ #### [continuedev/continue](https://github.com/continuedev/continue) *TypeScript · ★32,725 · Apache-2.0 · production · score:0.80 · hot:0.86 · rising:0.88 · durable:0.81 · board:rising · trend:up* ⏩ Source-controlled AI checks, enforceable in CI. Powered by the open-source Continue CLI **Why it matters.** Continue is a tool that enables source-controlled AI checks for codebases, allowing developers to define and enforce AI-powered agents in CI pipelines to automate code reviews and security checks. It matters now because as AI-assisted coding proliferates, ensuring code quality and catching vulnerabilities early is essential to mitigate risks, and this integrates seamlessly with existing workflows like GitHub pull requests. However, its reliance on external AI models means it could introduce dependencies and potential inaccuracies that require careful management. _themes: agents · ai-checks · ci-integration · llm_ #### [assafelovic/gpt-researcher](https://github.com/assafelovic/gpt-researcher) *Python · ★26,612 · Apache-2.0 · production · score:0.80 · hot:0.86 · rising:0.87 · durable:0.84 · board:rising · trend:up* An autonomous agent that conducts deep research on any data using any LLM providers **Why it matters.** GPT Researcher is an open-source autonomous agent that uses LLMs to perform deep research, including web scraping and report generation with citations, addressing issues like misinformation and token limits. It matters now as AI-driven research tools are increasingly needed for efficient, unbiased information gathering in a world flooded with outdated or hallucinatory data from standard LLMs, and its integration with platforms like Claude enhances practical applications. _themes: agents · rag · research · llms_ #### [anthropics/claude-code](https://github.com/anthropics/claude-code) *Shell · ★115,818 · no-license · production · score:0.80 · hot:0.83 · rising:0.87 · durable:0.79 · board:rising · trend:up* Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflows - all through natural language commands. **Why it matters.** Claude Code is a terminal-based AI tool from Anthropic that assists developers by understanding codebases, executing routine tasks, and managing git workflows through natural language commands, potentially improving productivity in everyday coding. It matters now due to the growing demand for AI-powered coding assistants amid the AI boom, but its lack of a specified license raises concerns about accessibility and long-term adoption. However, its high popularity with over 115k stars indicates strong community interest in integrating AI directly into development workflows. _themes: agents · ai-tools · code-assist · nlp_ #### [datawhalechina/hello-agents](https://github.com/datawhalechina/hello-agents) *Python · ★38,479 · NOASSERTION · production · score:0.85 · hot:0.83 · rising:0.84 · durable:0.83 · board:rising · trend:up* 📚 《从零开始构建智能体》——从零开始的智能体原理与实践教程 **Why it matters.** This repository provides a comprehensive, hands-on tutorial for building AI agents from scratch, covering core principles, practical implementations, and real-world applications using LLMs and RAG. It matters now because 2025 is emerging as the year of agents, shifting focus from large language models to intelligent applications, and this resource addresses the lack of systematic guides to help users become builders in this rapidly evolving field. _themes: agents · llm · rag · tutorial_ #### [QwenLM/qwen-code](https://github.com/QwenLM/qwen-code) *TypeScript · ★23,546 · Apache-2.0 · beta · score:0.80 · hot:0.81 · rising:0.84 · durable:0.74 · board:rising · trend:up* An open-source AI agent that lives in your terminal. **Why it matters.** Qwen Code is an open-source AI agent that operates in the terminal, assisting developers with understanding codebases, automating tasks, and integrating with various AI models like Qwen series via APIs. It matters now due to recent updates like the discontinuation of free tiers and new model releases, pushing users towards paid options amid growing demand for efficient code assistants in an evolving AI landscape. _themes: agents · code-assistant · terminal · inference_ #### [SWE-agent/SWE-agent](https://github.com/SWE-agent/SWE-agent) *Python · ★19,036 · MIT · beta · score:0.70 · hot:0.78 · rising:0.81 · durable:0.80 · board:rising · trend:up* SWE-agent takes a GitHub issue and tries to automatically fix it, using your LM of choice. It can also be employed for offensive cybersecurity or competitive coding challenges. [NeurIPS 2024] **Why it matters.** SWE-agent is an AI-powered tool that uses language models to autonomously fix GitHub issues, detect cybersecurity vulnerabilities, and handle custom tasks by interacting with tools, but its development has been superseded by a simpler version called mini-SWE-agent. It matters now as it represents a benchmark in automated software engineering research, achieving state-of-the-art results on SWE-bench, though users are recommended to switch to the newer alternative for ongoing work. _themes: agents · llm · automation · cybersecurity_ #### [bytedance/deer-flow](https://github.com/bytedance/deer-flow) *Python · ★62,686 · MIT · beta · score:0.80 · hot:0.77 · rising:0.79 · durable:0.75 · board:rising · trend:up* An open-source long-horizon SuperAgent harness that researches, codes, and creates. With the help of sandboxes, memories, tools, skill, subagents and message gateway, it handles different levels of tasks that could take minutes to hours. **Why it matters.** DeerFlow is an open-source framework for building advanced AI agents that handle complex, multi-step tasks like research and coding using sub-agents, tools, and memory systems. It matters now because the growing need for autonomous agents in AI development could improve efficiency in long-horizon workflows, though its heavy reliance on specific proprietary models and lack of formal releases raise questions about accessibility and stability. _themes: agents · multi-agent · workflow · research_ #### [microsoft/Multi-Agent-Custom-Automation-Engine-Solution-Accelerator](https://github.com/microsoft/Multi-Agent-Custom-Automation-Engine-Solution-Accelerator) *Python · ★780 · MIT · production · score:0.70 · hot:0.76 · rising:0.78 · durable:0.68 · board:rising · trend:up* The Multi-Agent Custom Automation Engine Solution Accelerator is an AI-driven system that manages a group of AI agents to accomplish tasks based on user input. Powered by Microsoft Agent Framework, Azure Foundry, Azure Cosmos DB, and infrastructure services, it provides a reference application, allowing you to hit the ground running. **Why it matters.** This repository provides a solution accelerator for orchestrating multiple AI agents to automate business tasks using Microsoft's tools like Azure services, serving as a reference for quick deployment. It matters now as enterprises increasingly adopt multi-agent systems for efficiency in complex workflows, though it requires users to manage risks and ensure compliance with AI regulations. However, its heavy reliance on Microsoft-specific technologies may limit broader applicability. _themes: agents · automation · orchestration · azure_ #### [anthropics/skills](https://github.com/anthropics/skills) *Python · ★120,389 · no-license · beta · score:0.85 · hot:0.75 · rising:0.79 · durable:0.77 · board:rising · trend:up* Public repository for Agent Skills **Why it matters.** This repository provides example implementations of skills for Anthropic's Claude AI, which are dynamic modules that enhance the model's performance on specialized tasks like document creation or data analysis. It matters right now because AI agents are rapidly evolving for real-world applications, offering customization options, but its lack of a license and official releases limits practical adoption and collaboration. _themes: agents · skills · ai-customization · python_ #### [mistralai/mistral-vibe](https://github.com/mistralai/mistral-vibe) *Python · ★3,917 · Apache-2.0 · beta · score:0.70 · hot:0.75 · rising:0.75 · durable:0.68 · board:rising · trend:up* Minimal CLI coding agent by Mistral **Why it matters.** Mistral Vibe is a CLI-based coding assistant that uses Mistral's AI models to enable natural language interactions for exploring and modifying codebases, though it's limited to primarily UNIX environments and may not handle complex tasks as effectively as more integrated IDE tools. It matters now because AI-assisted coding is increasingly popular for boosting developer productivity, but this implementation is straightforward and open-source, offering an accessible alternative amid growing competition from proprietary solutions, albeit with potential reliability issues on non-supported platforms. _themes: agents · cli · coding · llm_ #### [microsoft/RD-Agent](https://github.com/microsoft/RD-Agent) *Python · ★12,585 · MIT · beta · score:0.85 · hot:0.74 · rising:0.77 · durable:0.79 · board:durable · trend:up* Research and development (R&D) is crucial for the enhancement of industrial productivity, especially in the AI era, where the core aspects of R&D are mainly focused on data and models. We are committed to automating these high-value generic R&D processes through R&D-Agent, which lets AI drive data-driven AI. 🔗https://aka.ms/RD-Agent-Tech-Report **Why it matters.** RD-Agent is an AI framework that automates research and development processes, focusing on data mining, model development, and other R&D tasks using LLM-powered agents. It matters right now because it leads on benchmarks like MLE-Bench and has a NeurIPS 2025 acceptance, demonstrating its effectiveness in boosting AI-driven productivity amid growing demands for efficient data and model workflows. _themes: agents · automation · llm · data-science_ #### [huggingface/agents-course](https://github.com/huggingface/agents-course) *MDX · ★27,983 · Apache-2.0 · production · score:0.85 · hot:0.73 · rising:0.78 · durable:0.77 · board:rising · trend:up* This repository contains the Hugging Face Agents Course. **Why it matters.** This repository provides a free online course from Hugging Face on building and understanding AI agents, covering topics from basics to advanced implementations using tools like Langchain and LlamaIndex. It matters right now because AI agents are a rapidly evolving area in AI development, enabling more autonomous applications, and this course offers accessible education to help developers and enthusiasts stay current amid growing industry demand. _themes: agents · education · llms · ai-course_ #### [shareAI-lab/learn-claude-code](https://github.com/shareAI-lab/learn-claude-code) *TypeScript · ★54,799 · MIT · experimental · score:0.60 · hot:0.72 · rising:0.74 · durable:0.71 · board:rising · trend:up* Bash is all you need - A nano claude code–like 「agent harness」, built from 0 to 1 **Why it matters.** This repository provides tutorials and code in TypeScript to build a simple agent harness for AI models like Claude, focusing on the infrastructure that enables models to interact with environments rather than the models themselves. It matters now because, with the rapid growth of AI agents, developers need practical guidance on integrating LLMs into real applications, and this educational resource demystifies the process for those without deep AI expertise. However, its lack of a formal release and basic scope means it's more of a starting point than a comprehensive solution. _themes: agents · llm · tutorial · harness_ #### [microsoft/magentic-ui](https://github.com/microsoft/magentic-ui) *Python · ★9,780 · MIT · experimental · score:0.75 · hot:0.72 · rising:0.74 · durable:0.75 · board:durable · trend:up* A research prototype of a human-centered web agent **Why it matters.** Magentic-UI is a research prototype AI agent that automates web browsing, coding tasks, and file analysis while prioritizing user control through plan revelation and approval requests. It matters now because it tackles transparency and safety in AI agents amid growing concerns over autonomous systems, especially with integrations like the Fara-7B model that enhance real-time monitoring capabilities. _themes: agents · ai-ux · web-automation · human-in-the-loop_ #### [acon96/home-llm](https://github.com/acon96/home-llm) *Python · ★1,310 · no-license · beta · score:0.65 · hot:0.71 · rising:0.69 · durable:0.61 · board:hot · trend:up* A Home Assistant integration & Model to control your smart home using a Local LLM **Why it matters.** This repository offers a Home Assistant custom component that integrates a local Large Language Model for controlling smart home devices via voice and chat, emphasizing privacy by running entirely on the user's hardware. It matters now amid rising concerns over data privacy in AI and the increasing adoption of edge computing for home automation, though its lack of a specified license could limit broader use. _themes: local-llm · home-automation · voice-control · inference_ #### [openai/skills](https://github.com/openai/skills) *Python · ★17,097 · no-license · beta · score:0.85 · hot:0.70 · rising:0.73 · durable:0.69 · board:rising · trend:up* Skills Catalog for Codex **Why it matters.** This repository catalogs skills for OpenAI's Codex, which are collections of instructions, scripts, and resources that AI agents can use to perform specific tasks, enabling modular and reusable capabilities. It matters now because as AI agents become more prevalent in development workflows, this provides a standardized way to share and integrate skills, potentially improving efficiency for users of Codex amid growing demand for AI automation tools. _themes: agents · ai-tools · codex · skills_ #### [UfoMiao/zcf](https://github.com/UfoMiao/zcf) *TypeScript · ★5,926 · MIT · production · score:0.75 · hot:0.70 · rising:0.70 · durable:0.78 · board:durable · trend:stable* Zero-Config Code Flow for Claude code & Codex **Why it matters.** ZCF is a CLI tool that offers zero-configuration setup for AI coding assistants like Claude and Codex, including features for workflows, intelligent agents, and bilingual support, all built with TypeScript. It matters now because it simplifies AI integration for developers amid the growing adoption of LLMs in coding, but its sponsorship ties and focus on specific models could limit versatility and raise dependency concerns. _themes: agents · llm · workflow · cli_ #### [microsoft/fara](https://github.com/microsoft/fara) *Python · ★4,955 · MIT · experimental · score:0.70 · hot:0.70 · rising:0.71 · durable:0.66 · board:rising · trend:up* Fara-7B: An Efficient Agentic Model for Computer Use **Why it matters.** Fara-7B is a 7-billion parameter small language model from Microsoft designed as an agent for computer interactions, such as web browsing and task execution, emphasizing efficiency for local deployment. It matters now because it addresses the growing demand for lightweight AI agents that can run on consumer hardware, potentially making autonomous tools more accessible amid resource constraints in AI development, though its lack of a formal release raises questions about stability and broader applicability. _themes: agents · inference · efficiency · slm_ #### [huggingface/skills](https://github.com/huggingface/skills) *Python · ★10,228 · Apache-2.0 · beta · score:0.85 · hot:0.69 · rising:0.73 · durable:0.72 · board:rising · trend:up* Give your agents the power of the Hugging Face ecosystem **Why it matters.** Hugging Face Skills offers a collection of standardized definitions and scripts for AI agents to perform tasks like dataset creation, model training, and evaluation using the Hugging Face ecosystem, making them interoperable with tools like Codex and Claude Code. This matters now because the growing adoption of AI agents in development workflows demands reusable, plug-and-play components to reduce custom coding efforts and enhance productivity, though its lack of formal releases may indicate ongoing refinements rather than full stability. _themes: agents · skills · ml-tasks · huggingface_ #### [openai/symphony](https://github.com/openai/symphony) *Elixir · ★15,259 · Apache-2.0 · experimental · score:0.70 · hot:0.68 · rising:0.72 · durable:0.70 · board:rising · trend:up* Symphony turns project work into isolated, autonomous implementation runs, allowing teams to manage work instead of supervising coding agents. **Why it matters.** Symphony automates software development by converting project tasks into isolated runs managed by autonomous agents, reducing the need for direct supervision of coding processes. It integrates with tools like Linear for task monitoring and provides outputs such as CI status and PR reviews, enabling teams to focus on higher-level management. This matters now amid growing AI agent adoption, as it could streamline workflows but remains unproven and experimental. _themes: agents · automation · devops · ci-cd_ #### [openai/swarm](https://github.com/openai/swarm) *Python · ★21,344 · MIT · experimental · score:0.40 · hot:0.68 · rising:0.71 · durable:0.65 · board:rising · trend:up* Educational framework exploring ergonomic, lightweight multi-agent orchestration. Managed by OpenAI Solution team. **Why it matters.** Swarm is an experimental Python framework from OpenAI for orchestrating multiple AI agents through simple abstractions like agents and handoffs, focusing on lightweight coordination and testing. It matters for educational purposes as it provides a basic way to experiment with multi-agent systems, but it's being replaced by the more advanced OpenAI Agents SDK, making it less relevant for current production or development needs. _themes: agents · orchestration · multi-agent_ #### [thinkwee/AgentsMeetRL](https://github.com/thinkwee/AgentsMeetRL) *HTML · ★958 · no-license · beta · score:0.70 · hot:0.67 · rising:0.64 · durable:0.62 · board:hot · trend:stable* Awesome List for Agentic RL **Why it matters.** This repository is a curated awesome list of open-source projects that use reinforcement learning to train large language model agents, focusing on aspects like RL frameworks, algorithms, rewards, and environments for multi-turn interactions or tool use. It matters right now because the rapid advancement of AI agents and LLMs requires accessible resources for integrating RL, helping researchers and developers explore and build upon existing work in this emerging field. _themes: agents · rl · llm · multiagent_ #### [microsoft/agent-lightning](https://github.com/microsoft/agent-lightning) *Python · ★16,950 · MIT · beta · score:0.75 · hot:0.65 · rising:0.69 · durable:0.79 · board:durable · trend:stable* The absolute trainer to light up AI agents. **Why it matters.** Agent Lightning is a Python library designed to optimize and train AI agents using techniques like reinforcement learning and prompt optimization, with compatibility across various frameworks like LangChain and AutoGen, requiring minimal code changes. It matters right now because the rapid evolution of AI agents demands efficient training tools to handle multi-agent systems and LLM-based workflows, potentially accelerating development in a competitive field, though its early version suggests it may still have stability issues. _themes: agents · llm · reinforcement-learning · fine-tuning_ #### [microsoft/TypeAgent](https://github.com/microsoft/TypeAgent) *TypeScript · ★683 · MIT · experimental · score:0.70 · hot:0.65 · rising:0.65 · durable:0.58 · board:hot · trend:stable* Sample code that explores an architecture for using language models to build a personal agent that can work with application agents. **Why it matters.** TypeAgent is sample code that demonstrates an architecture for building a personal agent using language models, focusing on principles like distilling models into logical structures, controlling information density, and enabling collaboration between LLMs and traditional software. It matters now because the rapid growth of AI agents requires safer and more efficient integration of stochastic systems like LLMs with deterministic components, potentially addressing challenges in precision, recall, and human-AI interaction in real-world applications. _themes: agents · llms · architecture · collaboration_ #### [facebookresearch/HyperAgents](https://github.com/facebookresearch/HyperAgents) *Python · ★2,345 · NOASSERTION · experimental · score:0.60 · hot:0.64 · rising:0.64 · durable:0.59 · board:hot · trend:stable* Self-referential self-improving agents that can optimize for any computable task **Why it matters.** HyperAgents is a research project from Facebook that implements self-referential agents using external LLMs to generate and execute code for self-improvement and task optimization, though it's limited by dependencies on proprietary APIs and lacks robust safety measures. It matters now because the growing interest in autonomous AI agents could advance task automation, but its experimental nature and potential security risks underscore the need for cautious development in this area. _themes: agents · self-improving · llm · code-generation_ #### [allenai/molmoweb](https://github.com/allenai/molmoweb) *Python · ★527 · Apache-2.0 · experimental · score:0.70 · hot:0.60 · rising:0.61 · durable:0.61 · board:durable · trend:stable* **Why it matters.** MolmoWeb is an open-source multimodal agent that uses AI to autonomously interact with web browsers via natural language commands, performing actions like clicking and scrolling, as demonstrated in its associated paper and demo. It matters now because it contributes to the growing field of AI agents for web automation, potentially aiding in tasks like accessibility and data extraction, but its lack of a formal release and experimental status mean it's not yet ready for widespread production use. _themes: agents · multimodal · inference · fine-tuning_ #### [QwenLM/Qwen-Agent](https://github.com/QwenLM/Qwen-Agent) *Python · ★16,098 · Apache-2.0 · beta · score:0.80 · hot:0.51 · rising:0.58 · durable:0.73 · board:durable · trend:stable* Agent framework and applications built upon Qwen>=3.0, featuring Function Calling, MCP, Code Interpreter, RAG, Chrome extension, etc. **Why it matters.** Qwen-Agent is a Python framework for building AI agents using Qwen language models, offering features like function calling, code interpretation, RAG, and tools for planning and memory management. It matters now because the growing demand for agentic AI in applications requires accessible frameworks, and this one integrates with popular Qwen models while providing benchmarks and examples, though its early version (v0.0.26) suggests potential stability issues. _themes: agents · rag · tool-usage · llm-framework_ #### [microsoft/JARVIS](https://github.com/microsoft/JARVIS) *Python · ★24,653 · MIT · experimental · score:0.75 · hot:0.50 · rising:0.55 · durable:0.64 · board:durable · trend:stable* JARVIS, a system to connect LLMs with ML community. Paper: https://arxiv.org/pdf/2303.17580.pdf **Why it matters.** JARVIS is a platform that connects large language models with tools and services to enable task automation, benchmarking, and AGI research, as outlined in its paper. It matters now because it addresses the need for practical integration and evaluation of LLMs in real-world applications, especially with the rise of AI agents, and includes recent additions like Easytool for simplified tool usage and TaskBench for assessing LLM capabilities. _themes: agents · llms · benchmarking · tools_ #### [facebookresearch/drzero](https://github.com/facebookresearch/drzero) *Python · ★508 · NOASSERTION · experimental · score:0.70 · hot:0.49 · rising:0.50 · durable:0.58 · board:durable · trend:stable* Dr. Zero Self-Evolving Search Agents without Training Data **Why it matters.** Dr. Zero is a framework that allows search agents to self-evolve without any training data by using a feedback loop where a proposer generates tasks and a solver learns from them iteratively, incorporating techniques like hop-grouped relative policy optimization for efficiency. This matters now because it addresses the challenges of data scarcity and high computational costs in AI development, potentially enabling more autonomous and scalable agent training in resource-constrained environments. However, its reliance on experimental methods and lack of a formal release raise questions about practical applicability and reproducibility. _themes: agents · self-evolution · reinforcement-learning · optimization_ #### [openai/Video-Pre-Training](https://github.com/openai/Video-Pre-Training) *Python · ★1,677 · MIT · beta · score:0.75 · hot:0.47 · rising:0.54 · durable:0.70 · board:durable · trend:down* Video PreTraining (VPT): Learning to Act by Watching Unlabeled Online Videos **Why it matters.** This repository provides code and models for Video PreTraining (VPT), enabling AI agents to learn actions from unlabeled online videos, as demonstrated in the Minecraft environment. It advances unsupervised reinforcement learning by reducing the need for labeled data, which is relevant for research in scalable AI training, though its practical impact is limited to specific simulation domains like gaming. However, the pinned dependencies and potential compatibility issues may hinder broader adoption. _themes: reinforcement-learning · agents · video-pretraining · unsupervised-learning_ #### [google-deepmind/scalable_agent](https://github.com/google-deepmind/scalable_agent) *Python · ★1,023 · Apache-2.0 · experimental · score:0.60 · hot:0.43 · rising:0.47 · durable:0.55 · board:durable · trend:stable* A TensorFlow implementation of Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures. **Why it matters.** This repository provides a TensorFlow implementation of the IMPALA algorithm for scalable distributed deep reinforcement learning, enabling efficient training of agents across multiple machines using importance weighted actor-learner architectures. It matters now for researchers exploring foundational RL techniques, as it demonstrates key concepts in distributed training, though its 2018 origins and lack of updates mean it's more relevant for educational purposes than cutting-edge applications. _themes: reinforcement-learning · distributed-training · deep-rl · tensorflow_ #### [allenai/visprog](https://github.com/allenai/visprog) *Python · ★766 · Apache-2.0 · experimental · score:0.80 · hot:0.43 · rising:0.46 · durable:0.57 · board:durable · trend:down* Official code for VisProg (CVPR 2023 Best Paper!) **Why it matters.** VisProg is a framework for compositional visual reasoning that enables writing and executing programs for visual tasks without additional training, as demonstrated in the CVPR 2023 Best Paper. It matters now because it advances AI's ability to handle complex visual queries through programmatic interfaces, though its limitations are addressed in the follow-up CodeNav, which improves code generation and execution iteratively. This evolution highlights ongoing needs in multimodal AI for more robust, error-handling agents. _themes: visual-reasoning · code-generation · agents · multimodal_ #### [google-research/dreamer](https://github.com/google-research/dreamer) *Python · ★714 · Apache-2.0 · archived · score:0.60 · hot:0.43 · rising:0.45 · durable:0.54 · board:durable · trend:down* Dream to Control: Learning Behaviors by Latent Imagination **Why it matters.** This repository implements the original Dreamer agent, a reinforcement learning model that learns behaviors by imagining trajectories in a latent space, enabling efficient learning from images with fewer episodes than model-free alternatives. However, it has been largely superseded by a newer, simpler TensorFlow 2 implementation, making this version more relevant for historical reference or educational purposes rather than active development. Its core ideas remain influential in model-based RL research. _themes: rl · model-based · latent-imagination · control_ #### [microsoft/DeepVideoDiscovery](https://github.com/microsoft/DeepVideoDiscovery) *Python · ★377 · MIT · experimental · score:0.70 · hot:0.41 · rising:0.41 · durable:0.59 · board:durable · trend:down* **Deep Video Discovery (DVD)** is a deep-research style question answering agent designed for understanding extra-long videos. **Why it matters.** Deep Video Discovery is an AI agent that uses tool-based search to answer questions about extra-long videos, leveraging techniques from a recent research paper to achieve state-of-the-art results on benchmarks like LVBench. It matters now because video content is proliferating, and effective long-video understanding could enhance applications in education, analysis, and entertainment, though its experimental nature and lack of formal releases mean it's still unproven for production use. _themes: agents · video-processing · video-understanding · rag_ ### data (26) #### [google-deepmind/formal-conjectures](https://github.com/google-deepmind/formal-conjectures) *Lean · ★927 · Apache-2.0 · experimental · score:0.70 · hot:0.63 · rising:0.65 · durable:0.55 · board:rising · trend:up* A collection of formalized statements of conjectures in Lean. **Why it matters.** This repository collects and formalizes statements of mathematical conjectures in the Lean proof assistant, without including proofs, to serve as benchmarks for automated theorem provers and to clarify conjecture meanings. It matters right now because the growing interest in AI-assisted formal mathematics, such as tools like AlphaProof, highlights the need for high-quality datasets to evaluate and improve these systems, while also encouraging community contributions to expand formal libraries. _themes: formal-mathematics · lean · conjectures · automated-proving_ #### [facebookresearch/Ego4d](https://github.com/facebookresearch/Ego4d) *Jupyter Notebook · ★579 · MIT · beta · score:0.75 · hot:0.61 · rising:0.65 · durable:0.61 · board:rising · trend:stable* Ego4d dataset repository. Download the dataset, visualize, extract features & example usage of the dataset **Why it matters.** This repository provides access to the Ego4D and Ego-Exo4D datasets, which are large-scale collections of egocentric and multi-view videos with annotations for computer vision tasks, including tools for downloading, visualizing, and extracting features. It matters now because recent updates with expanded annotations and data volumes support advancing research in video understanding and perception, amid growing needs for real-world AI applications in robotics and embodied AI. _themes: computer-vision · video-dataset · feature-extraction · visualization_ #### [google-deepmind/materials_discovery](https://github.com/google-deepmind/materials_discovery) *Jupyter Notebook · ★1,161 · Apache-2.0 · experimental · score:0.75 · hot:0.51 · rising:0.55 · durable:0.56 · board:durable · trend:stable* **Why it matters.** This repository from Google DeepMind shares a dataset of over 520,000 stable materials discovered using graph networks for materials science, aiming to accelerate research in areas like batteries and microchips. It matters now because it provides a large-scale resource for exploring novel inorganic crystals, potentially enabling breakthroughs in technology, though it's still a research project with possible limitations and no formal releases. _themes: materials-science · graph-networks · machine-learning · dataset_ #### [google-research/language-table](https://github.com/google-research/language-table) *Jupyter Notebook · ★356 · Apache-2.0 · experimental · score:0.65 · hot:0.51 · rising:0.52 · durable:0.49 · board:rising · trend:stable* Suite of human-collected datasets and a multi-task continuous control benchmark for open vocabulary visuolinguomotor learning. **Why it matters.** This repo provides a collection of human-annotated datasets and a benchmark for training AI systems that integrate vision, language, and motor control in robotics simulations, focusing on open-vocabulary tasks. It matters now because advancements in multimodal AI and embodied robotics are accelerating, offering researchers tools to develop more natural language-driven robotic agents, though its lack of formal releases and modest adoption (356 stars) suggests it's still early in development and not yet widely validated. _themes: robotics · multimodal · benchmarks · simulation_ #### [allenai/dolma](https://github.com/allenai/dolma) *Python · ★1,483 · Apache-2.0 · production · score:0.85 · hot:0.47 · rising:0.55 · durable:0.70 · board:durable · trend:down* Data and tools for generating and inspecting OLMo pre-training data. **Why it matters.** Dolma provides a massive open dataset of 3 trillion tokens sourced from diverse materials for training large language models, along with a high-performance toolkit for curating and processing such datasets. It matters now because the AI community is grappling with the need for high-quality, ethically sourced data to advance open LLMs, especially as proprietary datasets dominate and concerns about data diversity and licensing grow. _themes: llm · data-processing · nlp · pre-training_ #### [google-research/arco-era5](https://github.com/google-research/arco-era5) *Python · ★465 · Apache-2.0 · beta · score:0.70 · hot:0.47 · rising:0.51 · durable:0.55 · board:durable · trend:stable* Recipes for reproducing Analysis-Ready & Cloud Optimized (ARCO) ERA5 datasets. **Why it matters.** This repository provides Python recipes and Colab notebooks to reproduce analysis-ready and cloud-optimized versions of the ERA5 climate dataset, making it easier to access and process massive historical weather data in the cloud. It matters now as climate research accelerates due to global warming concerns, and cloud-optimized formats enable scalable analysis without large downloads, supporting open science initiatives in meteorology and environmental AI. _themes: climate-data · cloud-optimized · zarr · open-science_ #### [facebookresearch/uco3d](https://github.com/facebookresearch/uco3d) *Python · ★1,319 · CC-BY-4.0 · experimental · score:0.70 · hot:0.45 · rising:0.53 · durable:0.62 · board:durable · trend:stable* Uncommon Objects in 3D dataset **Why it matters.** uCO3D is a dataset offering over 170,000 turn-table videos of uncommon objects from the LVIS taxonomy, complete with annotations like segmentation, camera poses, point clouds, and 3D Gaussian splat reconstructions, making it useful for computer vision tasks. It matters now as the AI community increasingly needs diverse 3D data for applications in robotics and AR/VR, providing an improvement over predecessors like CO3Dv2 by releasing full videos and enhanced annotations, though its massive 19.3 TB size poses practical challenges for adoption. _themes: 3d-vision · datasets · computer-vision · point-clouds_ #### [openai/prm800k](https://github.com/openai/prm800k) *Python · ★2,117 · MIT · experimental · score:0.70 · hot:0.45 · rising:0.52 · durable:0.62 · board:durable · trend:stable* 800,000 step-level correctness labels on LLM solutions to MATH problems **Why it matters.** This repository provides the PRM800K dataset, containing 800,000 step-level correctness labels for LLM-generated solutions to math problems from the MATH dataset, enabling process supervision to improve mathematical reasoning in AI models. It matters now because ongoing research in LLMs highlights the need for better step-by-step verification to enhance accuracy in complex tasks like math, and this dataset supports fine-tuning efforts amid increasing demands for reliable AI. The release includes raw labels and labeling instructions, as detailed in an accompanying paper, making it a practical resource for advancing model evaluation. _themes: dataset · process-supervision · math-reasoning · llm_ #### [google-deepmind/mathematics_dataset](https://github.com/google-deepmind/mathematics_dataset) *Python · ★1,951 · Apache-2.0 · production · score:0.80 · hot:0.44 · rising:0.54 · durable:0.66 · board:durable · trend:stable* This dataset code generates mathematical question and answer pairs, from a range of question types at roughly school-level difficulty. **Why it matters.** This repository provides Python code to generate a dataset of mathematical question-answer pairs focused on school-level algebra and reasoning, primarily for evaluating AI models' capabilities in mathematical learning. It matters now because, amid rapid advancements in AI, robust benchmarks for reasoning skills are essential to identify limitations in models used for educational tools or problem-solving applications, though the dataset hasn't seen recent updates since its initial release. _themes: math-reasoning · dataset · ai-evaluation · algebra_ #### [allenai/natural-instructions](https://github.com/allenai/natural-instructions) *Python · ★1,041 · Apache-2.0 · beta · score:0.75 · hot:0.44 · rising:0.51 · durable:0.63 · board:durable · trend:down* Expanding natural instructions **Why it matters.** This repository maintains a community-driven collection of over 1,500 natural language instructions for various NLP tasks, aiming to improve model generalization to unseen tasks by defining them in everyday language rather than task-specific labels. It matters now because the rise of instruction-tuned language models highlights the need for such datasets to advance zero-shot and few-shot learning, though its impact is limited by the quality and coverage of community contributions. _themes: nlp · instructions · zero-shot · datasets_ #### [allenai/objaverse-xl](https://github.com/allenai/objaverse-xl) *Python · ★1,264 · Apache-2.0 · beta · score:0.80 · hot:0.43 · rising:0.51 · durable:0.64 · board:durable · trend:down* 🪐 Objaverse-XL is a Universe of 10M+ 3D Objects. Contains API Scripts for Downloading and Processing! **Why it matters.** This repository provides scripts to download and process Objaverse-XL, a massive dataset of over 10 million 3D objects, which is used to train advanced 3D foundation models like Zero123-XL for tasks such as novel view synthesis and image-to-3D generation. It matters right now because the growing demand for scalable 3D data in AI research is driving innovations in computer vision and generative models, enabling better generalization in real-world applications like AR/VR and content creation, though its scale also raises concerns about data quality and processing efficiency. _themes: 3d · computer-vision · image-to-3d · text-to-3d_ #### [facebookresearch/Replica-Dataset](https://github.com/facebookresearch/Replica-Dataset) *C++ · ★1,249 · NOASSERTION · production · score:0.75 · hot:0.43 · rising:0.51 · durable:0.62 · board:durable · trend:down* The Replica Dataset v1 as published in https://arxiv.org/abs/1906.05797 . **Why it matters.** The Replica Dataset offers high-quality 3D reconstructions of indoor spaces with detailed geometry, textures, and segmentation, along with an SDK for visualization and rendering, making it suitable for computer vision and AI training tasks. It matters now because it supports embodied AI research by integrating with frameworks like AI Habitat, which is relevant amid growing interest in simulated environments for agent training, though its 2019 origins mean it may lack updates compared to newer datasets. _themes: 3d-reconstruction · computer-vision · semantic-segmentation · ai-training_ #### [openai/gpt-2-output-dataset](https://github.com/openai/gpt-2-output-dataset) *Python · ★2,027 · MIT · beta · score:0.70 · hot:0.42 · rising:0.51 · durable:0.60 · board:durable · trend:stable* Dataset of GPT-2 outputs for research in detection, biases, and more **Why it matters.** This repository provides a dataset of outputs from various GPT-2 models, including 250K samples per model with different configurations, aimed at research in AI-generated text detection, biases, and ethical issues. It matters now as the proliferation of generative AI raises concerns about misinformation and content authenticity, making tools for detection and bias analysis increasingly vital for AI safety research and applications in content moderation. _themes: text-generation · bias-detection · ai-research · datasets_ #### [deepseek-ai/profile-data](https://github.com/deepseek-ai/profile-data) *? · ★1,149 · no-license · experimental · score:0.70 · hot:0.42 · rising:0.47 · durable:0.57 · board:durable · trend:down* Analyze computation-communication overlap in V3/R1. **Why it matters.** This repository shares profiling data from DeepSeek's AI training and inference framework, focusing on strategies for overlapping computation and communication to improve efficiency in distributed setups. It matters now because optimizing these overlaps is critical for scaling large language models amid growing computational demands, providing valuable insights for developers working on similar systems. Such data promotes transparency and can help accelerate advancements in AI infrastructure. _themes: profiling · distributed-training · inference · optimization_ #### [EleutherAI/the-pile](https://github.com/EleutherAI/the-pile) *Python · ★1,644 · MIT · beta · score:0.70 · hot:0.42 · rising:0.49 · durable:0.58 · board:durable · trend:down* **Why it matters.** This repository offers code for replicating or creating variants of The Pile, a large-scale dataset used for training language models, but it does not provide the dataset itself, directing users elsewhere for that. It matters for researchers needing to customize datasets for AI experiments, though its utility is limited by the lack of recent updates and focus on replication rather than direct use, in a field where dataset quality and availability are critical for advancing model generalization. _themes: dataset · language-modeling · replication · open-source_ #### [google-deepmind/open_x_embodiment](https://github.com/google-deepmind/open_x_embodiment) *Jupyter Notebook · ★1,783 · Apache-2.0 · experimental · score:0.80 · hot:0.42 · rising:0.47 · durable:0.56 · board:durable · trend:down* **Why it matters.** This repository from Google DeepMind unifies various open-sourced robotic datasets into a standardized format for easy access and use in training models, including providing a model checkpoint for RT-1-X. It matters now because the robotics and AI fields are experiencing rapid growth in embodied intelligence, and this standardization accelerates research by simplifying data integration, especially amid increasing demands for open, reusable datasets. _themes: robotics · datasets · rl · embodied-ai_ #### [allenai/s2orc](https://github.com/allenai/s2orc) *Python · ★1,042 · no-license · production · score:0.80 · hot:0.41 · rising:0.47 · durable:0.57 · board:durable · trend:down* S2ORC: The Semantic Scholar Open Research Corpus: https://www.aclweb.org/anthology/2020.acl-main.447/ **Why it matters.** S2ORC is a large corpus of scientific papers designed for NLP and text mining research, containing structured data from Semantic Scholar. It matters right now because it's actively maintained through the Semantic Scholar API, providing ongoing access to new papers for real-time research needs. This makes it a reliable resource amid growing demands for high-quality scientific datasets in AI. _themes: nlp · text-mining · scientific-corpus · dataset_ #### [facebookresearch/seamless_interaction](https://github.com/facebookresearch/seamless_interaction) *Python · ★370 · NOASSERTION · experimental · score:0.70 · hot:0.40 · rising:0.44 · durable:0.54 · board:durable · trend:down* Foundation Models and Data for Human-Human and Human-AI interactions. **Why it matters.** This repository offers a large-scale multimodal dataset with over 4,000 hours of human interaction footage, including verbal and nonverbal signals, to support AI research on human-AI and human-human communication. It matters now because the growing demand for advanced virtual agents and natural interaction systems requires high-quality data to improve AI's understanding of emotions and contexts, especially in applications like telepresence and embodied AI. _themes: multimodal · dataset · human-interaction · ai-research_ #### [google-deepmind/dsprites-dataset](https://github.com/google-deepmind/dsprites-dataset) *Jupyter Notebook · ★532 · Apache-2.0 · production · score:0.70 · hot:0.40 · rising:0.48 · durable:0.60 · board:durable · trend:down* Dataset to assess the disentanglement properties of unsupervised learning methods **Why it matters.** This repository provides the dSprites dataset, a collection of procedurally generated 2D shapes designed to evaluate how well unsupervised learning models can disentangle latent factors like shape and position. It matters for ongoing AI research on interpretable representations, though its 2017 origin means it's foundational rather than cutting-edge, with limited updates potentially making it less relevant for current high-resolution or dynamic applications. _themes: disentanglement · vae · dataset · unsupervised-learning_ #### [google-deepmind/narrativeqa](https://github.com/google-deepmind/narrativeqa) *Shell · ★509 · Apache-2.0 · production · score:0.65 · hot:0.40 · rising:0.48 · durable:0.60 · board:durable · trend:down* This repository contains the NarrativeQA dataset. It includes the list of documents with Wikipedia summaries, links to full stories, and questions and answers. **Why it matters.** This repository provides the NarrativeQA dataset, which consists of stories from books and movies with accompanying Wikipedia summaries, questions, and answers, designed for evaluating reading comprehension in AI models. It matters because it's a foundational benchmark for narrative understanding tasks in NLP, though its 2018 origins mean it lacks integration with modern large language models and may not fully address current challenges like multimodal data. _themes: qa · reading-comprehension · nlp · dataset_ #### [huggingface/cosmopedia](https://github.com/huggingface/cosmopedia) *Python · ★567 · Apache-2.0 · experimental · score:0.75 · hot:0.40 · rising:0.45 · durable:0.54 · board:durable · trend:down* **Why it matters.** This repository contains code for generating Cosmopedia, a large synthetic dataset with over 25 billion tokens created using Mixtral-8x7B, covering topics like textbooks and blog posts to mimic web knowledge. It matters now because synthetic data is increasingly vital for AI research to address real-world data shortages and biases, though its v0.1 status suggests it's still rough around the edges and not yet optimized for broad use. _themes: synthetic-data · dataset-generation · llm · deduplication_ #### [facebookresearch/ContactPose](https://github.com/facebookresearch/ContactPose) *Jupyter Notebook · ★406 · MIT · beta · score:0.70 · hot:0.39 · rising:0.46 · durable:0.56 · board:durable · trend:down* Large dataset of hand-object contact, hand- and object-pose, and 2.9 M RGB-D grasp images. **Why it matters.** This repository provides a large dataset of hand-object interactions, including 2.9 million RGB-D images with annotations for hand pose, object pose, and contact points, aimed at advancing computer vision and robotics research. It matters now because accurate hand-object modeling is essential for real-world applications like robotic grasping and augmented reality, especially as AI-driven manipulation tasks gain prominence, though accessibility issues with the data could hinder its immediate utility. _themes: computer-vision · dataset · robotics · grasp-detection_ #### [allenai/mmc4](https://github.com/allenai/mmc4) *Python · ★955 · MIT · beta · score:0.70 · hot:0.39 · rising:0.44 · durable:0.60 · board:durable · trend:down* MultimodalC4 is a multimodal extension of c4 that interleaves millions of images with text. **Why it matters.** MMC4 is a large-scale dataset that extends the C4 corpus by integrating millions of images with text, enabling training and evaluation of multimodal AI models. It matters now because the growing focus on vision-language models requires diverse, high-quality datasets, and this one provides substantial resources despite recent availability issues from accidental deletions. _themes: multimodal · dataset · vision-language · large-scale_ #### [allenai/scitldr](https://github.com/allenai/scitldr) *Python · ★758 · Apache-2.0 · experimental · score:0.70 · hot:0.38 · rising:0.44 · durable:0.54 · board:durable · trend:down* **Why it matters.** SciTLDR offers a dataset, pre-trained models, and code for extreme summarization of scientific documents, based on a 2020 paper, focusing on generating TLDRs from research papers. It matters now as the need for efficient scientific literature summarization grows with increasing publication volumes, aiding researchers in quick comprehension, but its age and lack of updates mean it may not incorporate recent NLP advancements. The repo provides valuable resources for experimentation, though reliance on external libraries like Fairseq adds setup complexity. _themes: summarization · nlp · datasets · scientific-documents_ #### [google-research/sound-separation](https://github.com/google-research/sound-separation) *Python · ★707 · Apache-2.0 · experimental · score:0.70 · hot:0.38 · rising:0.44 · durable:0.53 · board:durable · trend:down* **Why it matters.** This repository provides open-source datasets and deep learning models for sound separation, focusing on tasks like separating mixed audio sources using techniques such as mixture-invariant training. It matters right now because sound separation is crucial for advancing AI applications in audio processing, like improving speech recognition and multimedia analysis, especially with the growing demand for unsupervised methods in resource-scarce environments. _themes: audio · separation · deep-learning · unsupervised_ #### [facebookresearch/Action100M](https://github.com/facebookresearch/Action100M) *Python · ★447 · NOASSERTION · experimental · score:0.70 · hot:0.35 · rising:0.36 · durable:0.54 · board:durable · trend:down* A Large-scale Video Action Dataset **Why it matters.** Action100M is a large-scale dataset for video action recognition, containing annotations for training and evaluating AI models on real-world video data. It matters right now because the growing demand for advanced video understanding in AI applications, such as autonomous systems and content analysis, requires diverse and scalable datasets to improve model accuracy and generalization. _themes: video · dataset · action-recognition · computer-vision_ ### demo (64) #### [Shubhamsaboo/awesome-llm-apps](https://github.com/Shubhamsaboo/awesome-llm-apps) *Python · ★106,397 · Apache-2.0 · beta · score:0.85 · hot:0.76 · rising:0.82 · durable:0.77 · board:rising · trend:up* 100+ AI Agent & RAG apps you can actually run — clone, customize, ship. **Why it matters.** This repository offers over 100 ready-to-run templates for AI agents and RAG applications, allowing users to clone, customize, and deploy them with minimal setup, covering areas like multi-agent systems and fine-tuning. It matters now because the rapid proliferation of LLMs demands efficient prototyping tools to reduce boilerplate code, enabling developers to iterate faster in a competitive AI landscape. _themes: rag · agents · llms · fine-tuning_ #### [meta-llama/llama-cookbook](https://github.com/meta-llama/llama-cookbook) *Jupyter Notebook · ★18,292 · MIT · beta · score:0.85 · hot:0.73 · rising:0.80 · durable:0.79 · board:rising · trend:up* Welcome to the Llama Cookbook! This is your go to guide for Building with Llama: Getting started with Inference, Fine-Tuning, RAG. We also show you how to solve end to end problems using Llama model family and using them on various provider services **Why it matters.** This repository provides a collection of Jupyter notebooks and guides for getting started with Meta's Llama family of language models, covering inference, fine-tuning, RAG, and end-to-end use cases across various providers. It matters now because the rapid adoption of LLMs in AI development requires accessible, practical resources, and this cookbook helps developers quickly implement Llama models amid ongoing advancements like Llama 4. _themes: inference · fine-tuning · rag · llama_ #### [openai/openai-cookbook](https://github.com/openai/openai-cookbook) *Jupyter Notebook · ★72,830 · MIT · production · score:0.90 · hot:0.73 · rising:0.83 · durable:0.78 · board:rising · trend:up* Examples and guides for using the OpenAI API **Why it matters.** This repository provides a collection of examples and guides for using the OpenAI API, including tasks like text generation and chat interactions, primarily through Jupyter Notebooks in Python. It matters right now because the rapid adoption of generative AI models like GPT-4 requires accessible resources for developers to integrate these APIs quickly, helping bridge the gap between theoretical AI capabilities and practical application amid ongoing advancements in the field. _themes: inference · api · chat · fine-tuning_ #### [microsoft/generative-ai-for-beginners](https://github.com/microsoft/generative-ai-for-beginners) *Jupyter Notebook · ★109,513 · MIT · production · score:0.80 · hot:0.73 · rising:0.81 · durable:0.81 · board:rising · trend:up* 21 Lessons, Get Started Building with Generative AI **Why it matters.** This repository provides 21 Jupyter Notebook lessons aimed at teaching beginners the basics of generative AI, including tools like ChatGPT and DALL-E, as well as concepts like prompt engineering and semantic search. It matters now because the rapid growth of generative AI demands accessible entry-level resources to build foundational skills, especially amid increasing industry adoption and the need for diverse language support to reach a global audience. _themes: generative-ai · prompt-engineering · llms · transformers_ #### [microsoft/WinUI-Gallery](https://github.com/microsoft/WinUI-Gallery) *C# · ★3,448 · MIT · production · score:0.80 · hot:0.73 · rising:0.80 · durable:0.71 · board:rising · trend:up* This app demonstrates the controls available in WinUI and the Fluent Design System. **Why it matters.** This repository provides an interactive gallery app that demonstrates WinUI 3 controls and the Fluent Design System, offering code samples and guidance for developers building modern Windows applications. It matters right now because Microsoft is actively evolving the Windows App SDK, making it essential for Windows app developers to understand these UI components for creating responsive and accessible interfaces. Additionally, its open contribution model supports community involvement, especially during events like Hacktoberfest. _themes: ui-design · windows-dev · xaml · fluent-design_ #### [genieincodebottle/generative-ai](https://github.com/genieincodebottle/generative-ai) *Jupyter Notebook · ★2,222 · MIT · beta · score:0.60 · hot:0.70 · rising:0.70 · durable:0.66 · board:hot · trend:up* Comprehensive resources on Generative AI, including a detailed roadmap, projects, use cases, interview preparation, and coding preparation. **Why it matters.** This repository offers a collection of resources, roadmaps, projects, and preparation materials for Generative AI, focusing on learning, use cases, and interview readiness, primarily through Jupyter Notebooks. It matters now because Generative AI is a rapidly expanding field with high demand for accessible educational content, though its value is limited by potential outdated materials and lack of formal releases. _themes: generative-ai · agents · llm-evaluation · interview-prep_ #### [microsoft/ai-agents-for-beginners](https://github.com/microsoft/ai-agents-for-beginners) *Jupyter Notebook · ★56,967 · MIT · beta · score:0.85 · hot:0.70 · rising:0.77 · durable:0.77 · board:rising · trend:up* 12 Lessons to Get Started Building AI Agents **Why it matters.** This repository provides a free, beginner-oriented course with 12 lessons using Jupyter Notebooks to teach the basics of building AI agents with tools like Autogen and Semantic Kernel. It matters right now because AI agents are rapidly evolving in the generative AI landscape, and this resource helps democratize access for newcomers, potentially accelerating adoption amid growing interest in agentic frameworks. _themes: agents · rag · generative-ai · autogen_ #### [anthropics/claude-cookbooks](https://github.com/anthropics/claude-cookbooks) *Jupyter Notebook · ★41,065 · MIT · beta · score:0.85 · hot:0.69 · rising:0.79 · durable:0.73 · board:rising · trend:up* A collection of notebooks/recipes showcasing some fun and effective ways of using Claude. **Why it matters.** This repository is a collection of Jupyter notebooks and recipes that demonstrate practical ways to use Anthropic's Claude AI model via its API, providing copyable code for developers to integrate into their projects. It matters right now because the rapid adoption of AI assistants like Claude requires accessible resources for quick learning and prototyping, and with over 40,000 stars, it serves as a valuable, community-driven guide amid the growing AI development landscape. _themes: api · notebooks · ai-assistants · prompt-engineering_ #### [microsoft/mcp-for-beginners](https://github.com/microsoft/mcp-for-beginners) *Jupyter Notebook · ★15,900 · MIT · production · score:0.70 · hot:0.69 · rising:0.78 · durable:0.73 · board:rising · trend:up* This open-source curriculum introduces the fundamentals of Model Context Protocol (MCP) through real-world, cross-language examples in .NET, Java, TypeScript, JavaScript, Rust and Python. Designed for developers, it focuses on practical techniques for building modular, scalable, and secure AI workflows from session setup to service orchestration. **Why it matters.** This repository provides an open-source curriculum for learning the Model Context Protocol (MCP), offering practical, cross-language examples in languages like .NET, Java, TypeScript, JavaScript, Rust, and Python to teach developers how to build modular, scalable, and secure AI workflows. It matters now because the rapid growth of AI development demands standardized protocols for interoperability, and this resource helps beginners quickly grasp these concepts amid increasing demand for secure, multi-language AI systems, especially with Microsoft's backing and global accessibility through translations. _themes: mcp · ai-workflows · security · cross-language_ #### [microsoft/PhiCookBook](https://github.com/microsoft/PhiCookBook) *Jupyter Notebook · ★3,730 · MIT · beta · score:0.70 · hot:0.68 · rising:0.73 · durable:0.65 · board:rising · trend:up* This is a Phi Family of SLMs book for getting started with Phi Models. Phi a family of open sourced AI models developed by Microsoft. Phi models are the most capable and cost-effective small language models (SLMs) available, outperforming models of the same size and next size up across a variety of language, reasoning, coding, and math benchmarks **Why it matters.** The Phi Cookbook provides hands-on examples and tutorials for Microsoft's Phi family of small language models, focusing on deployment, inference, and applications in areas like reasoning, coding, and multimodal tasks. It matters right now because there's growing interest in efficient, cost-effective SLMs for edge devices and resource-limited environments, and this resource helps developers quickly experiment with Phi models amid the competitive landscape of open-source AI. Microsoft's backing adds credibility, but the lack of a formal release means it's still evolving. _themes: slm · inference · multimodal · cookbook_ #### [microsoft/ML-For-Beginners](https://github.com/microsoft/ML-For-Beginners) *Jupyter Notebook · ★85,296 · MIT · production · score:0.80 · hot:0.68 · rising:0.79 · durable:0.76 · board:rising · trend:up* 12 weeks, 26 lessons, 52 quizzes, classic Machine Learning for all **Why it matters.** This repository offers a structured 12-week curriculum with 26 lessons and 52 quizzes to teach fundamental machine learning concepts using Python and R, primarily through Jupyter Notebooks, and includes multi-language support for broader accessibility. It matters now because the growing demand for AI skills requires accessible, high-quality educational resources, and this repo fills that gap by providing free, community-maintained content that helps beginners build foundational knowledge amid the AI talent shortage. _themes: machine-learning · education · jupyter · beginners_ #### [microsoft/Data-Science-For-Beginners](https://github.com/microsoft/Data-Science-For-Beginners) *Jupyter Notebook · ★34,860 · MIT · production · score:0.85 · hot:0.68 · rising:0.78 · durable:0.75 · board:rising · trend:up* 10 Weeks, 20 Lessons, Data Science for All! **Why it matters.** This repository provides a structured 10-week curriculum for beginners to learn data science basics through 20 hands-on lessons using Python, Pandas, and data visualization tools, including quizzes and assignments. It matters right now because the growing demand for data science skills in AI and business analytics makes accessible, free educational resources essential for entry-level learners to build foundational knowledge and apply it practically. _themes: data-science · python · beginners · data-analysis_ #### [huggingface/diffusion-models-class](https://github.com/huggingface/diffusion-models-class) *Jupyter Notebook · ★4,311 · Apache-2.0 · production · score:0.80 · hot:0.65 · rising:0.73 · durable:0.68 · board:rising · trend:up* Materials for the Hugging Face Diffusion Models Course **Why it matters.** This repository provides free educational materials for Hugging Face's Diffusion Models Course, including Jupyter notebooks on theory, implementation with the Diffusers library, training, and fine-tuning diffusion models for image and audio generation. It matters now because diffusion models are a leading technique in generative AI, with applications in creative tools and research, and this resource helps developers and researchers quickly upskill amid rapid advancements in the field. _themes: diffusion · generative-ai · fine-tuning · notebooks_ #### [microsoft/DirectX-Graphics-Samples](https://github.com/microsoft/DirectX-Graphics-Samples) *C++ · ★6,719 · MIT · production · score:0.85 · hot:0.64 · rising:0.74 · durable:0.72 · board:rising · trend:up* This repo contains the DirectX Graphics samples that demonstrate how to build graphics intensive applications on Windows. **Why it matters.** This repository contains sample code for DirectX 12 graphics programming, demonstrating features like mesh shaders and variable rate shading to help developers build high-performance applications on Windows. It matters now because DirectX remains a core technology for Windows-based graphics development, especially in gaming and professional applications, where understanding these samples can aid in optimizing performance amid evolving hardware and API standards. _themes: graphics · directx · shaders · rendering_ #### [mistralai/cookbook](https://github.com/mistralai/cookbook) *Jupyter Notebook · ★2,224 · MIT · beta · score:0.65 · hot:0.62 · rising:0.68 · durable:0.59 · board:rising · trend:up* **Why it matters.** This repository serves as a community-driven collection of examples and tutorials for Mistral AI models, including notebooks and markdown files that demonstrate practical applications. It matters now because it provides accessible, reproducible code amid the growing adoption of open-source AI models, helping developers and researchers experiment efficiently without relying solely on official documentation. _themes: inference · fine-tuning · notebooks · community_ #### [huggingface/notebooks](https://github.com/huggingface/notebooks) *Jupyter Notebook · ★4,518 · Apache-2.0 · beta · score:0.80 · hot:0.62 · rising:0.69 · durable:0.59 · board:rising · trend:up* Notebooks using the Hugging Face libraries 🤗 **Why it matters.** This repository contains a collection of Jupyter notebooks that demonstrate how to use Hugging Face libraries for tasks such as natural language processing, computer vision, and model fine-tuning. It matters right now because it provides accessible, hands-on examples that help developers and researchers quickly adopt Hugging Face's tools amid the growing demand for AI prototyping and experimentation in a rapidly evolving AI landscape. _themes: fine-tuning · inference · nlp · computer-vision_ #### [openai/build-hours](https://github.com/openai/build-hours) *Jupyter Notebook · ★742 · MIT · experimental · score:0.70 · hot:0.62 · rising:0.67 · durable:0.59 · board:rising · trend:stable* Build hours code to share. **Why it matters.** This repository shares code from OpenAI's Build Hours sessions, which are live demos and tutorials for building with their latest AI tools, including hands-on examples in Jupyter Notebooks. It matters now because it provides practical resources for developers to quickly experiment with evolving AI capabilities, though its utility is limited to supplementary learning rather than comprehensive tools, as it's tied to specific events without formal releases. _themes: demos · ai-education · openai · hands-on_ #### [huggingface/cookbook](https://github.com/huggingface/cookbook) *Jupyter Notebook · ★2,647 · Apache-2.0 · beta · score:0.75 · hot:0.62 · rising:0.68 · durable:0.60 · board:rising · trend:up* Open-source AI cookbook **Why it matters.** This repository serves as a community-driven collection of Jupyter notebooks providing practical, end-to-end examples for building AI applications using open-source tools and models from Hugging Face. It matters right now because the rapid growth of AI development demands accessible resources for hands-on learning and experimentation, helping developers quickly apply AI in real-world scenarios without proprietary dependencies. As open-source AI ecosystems expand, this cookbook fosters collaboration and knowledge sharing among diverse users. _themes: ai-examples · open-source · jupyter-notebooks · fine-tuning_ #### [microsoft/LangChain4j-for-Beginners](https://github.com/microsoft/LangChain4j-for-Beginners) *Java · ★305 · MIT · beta · score:0.60 · hot:0.62 · rising:0.67 · durable:0.59 · board:rising · trend:stable* A course for AI applications with LangChain4j from simple chat to AI agents. **Why it matters.** This repository offers a beginner-oriented course on building AI applications using LangChain4j, a Java library for integrating LLMs like Azure OpenAI, covering topics from basic chat interfaces to advanced AI agents. It matters now because the growing demand for AI development in Java ecosystems requires accessible educational resources, though its lack of formal releases and modest 304 stars suggest it's still niche and not yet polished for widespread adoption. _themes: agents · chat · java · langchain_ #### [NVIDIA/cuopt-examples](https://github.com/NVIDIA/cuopt-examples) *Jupyter Notebook · ★434 · Apache-2.0 · beta · score:0.60 · hot:0.61 · rising:0.66 · durable:0.59 · board:rising · trend:stable* NVIDIA cuOpt examples for decision optimization **Why it matters.** This repository provides examples for NVIDIA cuOpt, a GPU-accelerated engine for solving optimization problems like mixed integer programming and vehicle routing, demonstrating its use via APIs and integrations. It matters now because industries such as logistics and supply chain are increasingly adopting GPU-based acceleration for faster decision-making in real-time scenarios, though its niche focus and moderate adoption (434 stars) suggest it's not yet a mainstream solution. _themes: optimization · gpu · routing · algorithms_ #### [NVIDIA/GenerativeAIExamples](https://github.com/NVIDIA/GenerativeAIExamples) *Jupyter Notebook · ★3,920 · Apache-2.0 · beta · score:0.80 · hot:0.61 · rising:0.65 · durable:0.73 · board:durable · trend:stable* Generative AI reference workflows optimized for accelerated infrastructure and microservice architecture. **Why it matters.** This repository provides Jupyter Notebook examples for generative AI workflows, focusing on NVIDIA's ecosystem like NeMo and NIMs for tasks such as RAG, agentic AI, and fine-tuning, all optimized for GPU acceleration and microservices. It matters now because developers need efficient, hardware-accelerated implementations to handle the growing complexity and scale of AI applications, especially with the surge in LLM deployments and real-world integrations. These examples bridge the gap between NVIDIA tools and practical development, potentially saving time but requiring familiarity with their ecosystem. _themes: rag · agents · inference · fine-tuning_ #### [patchy631/ai-engineering-hub](https://github.com/patchy631/ai-engineering-hub) *Jupyter Notebook · ★33,759 · MIT · production · score:0.80 · hot:0.60 · rising:0.67 · durable:0.76 · board:durable · trend:stable* In-depth tutorials on LLMs, RAGs and real-world AI agent applications. **Why it matters.** This repository provides in-depth tutorials and over 90 production-ready projects focused on large language models (LLMs), retrieval-augmented generation (RAG), and real-world AI agent applications, serving as a comprehensive learning resource for AI engineering. It matters now because the rapid advancement of AI technologies demands accessible, hands-on materials to bridge theory and practice, especially amid growing interest in generative AI and agent-based systems. _themes: rag · agents · llms · machine-learning_ #### [openai/openai-apps-sdk-examples](https://github.com/openai/openai-apps-sdk-examples) *TypeScript · ★2,193 · MIT · beta · score:0.70 · hot:0.60 · rising:0.66 · durable:0.58 · board:rising · trend:up* Example apps for the Apps SDK **Why it matters.** This repository provides example applications and UI components for OpenAI's Apps SDK, illustrating how to integrate external tools and rich interfaces with ChatGPT using the Model Context Protocol (MCP). It serves as a practical starting point for developers building custom AI assistants, which is increasingly relevant as AI platforms evolve and demand seamless tool integrations. However, its utility is limited by the lack of official releases and potential compatibility issues, like Chrome flag requirements, making it more of a reference than a polished solution. _themes: sdk · chatgpt · tools · ui_ #### [openai/plugins](https://github.com/openai/plugins) *Python · ★863 · no-license · experimental · score:0.60 · hot:0.60 · rising:0.62 · durable:0.52 · board:rising · trend:stable* OpenAI Plugins **Why it matters.** This repository provides a collection of example plugins for OpenAI's Codex, demonstrating integrations with tools like Figma and Notion to extend AI capabilities in coding and design workflows. It matters for developers exploring AI-assisted development, but its lack of a license, releases, and low activity (851 stars) suggests it's more of an illustrative resource than a ready-to-use tool, potentially limiting broader adoption in the current AI integration boom. _themes: agents · plugins · code-generation · ai-integration_ #### [microsoft/agent-academy](https://github.com/microsoft/agent-academy) *JavaScript · ★2,124 · MIT · beta · score:0.60 · hot:0.60 · rising:0.63 · durable:0.56 · board:rising · trend:stable* Curated lessons on getting started building agents with Copilot Studio **Why it matters.** This repository offers curated lessons and a three-part course for beginners to learn building agents using Microsoft Copilot Studio, focusing on practical examples and contributions. It matters now amid the growing interest in AI agents and Microsoft's push for Copilot tools, providing accessible education, though it's limited to Microsoft's ecosystem and lacks a formal release, making it more supplementary than essential. _themes: agents · copilot · education · javascript_ #### [NVIDIA/accelerated-computing-hub](https://github.com/NVIDIA/accelerated-computing-hub) *Jupyter Notebook · ★1,483 · NOASSERTION · beta · score:0.70 · hot:0.59 · rising:0.64 · durable:0.56 · board:rising · trend:stable* NVIDIA curated collection of educational resources related to general purpose GPU programming. **Why it matters.** This repository serves as a curated collection of educational resources, including tutorials and guides for GPU programming with a focus on NVIDIA technologies, aimed at helping users learn accelerated computing. It matters right now because the growing demand for GPU skills in AI, machine learning, and high-performance computing makes accessible learning materials essential for developers and researchers navigating these technologies. _themes: gpu · cuda · tutorials · parallelism_ #### [anthropics/claude-desktop-buddy](https://github.com/anthropics/claude-desktop-buddy) *C++ · ★721 · NOASSERTION · experimental · score:0.50 · hot:0.59 · rising:0.61 · durable:0.53 · board:rising · trend:stable* Reference and an example for the Bluetooth API for makers in Claude Cowork & Claude Code Desktop **Why it matters.** This repository offers a reference and example implementation for integrating hardware devices via Bluetooth with Anthropic's Claude AI tools on macOS and Windows, enabling makers to build interactive gadgets that display AI prompts and messages. It matters now as AI-hardware integrations are growing, allowing creative experimentation in the maker community, though its specificity to Claude limits broader applicability and it lacks formal releases or licensing clarity. _themes: ble · ai-hardware · esp32 · iot_ #### [beatai-org/BeatAI](https://github.com/beatai-org/BeatAI) *Handlebars · ★4,655 · MIT · beta · score:0.60 · hot:0.54 · rising:0.58 · durable:0.62 · board:durable · trend:stable* 🌶️ 通过 AI 辣评学习 AI，模拟各种明星角色，给大家不一样的学习体验。🦄 BeatAI，一片简单有趣的 AI 大陆，欢迎大家常来住住。 **Why it matters.** BeatAI is a community-driven platform focused on making AI learning engaging by simulating celebrity roles and providing insights through articles on topics like agent architectures and prompt engineering, primarily in Chinese. It matters right now as accessible AI education is crucial amid the LLM proliferation, but its informal nature and lack of original code raise questions about depth and practical utility compared to structured resources. _themes: ai-learning · agents · llm · prompt-engineering_ #### [openai/openai-cua-sample-app](https://github.com/openai/openai-cua-sample-app) *TypeScript · ★1,695 · MIT · experimental · score:0.60 · hot:0.51 · rising:0.57 · durable:0.55 · board:rising · trend:stable* Learn how to use CUA (our Computer Using Agent) via the API on multiple computer environments. **Why it matters.** This repository provides a TypeScript sample app for integrating OpenAI's CUA agent in browser-focused workflows, demonstrating API usage for managing AI-driven computer interactions and scenarios. It matters now as AI agents are advancing quickly, offering developers practical insights into building automation tools with models like GPT-5.4, though it's limited to demonstration purposes and may not cover production-ready implementations. _themes: agents · api-integration · browser-automation · workflows_ #### [openai/openai-chatkit-advanced-samples](https://github.com/openai/openai-chatkit-advanced-samples) *? · ★612 · NOASSERTION · beta · score:0.70 · hot:0.51 · rising:0.56 · durable:0.59 · board:durable · trend:stable* Starter app to build with OpenAI ChatKit SDK **Why it matters.** This repository provides advanced sample applications using OpenAI's ChatKit SDK, demonstrating how to build chat-based interfaces with FastAPI backends and React frontends for scenarios like virtual pet management and customer support. It matters right now because developers are rapidly adopting AI chat technologies, and these examples offer practical, ready-to-run code to accelerate prototyping amid the growing demand for generative AI integrations. _themes: agents · chat · sdk · demo_ #### [openai/openai-chatkit-starter-app](https://github.com/openai/openai-chatkit-starter-app) *Python · ★836 · MIT · beta · score:0.60 · hot:0.50 · rising:0.56 · durable:0.56 · board:durable · trend:stable* Starter app to build with OpenAI ChatKit + Agent Builder **Why it matters.** This repository provides starter templates for integrating OpenAI's ChatKit, offering simple examples for self-hosted and managed chat applications using Python. It matters as a quick onboarding tool for developers exploring OpenAI's AI agent capabilities, especially with the increasing focus on conversational AI, but its lack of official releases and basic scope limits its utility for advanced use cases. _themes: chatkit · agents · openai-api · inference_ #### [anthropics/prompt-eng-interactive-tutorial](https://github.com/anthropics/prompt-eng-interactive-tutorial) *Jupyter Notebook · ★34,818 · no-license · production · score:0.85 · hot:0.49 · rising:0.61 · durable:0.73 · board:durable · trend:stable* Anthropic's Interactive Prompt Engineering Tutorial **Why it matters.** This repository provides an interactive tutorial for engineering prompts specifically for Anthropic's Claude AI models, featuring structured lessons, exercises, and a playground for experimentation to help users build effective prompts. It matters now because prompt engineering is a critical skill for optimizing AI interactions amid widespread LLM adoption, and this resource offers practical, model-specific guidance directly from the developers, though its lack of a formal license may limit broader accessibility. _themes: prompt-engineering · llm · tutorial · ai-education_ #### [openai/openai-fm](https://github.com/openai/openai-fm) *TypeScript · ★2,843 · MIT · beta · score:0.70 · hot:0.49 · rising:0.62 · durable:0.68 · board:durable · trend:stable* Code for openai.fm, a demo for the OpenAI Speech API **Why it matters.** This repository provides a demo application for OpenAI's text-to-speech API, built with NextJS, allowing users to interactively generate speech from text. It matters now because text-to-speech technology is advancing quickly in AI, enabling developers to experiment with OpenAI's models for potential applications in voice interfaces, though it's limited to showcasing rather than production use. _themes: tts · speech · api · demo_ #### [anthropics/courses](https://github.com/anthropics/courses) *Jupyter Notebook · ★20,696 · NOASSERTION · production · score:0.85 · hot:0.48 · rising:0.60 · durable:0.71 · board:durable · trend:stable* Anthropic's educational courses **Why it matters.** This repository provides a series of free, interactive courses from Anthropic on using their Claude AI models, covering topics from API basics to advanced prompt engineering and tool integration. It matters now because the AI industry is rapidly evolving, and these courses offer practical, cost-effective training that helps developers and beginners quickly build skills in generative AI amid increasing demand for AI expertise. _themes: prompting · api · evaluations · tools_ #### [anthropics/claude-quickstarts](https://github.com/anthropics/claude-quickstarts) *Python · ★16,167 · MIT · beta · score:0.80 · hot:0.47 · rising:0.59 · durable:0.69 · board:durable · trend:stable* A collection of projects designed to help developers quickly get started with building deployable applications using the Claude API **Why it matters.** This repository provides a collection of starter projects in Python for developers to quickly build and deploy applications using Anthropic's Claude API, including examples like customer support agents and browser automation. It matters now because the growing adoption of AI models like Claude requires accessible resources for rapid prototyping, helping developers integrate advanced AI capabilities amid increasing competition in the AI tools market. _themes: agents · api · quickstart · demo_ #### [microsoft/vscode-extension-samples](https://github.com/microsoft/vscode-extension-samples) *TypeScript · ★10,034 · MIT · production · score:0.85 · hot:0.47 · rising:0.59 · durable:0.68 · board:durable · trend:stable* Sample code illustrating the VS Code extension API. **Why it matters.** This repository provides a collection of sample code snippets that illustrate how to use the VS Code extension API, helping developers build and understand custom extensions for the popular code editor. It matters now because VS Code is a leading IDE with a vast ecosystem, and these samples offer practical, educational resources for creating extensions that enhance productivity, especially as more developers seek to customize their tools amid growing demands for integrated development environments. _themes: extensions · api · samples · ide_ #### [openai/openai-realtime-agents](https://github.com/openai/openai-realtime-agents) *TypeScript · ★6,826 · MIT · beta · score:0.70 · hot:0.46 · rising:0.57 · durable:0.67 · board:durable · trend:stable* This is a simple demonstration of more advanced, agentic patterns built on top of the Realtime API. **Why it matters.** This repo demonstrates advanced patterns for building voice agents using OpenAI's Realtime API and Agents SDK, including chat supervision and sequential handoffs for handling user intents in real-time applications. It matters now because AI agents are increasingly used in customer service and interactive systems, where low-latency interactions can improve user experiences, though it remains a basic demo that may not address scalability or production challenges. As AI agent frameworks evolve rapidly, this provides a practical starting point for developers exploring these patterns. _themes: agents · realtime · voice-interactions · sdk_ #### [openai/openai-quickstart-python](https://github.com/openai/openai-quickstart-python) *? · ★1,798 · MIT · beta · score:0.60 · hot:0.46 · rising:0.55 · durable:0.63 · board:durable · trend:stable* Python example app from the OpenAI API quickstart tutorial **Why it matters.** This repository provides simple Python examples and setup instructions for interacting with OpenAI's API, focusing on endpoints like chat completions and assistants, making it easy for newcomers to send their first requests. It matters right now because OpenAI's APIs are widely used in AI development, and this repo serves as an official entry point for beginners amid growing interest in generative AI, though it's limited to basic tutorials without advanced features or updates. _themes: api · chat · quickstart · python_ #### [openai/openai-realtime-console](https://github.com/openai/openai-realtime-console) *JavaScript · ★3,571 · MIT · experimental · score:0.70 · hot:0.45 · rising:0.54 · durable:0.62 · board:durable · trend:stable* React app for inspecting, building and debugging with the Realtime API **Why it matters.** This repository offers a React-based application for developers to inspect, build, and debug interactions with OpenAI's Realtime API using WebRTC, including viewing event payloads and configuring client-side function calls. It matters now because real-time AI capabilities are essential for emerging applications like live conversational agents, and this tool provides a straightforward way to prototype and troubleshoot amid growing adoption of OpenAI's features. _themes: realtime · webrtc · debugging · openai_ #### [openai/openai-quickstart-node](https://github.com/openai/openai-quickstart-node) *JavaScript · ★2,631 · MIT · beta · score:0.70 · hot:0.45 · rising:0.56 · durable:0.65 · board:durable · trend:stable* Node.js example app from the OpenAI API quickstart tutorial **Why it matters.** This repository offers Node.js examples for using OpenAI's APIs, covering basics like chat completions and fine-tuning, serving as a straightforward starting point for developers integrating AI features. It matters now because OpenAI's APIs are widely adopted for applications like chatbots and content generation, but the examples are basic and lack depth, making it more suitable for initial learning rather than advanced production use. _themes: openai-api · inference · fine-tuning · node.js_ #### [openai/openai-cs-agents-demo](https://github.com/openai/openai-cs-agents-demo) *Python · ★5,961 · MIT · experimental · score:0.70 · hot:0.45 · rising:0.54 · durable:0.64 · board:durable · trend:stable* Demo of a customer service use case implemented with the OpenAI Agents SDK **Why it matters.** This repository demonstrates a customer service agent using OpenAI's Agents SDK, with a Python backend for orchestration and a Next.js frontend for chat interaction, allowing users to visualize and test agent workflows. It matters now because AI agents are increasingly adopted for real-world applications like customer support, providing developers with a timely example to experiment with OpenAI's tools amid the rapid evolution of agentic AI systems. _themes: agents · sdk · chat-ui · customer-service_ #### [microsoft/MLOpsPython](https://github.com/microsoft/MLOpsPython) *Python · ★1,303 · MIT · production · score:0.70 · hot:0.44 · rising:0.56 · durable:0.64 · board:durable · trend:stable* MLOps using Azure ML Services and Azure DevOps **Why it matters.** This repository offers a template for setting up CI/CD pipelines for machine learning projects using Azure Machine Learning and Azure DevOps, focusing on automation of model training, evaluation, and deployment. It matters now because MLOps practices are critical for scaling AI in production environments, but its Azure-specific approach may limit broader adoption compared to more vendor-agnostic tools. _themes: mlops · ci-cd · azure · python_ #### [openai/gpt-5-coding-examples](https://github.com/openai/gpt-5-coding-examples) *HTML · ★1,876 · MIT · experimental · score:0.70 · hot:0.44 · rising:0.51 · durable:0.63 · board:durable · trend:down* GPT-5 coding examples **Why it matters.** This repository features a collection of front-end demo applications generated entirely by GPT-5 prompts, showcasing AI's ability to create code for websites and interactive UIs without manual intervention. It matters now as AI models like GPT-5 advance rapidly, offering developers quick prototyping tools to boost productivity amid growing interest in automated coding, though its reliance on unreleased models makes it more conceptual than practical. _themes: ai-coding · prompt-engineering · front-end · demos_ #### [anthropics/claude-agent-sdk-demos](https://github.com/anthropics/claude-agent-sdk-demos) *TypeScript · ★2,212 · no-license · experimental · score:0.60 · hot:0.44 · rising:0.50 · durable:0.55 · board:durable · trend:stable* Claude Code SDK Demos **Why it matters.** This repository provides demo applications for Anthropic's Claude Agent SDK, showcasing how to build AI-powered agents for tasks like email assistance, research, and spreadsheet handling using TypeScript. It matters right now because the growing interest in AI agents requires accessible examples for developers to experiment with Claude's capabilities, though it's explicitly for local use only and not production-ready. However, its lack of a license and formal releases limits broader adoption. _themes: agents · sdk · demos · ai-assistants_ #### [openai/openai-assistants-quickstart](https://github.com/openai/openai-assistants-quickstart) *TypeScript · ★1,963 · MIT · beta · score:0.70 · hot:0.44 · rising:0.54 · durable:0.62 · board:durable · trend:stable* OpenAI Assistants API quickstart with Next.js. **Why it matters.** This repository provides a quickstart template for integrating OpenAI's Assistants API into a Next.js application, demonstrating features like streaming, tool use, and function calling to help developers build AI assistants rapidly. It matters now because the Assistants API represents a growing trend in AI for creating advanced conversational agents, and this template lowers the barrier for developers amid increasing demand for AI-powered apps, though it's basic and not comprehensive for production use. _themes: agents · streaming · function-calling · nextjs_ #### [microsoft/AutonomousDrivingCookbook](https://github.com/microsoft/AutonomousDrivingCookbook) *Jupyter Notebook · ★2,429 · MIT · beta · score:0.70 · hot:0.44 · rising:0.52 · durable:0.59 · board:durable · trend:stable* Scenarios, tutorials and demos for Autonomous Driving **Why it matters.** This repository provides tutorials, scenarios, and demos for autonomous driving, primarily using simulation tools like AirSim to help developers train and test AI models without real-world hardware. It matters now because the autonomous driving industry is advancing rapidly, with simulations addressing the critical need for vast amounts of training data to improve safety and reliability in self-driving systems, though it's still a work in progress and may lack comprehensive coverage. _themes: simulation · autonomous-driving · deep-learning · tutorials_ #### [huggingface/transformers.js-examples](https://github.com/huggingface/transformers.js-examples) *JavaScript · ★2,020 · Apache-2.0 · beta · score:0.70 · hot:0.43 · rising:0.51 · durable:0.61 · board:durable · trend:stable* A collection of 🤗 Transformers.js demos and example applications **Why it matters.** This repository provides a collection of demos and example applications for Transformers.js, a JavaScript library that enables running Hugging Face transformer models in the browser via WebGPU, covering tasks like language modeling and image segmentation. It matters right now because it lowers the barrier for web developers to experiment with on-device AI, promoting efficient, privacy-focused applications amid the growing demand for edge-based machine learning solutions, though it lacks formal releases which may indicate ongoing development. _themes: inference · webgpu · transformers · demos_ #### [openai/plugins-quickstart](https://github.com/openai/plugins-quickstart) *Python · ★4,235 · MIT · archived · score:0.60 · hot:0.43 · rising:0.50 · durable:0.57 · board:durable · trend:stable* Get a ChatGPT plugin up and running in under 5 minutes! **Why it matters.** This repository provides a quickstart guide to set up a simple TODO list plugin for ChatGPT using Python, allowing users to run it locally in under 5 minutes for basic experimentation. It matters as an educational resource for understanding legacy ChatGPT plugins, though they've been superseded by GPTs with actions, making it less relevant for new development but still useful for beginners transitioning to modern OpenAI tools. _themes: chatgpt · plugins · quickstart · ai-development_ #### [openai/gpt-discord-bot](https://github.com/openai/gpt-discord-bot) *Python · ★1,835 · MIT · beta · score:0.60 · hot:0.43 · rising:0.52 · durable:0.60 · board:durable · trend:stable* Example Discord bot written in Python that uses the completions API to have conversations with the `text-davinci-003` model, and the moderations API to filter the messages. **Why it matters.** This repository offers an example Python-based Discord bot that leverages OpenAI's API for conversational interactions using the GPT-3.5-Turbo model and includes message moderation via the moderations API, making it a basic template for integrating AI chat features. It matters for developers experimenting with AI in social platforms, as Discord remains popular for community interactions, but its value is limited since it's not actively updated beyond bug fixes and lacks advanced features. While timely amid growing AI chatbot adoption, it's more of a starting point than a robust solution. _themes: inference · chatbot · moderation_ #### [huggingface/swift-coreml-diffusers](https://github.com/huggingface/swift-coreml-diffusers) *Swift · ★2,750 · Apache-2.0 · beta · score:0.75 · hot:0.42 · rising:0.51 · durable:0.61 · board:durable · trend:stable* Swift app demonstrating Core ML Stable Diffusion **Why it matters.** This repository provides a Swift application that demonstrates how to integrate Apple's Core ML implementation of Stable Diffusion for on-device image generation, simplifying the diffusers library for native Apple platforms. It matters right now because it facilitates faster development and deployment of AI features on Apple devices, amid growing demand for efficient, hardware-accelerated ML in mobile and desktop apps, especially with recent OS updates enhancing Core ML capabilities. _themes: inference · on-device · swift · stable-diffusion_ #### [openai/openai-responses-starter-app](https://github.com/openai/openai-responses-starter-app) *TypeScript · ★810 · NOASSERTION · beta · score:0.70 · hot:0.42 · rising:0.50 · durable:0.58 · board:durable · trend:stable* Starter app to build with the OpenAI Responses API **Why it matters.** This repository provides a NextJS starter app for building conversational assistants using OpenAI's Responses API, including features like multi-turn chats, web search, and file search integrations. It matters now because it helps developers quickly prototype AI agents in a rapidly evolving AI landscape, but its lack of a formal release and basic implementation means it's more of a template than a robust solution, potentially limiting its immediate utility for production environments. _themes: agents · chat · tools · integration_ #### [openai/web-crawl-q-and-a-example](https://github.com/openai/web-crawl-q-and-a-example) *Jupyter Notebook · ★322 · no-license · experimental · score:0.60 · hot:0.41 · rising:0.47 · durable:0.51 · board:durable · trend:stable* Learn how to crawl your website and build a Q/A bot with the OpenAI API **Why it matters.** This repository provides a Jupyter Notebook example for crawling a website and building a Q&A bot using OpenAI's API and embeddings, serving as a basic tutorial for integrating AI into search applications. It matters because it illustrates practical use of LLMs for retrieval-augmented generation amid growing interest in AI-driven tools, but its lack of licensing and updates makes it less reliable for production use. _themes: rag · embeddings · web-crawling · inference_ #### [microsoft/MLOps](https://github.com/microsoft/MLOps) *Jupyter Notebook · ★2,084 · MIT · archived · score:0.40 · hot:0.41 · rising:0.48 · durable:0.51 · board:durable · trend:stable* MLOps examples **Why it matters.** This repository provides a collection of end-to-end MLOps examples using Azure Machine Learning and related services to operationalize ML workflows, integrated with tools like GitHub and Azure DevOps. It matters because MLOps is critical for scaling and managing ML in production, but this repo appears outdated, as it redirects to newer resources like Azure/mlops-v2, making it less relevant for current practitioners. _themes: mlops · azureml · deployment · pipelines_ #### [NVIDIA/workbench-example-hybrid-rag](https://github.com/NVIDIA/workbench-example-hybrid-rag) *Python · ★367 · Apache-2.0 · experimental · score:0.60 · hot:0.41 · rising:0.48 · durable:0.53 · board:durable · trend:stable* An NVIDIA AI Workbench example project for Retrieval Augmented Generation (RAG) **Why it matters.** This repository provides an example chat application for Retrieval Augmented Generation (RAG) that integrates with NVIDIA APIs, NIM containers, and Hugging Face models, simplifying testing on NVIDIA hardware. It matters because RAG is a key technique for enhancing AI accuracy with external data, and this offers a straightforward way for developers to prototype amid growing demand for efficient AI workflows, though it's limited to NVIDIA ecosystems and may not be as versatile as broader tools. _themes: rag · nvidia · inference · chatbot_ #### [allenai/OLMoE.swift](https://github.com/allenai/OLMoE.swift) *Swift · ★310 · Apache-2.0 · beta · score:0.60 · hot:0.40 · rising:0.49 · durable:0.62 · board:durable · trend:down* **Why it matters.** This repository provides an iOS app built in Swift for running an offline large language model, specifically OLMoE, allowing users to ask questions and get responses directly on their device without internet access, prioritizing privacy. It matters now due to growing demand for on-device AI amid privacy concerns and advancements in mobile hardware, though its niche focus on Apple ecosystems and modest adoption (310 stars) limits broader impact. _themes: inference · on-device · llm · swift_ #### [openai/openai-realtime-twilio-demo](https://github.com/openai/openai-realtime-twilio-demo) *TypeScript · ★517 · MIT · experimental · score:0.65 · hot:0.40 · rising:0.49 · durable:0.55 · board:durable · trend:down* **Why it matters.** This repo demonstrates how to integrate OpenAI's Realtime API with Twilio for building a real-time AI voice assistant that handles phone calls, forwarding audio streams for AI processing. It matters now because voice AI integrations are increasingly relevant for applications like customer service, but this demo is basic and requires manual setup, making it useful for prototyping yet limited for production use. _themes: realtime · voice-ai · ai-integration · websocket_ #### [openai/openai-testing-agent-demo](https://github.com/openai/openai-testing-agent-demo) *TypeScript · ★759 · NOASSERTION · experimental · score:0.70 · hot:0.40 · rising:0.46 · durable:0.55 · board:durable · trend:down* Demo of a UI testing agent using the OpenAI CUA model and the Responses API. **Why it matters.** This repo demonstrates an AI agent using OpenAI's CUA model and Playwright to automate frontend UI testing by navigating and interacting with web apps based on test cases. It matters right now as it highlights the potential of AI for reducing manual testing efforts in software development, though its preview status limits it to exploratory use amid growing interest in AI-driven automation tools. _themes: agents · testing · automation · ai_ #### [google-deepmind/neural-processes](https://github.com/google-deepmind/neural-processes) *Jupyter Notebook · ★1,018 · Apache-2.0 · experimental · score:0.60 · hot:0.40 · rising:0.47 · durable:0.54 · board:durable · trend:down* This repository contains notebook implementations of the following Neural Process variants: Conditional Neural Processes (CNPs), Neural Processes (NPs), Attentive Neural Processes (ANPs). **Why it matters.** This repository provides Jupyter notebook implementations of Neural Process variants like CNPs, NPs, and ANPs, which are meta-learning models for tasks such as function regression and uncertainty estimation. It matters for researchers exploring probabilistic machine learning, as these models remain relevant for few-shot learning applications, though the implementations use outdated dependencies like TensorFlow 1.13.1, limiting their practicality for current production use. _themes: meta-learning · neural-processes · probabilistic-models · few-shot-learning_ #### [google-deepmind/educational](https://github.com/google-deepmind/educational) *Jupyter Notebook · ★1,482 · Apache-2.0 · beta · score:0.70 · hot:0.40 · rising:0.49 · durable:0.60 · board:durable · trend:down* **Why it matters.** This repository provides a collection of beginner-friendly Jupyter notebooks on basic machine learning concepts, such as AI agents in games and language models, accessible via Google Colab. It matters now because the rapid growth of AI technologies has increased the need for accessible educational resources to help newcomers build foundational skills, especially amid efforts to democratize AI knowledge. _themes: education · ml-tutorials · agents · nlp_ #### [openai/openai-realtime-solar-system](https://github.com/openai/openai-realtime-solar-system) *TypeScript · ★490 · MIT · experimental · score:0.60 · hot:0.39 · rising:0.46 · durable:0.57 · board:durable · trend:down* Demo showing how to use the OpenAI Realtime API to navigate a 3D scene via tool calling **Why it matters.** This repo is a demo that uses OpenAI's Realtime API to enable voice commands for navigating a simple 3D solar system scene, leveraging WebRTC and function calling; it serves as an educational example for integrating AI with interactive web apps. While it demonstrates emerging real-time AI capabilities, its value is limited by being a basic, non-production demo that doesn't introduce novel concepts beyond OpenAI's own tools, making it useful mainly for developers experimenting with the API but not essential for broader adoption right now. _themes: realtime · agents · inference · voice-interaction_ #### [openai/openai-structured-outputs-samples](https://github.com/openai/openai-structured-outputs-samples) *TypeScript · ★672 · MIT · experimental · score:0.70 · hot:0.38 · rising:0.46 · durable:0.58 · board:durable · trend:down* Sample apps to help developers get started with Structured Outputs **Why it matters.** This repository provides sample applications using OpenAI's Structured Outputs feature to ensure API responses conform to a specified JSON schema, making AI outputs more reliable for integration into apps. It matters now as developers face challenges with unpredictable model responses in production workflows, and these examples offer practical guidance using NextJS to bridge that gap quickly amid growing AI adoption. _themes: api · structured-outputs · json-schema · samples_ #### [huggingface/huggingface-llama-recipes](https://github.com/huggingface/huggingface-llama-recipes) *Jupyter Notebook · ★698 · no-license · experimental · score:0.70 · hot:0.38 · rising:0.44 · durable:0.50 · board:durable · trend:down* **Why it matters.** This repository provides minimal Jupyter Notebook recipes for quickly setting up and running Meta's Llama 3.x models, such as Llama 3.1 and 3.2, using Hugging Face's transformers library, which is useful for basic experimentation and inference tasks. It matters right now because of the ongoing interest in these evolving LLMs and the need for straightforward onboarding amid frequent model updates, but its work-in-progress status means it lacks stability and comprehensive coverage, potentially leading to incomplete or changing examples. _themes: inference · llm · quickstart · huggingface_ #### [huggingface/sam2-studio](https://github.com/huggingface/sam2-studio) *Swift · ★412 · Apache-2.0 · experimental · score:0.60 · hot:0.37 · rising:0.44 · durable:0.53 · board:durable · trend:down* **Why it matters.** This repository provides a Swift-based demo application for the SAM 2 Core ML models, enabling interactive object segmentation in images on Apple devices through user inputs like points and boxes. It matters because it demonstrates on-device application of recent advancements in visual segmentation from FAIR, potentially aiding app developers in exploring AI features, but its lack of a formal release, low adoption (412 stars), and image-only support make it more of a niche proof-of-concept than a robust tool. _themes: segmentation · computer-vision · coreml · on-device-ml_ #### [openai/dalle-2-preview](https://github.com/openai/dalle-2-preview) *? · ★1,042 · no-license · experimental · score:0.60 · hot:0.35 · rising:0.42 · durable:0.50 · board:durable · trend:down* **Why it matters.** This repository appears to be a preview for OpenAI's DALL-E 2, an AI model that generates images from textual descriptions, but it lacks actual code, releases, or substantial content beyond what's typical for a promotional page. It matters now as DALL-E 2 represents advancements in generative AI that could influence creative industries, though its utility is limited without accessible resources, making it more of a placeholder than a practical tool. _themes: image-generation · generative-ai · ai-art_ ### eval (42) #### [comet-ml/opik](https://github.com/comet-ml/opik) *Python · ★18,980 · Apache-2.0 · production · score:0.80 · hot:0.88 · rising:0.89 · durable:0.82 · board:rising · trend:up* Debug, evaluate, and monitor your LLM applications, RAG systems, and agentic workflows with comprehensive tracing, automated evaluations, and production-ready dashboards. **Why it matters.** Opik is an open-source platform that enables debugging, evaluation, and monitoring of LLM applications, RAG systems, and agentic workflows through tracing, automated evaluations, and dashboards. It matters right now because the rapid adoption of generative AI in production requires robust tools for observability and optimization to ensure reliability, reduce errors, and comply with safety standards amid increasing complexity in AI development. _themes: llm · evaluation · agents · observability_ #### [EleutherAI/lm-evaluation-harness](https://github.com/EleutherAI/lm-evaluation-harness) *Python · ★12,290 · MIT · beta · score:0.85 · hot:0.80 · rising:0.81 · durable:0.72 · board:rising · trend:up* A framework for few-shot evaluation of language models. **Why it matters.** The LM Evaluation Harness is a Python framework for few-shot evaluation of language models, enabling users to benchmark model performance on various tasks with minimal examples. It matters right now because accurate and standardized evaluation is critical amid the rapid evolution of AI models, helping researchers compare and improve them efficiently. Recent updates, like support for multimodal inputs and a refactored CLI, enhance its usability for emerging use cases such as text-image tasks. _themes: evaluation · language-models · benchmarking · few-shot_ #### [NVIDIA/garak](https://github.com/NVIDIA/garak) *HTML · ★7,570 · Apache-2.0 · beta · score:0.80 · hot:0.79 · rising:0.79 · durable:0.72 · board:rising · trend:up* the LLM vulnerability scanner **Why it matters.** Garak is an open-source tool from NVIDIA that scans large language models for vulnerabilities like hallucination, prompt injection, and toxicity by using probes similar to network security scanners. It matters now because the increasing deployment of LLMs in real-world applications heightens risks of security breaches and misinformation, making systematic evaluation tools essential for ensuring model safety amid growing regulatory and ethical scrutiny. _themes: llm-security · evaluation · red-teaming_ #### [stanford-crfm/helm](https://github.com/stanford-crfm/helm) *Python · ★2,758 · Apache-2.0 · beta · score:0.80 · hot:0.77 · rising:0.79 · durable:0.68 · board:rising · trend:up* Holistic Evaluation of Language Models (HELM) is an open source Python framework created by the Center for Research on Foundation Models (CRFM) at Stanford for holistic, reproducible and transparent evaluation of foundation models, including large language models (LLMs) and multimodal models. **Why it matters.** HELM is an open-source Python framework for holistic evaluation of language models, including LLMs and multimodal models, offering standardized benchmarks, metrics for bias and efficiency, and a web interface for results. It matters now because the AI field requires reproducible and transparent assessments to address ethical concerns and compare models amid rapid advancements, helping researchers and developers make informed decisions on model selection and improvement. _themes: eval · benchmarks · llms · multimodal_ #### [confident-ai/deepeval](https://github.com/confident-ai/deepeval) *Python · ★14,934 · Apache-2.0 · production · score:0.80 · hot:0.76 · rising:0.80 · durable:0.78 · board:rising · trend:up* The LLM Evaluation Framework **Why it matters.** DeepEval is an open-source framework for evaluating large language models (LLMs) using metrics like G-Eval and task completion, running locally on your machine similar to Pytest for unit testing LLM applications. It matters right now because the proliferation of LLMs in production systems demands reliable evaluation to detect issues like hallucinations and ensure prompt stability, helping developers optimize models amid rapid AI advancements and increasing regulatory scrutiny. _themes: llm-eval · metrics · testing · framework_ #### [open-compass/opencompass](https://github.com/open-compass/opencompass) *Python · ★6,895 · Apache-2.0 · beta · score:0.70 · hot:0.73 · rising:0.72 · durable:0.69 · board:hot · trend:up* OpenCompass is an LLM evaluation platform, supporting a wide range of models (Llama3, Mistral, InternLM2,GPT-4,LLaMa2, Qwen,GLM, Claude, etc) over 100+ datasets. **Why it matters.** OpenCompass is an open-source platform for evaluating large language models across over 100 datasets and supports models like Llama3 and GPT-4, enabling systematic benchmarking. It matters now because the rapid growth of LLMs requires reliable, standardized evaluation tools to compare performance and guide improvements, though its effectiveness depends on dataset quality and ease of use. _themes: evaluation · llm · benchmark · datasets_ #### [openai/evals](https://github.com/openai/evals) *Python · ★18,255 · NOASSERTION · beta · score:0.85 · hot:0.70 · rising:0.74 · durable:0.69 · board:rising · trend:up* Evals is a framework for evaluating LLMs and LLM systems, and an open-source registry of benchmarks. **Why it matters.** OpenAI Evals is a framework for evaluating large language models and systems, providing a registry of benchmarks and tools for custom evaluations. It matters now because the rapid proliferation of LLMs in real-world applications demands rigorous, standardized testing to ensure reliability, accuracy, and safety, helping developers avoid costly errors in deployment. _themes: llm-eval · benchmarks · evaluation · testing_ #### [huggingface/lighteval](https://github.com/huggingface/lighteval) *Python · ★2,381 · MIT · beta · score:0.80 · hot:0.68 · rising:0.71 · durable:0.69 · board:rising · trend:up* Lighteval is your all-in-one toolkit for evaluating LLMs across multiple backends **Why it matters.** Lighteval is a toolkit for evaluating large language models across various backends, supporting over 1000 tasks for detailed performance analysis and customization of metrics. It matters now because the rapid growth of LLMs demands robust, flexible evaluation tools to benchmark and debug models accurately, especially amid increasing scrutiny on AI reliability and safety. This integration with Hugging Face's ecosystem helps streamline workflows for developers and researchers working on model improvements. _themes: evaluation · llms · benchmarks · metrics_ #### [huggingface/evaluate](https://github.com/huggingface/evaluate) *Python · ★2,436 · Apache-2.0 · production · score:0.70 · hot:0.67 · rising:0.71 · durable:0.68 · board:rising · trend:up* 🤗 Evaluate: A library for easily evaluating machine learning models and datasets. **Why it matters.** Hugging Face's Evaluate library provides a straightforward way to load and apply various machine learning evaluation metrics for tasks in NLP and computer vision, supporting multiple frameworks like PyTorch and TensorFlow. It promotes standardization by allowing users to create and share custom metrics on the Hub, which is valuable for consistent model comparisons in research and development. However, its relevance is diminishing as the library is not the most actively maintained for emerging areas like LLM evaluation, with Hugging Face recommending their newer LightEval instead. _themes: evaluation · metrics · ml · nlp_ #### [google-research/weatherbench2](https://github.com/google-research/weatherbench2) *Python · ★602 · Apache-2.0 · beta · score:0.80 · hot:0.66 · rising:0.68 · durable:0.64 · board:rising · trend:stable* A benchmark for the next generation of data-driven global weather models. **Why it matters.** WeatherBench 2 provides datasets and evaluation tools for benchmarking data-driven global weather forecasting models against traditional methods, fostering fair comparisons and advancements in AI-based predictions. It matters now as climate change demands more accurate and efficient weather models, and this framework helps standardize evaluations amid growing interest in AI applications for environmental science. The recent update to WeatherBench-X indicates active evolution in this space. _themes: weather · benchmarking · datasets · AI-evaluation_ #### [openai/frontier-evals](https://github.com/openai/frontier-evals) *Python · ★1,170 · MIT · beta · score:0.75 · hot:0.64 · rising:0.67 · durable:0.64 · board:rising · trend:stable* OpenAI Frontier Evals **Why it matters.** This repository provides code and benchmarks for evaluating the capabilities of advanced AI models, including tools like PaperBench for replicating research papers, SWE-Lancer for software engineering tasks, and EVMbench for smart contract security. It matters now because the rapid evolution of AI requires standardized, rigorous evaluations to assess model performance, identify limitations, and guide development in a field increasingly focused on reliability and real-world applicability. _themes: benchmarks · ai-evals · model-testing · llm_ #### [google-deepmind/bsuite](https://github.com/google-deepmind/bsuite) *Python · ★1,540 · Apache-2.0 · beta · score:0.75 · hot:0.63 · rising:0.66 · durable:0.68 · board:durable · trend:stable* bsuite is a collection of carefully-designed experiments that investigate core capabilities of a reinforcement learning (RL) agent **Why it matters.** bsuite is a benchmark suite from Google DeepMind that offers a collection of experiments to systematically evaluate the core capabilities of reinforcement learning agents, such as learning efficiency and generalization. It matters now because reinforcement learning research is grappling with issues like scalability and reliability in real-world applications, and bsuite provides standardized, reproducible tests to help identify algorithmic flaws, potentially accelerating progress in AI safety and agent design. _themes: rl · benchmarking · agents · evaluation_ #### [facebookresearch/meta-agents-research-environments](https://github.com/facebookresearch/meta-agents-research-environments) *Python · ★477 · MIT · experimental · score:0.65 · hot:0.63 · rising:0.60 · durable:0.55 · board:hot · trend:stable* Meta Agents Research Environments is a comprehensive platform designed to evaluate AI agents in dynamic, realistic scenarios. Unlike static benchmarks, this platform introduces evolving environments where agents must adapt their strategies as new information becomes available, mirroring real-world challenges. **Why it matters.** This repository provides a platform for evaluating AI agents in dynamic, evolving environments that require multi-step reasoning and adaptation, building on the Gaia2 benchmark to address real-world challenges. It matters now because as AI agents and large language models advance, there's a growing need for more realistic benchmarks beyond static tests to ensure agents can handle uncertainty and change effectively. _themes: agents · benchmark · rl · evaluation_ #### [openai/parameter-golf](https://github.com/openai/parameter-golf) *Python · ★4,897 · MIT · experimental · score:0.75 · hot:0.62 · rising:0.65 · durable:0.61 · board:rising · trend:stable* Train the smallest LM you can that fits in 16MB. Best model wins! **Why it matters.** This repository is for an OpenAI challenge where participants train the smallest language model that fits in 16MB and trains in under 10 minutes on specific hardware, evaluated by compression performance on a validation set. It matters now because it highlights the growing need for efficient AI models amid resource constraints, potentially inspiring practical advancements in model compression and architecture that could reduce costs and environmental impact in real-world deployments. _themes: model-optimization · compression · efficiency · language-models_ #### [google-deepmind/meltingpot](https://github.com/google-deepmind/meltingpot) *Python · ★817 · Apache-2.0 · beta · score:0.75 · hot:0.62 · rising:0.64 · durable:0.66 · board:durable · trend:stable* A suite of test scenarios for multi-agent reinforcement learning. **Why it matters.** Melting Pot is a suite of over 50 multi-agent reinforcement learning substrates and 256 test scenarios designed to evaluate how well agents generalize to novel social situations, including cooperation, competition, and deception. It matters now because multi-agent RL is increasingly relevant for AI research in social dynamics and generalization, providing a standardized benchmark to compare algorithms and advance fields like cooperative AI, especially with ongoing developments in AI safety and large-scale simulations. _themes: multi-agent · rl · benchmarks · generalization_ #### [llm-jp/awesome-japanese-llm](https://github.com/llm-jp/awesome-japanese-llm) *TypeScript · ★1,380 · Apache-2.0 · beta · score:0.70 · hot:0.61 · rising:0.58 · durable:0.62 · board:durable · trend:stable* 日本語LLMまとめ - Overview of Japanese LLMs **Why it matters.** This repository curates a comprehensive overview of Japanese large language models, including their architectures, licenses, benchmarks, and related resources, making it a centralized hub for information on Japanese NLP. It matters now because the growing demand for localized AI models in non-English languages like Japanese requires accessible evaluations and comparisons, helping researchers and developers avoid redundant efforts in a rapidly evolving field. _themes: japanese-llm · benchmarks · language-models · evaluation_ #### [google-deepmind/long-form-factuality](https://github.com/google-deepmind/long-form-factuality) *Python · ★680 · NOASSERTION · experimental · score:0.70 · hot:0.60 · rising:0.59 · durable:0.52 · board:hot · trend:stable* Benchmarking long-form factuality in large language models. Original code for our paper "Long-form factuality in large language models". **Why it matters.** This repository provides a benchmark dataset called LongFact with 2,280 prompts for testing long-form responses from large language models, along with tools like the Search-Augmented Factuality Evaluator (SAFE) and metrics such as F1@K to assess factual accuracy. It matters now because the proliferation of LLMs generating detailed content has heightened concerns about misinformation, and this work offers a standardized way to evaluate and improve model reliability in real-world applications. _themes: factuality · evaluation · llms · benchmarking_ #### [NVIDIA/nccl-tests](https://github.com/NVIDIA/nccl-tests) *Cuda · ★1,489 · BSD-3-Clause · production · score:0.70 · hot:0.60 · rising:0.63 · durable:0.57 · board:rising · trend:stable* NCCL Tests **Why it matters.** This repository provides tests for NVIDIA's NCCL library, focusing on verifying the performance and correctness of multi-GPU and multi-node communication operations, which are critical for distributed training in AI workloads. It matters now because as large-scale machine learning models demand efficient GPU coordination, these tests help ensure reliability in production environments, though the lack of formal releases may indicate less polished maintenance compared to core NCCL. _themes: distributed-computing · gpu · performance · testing_ #### [google-research/android_world](https://github.com/google-research/android_world) *Python · ★724 · Apache-2.0 · experimental · score:0.70 · hot:0.56 · rising:0.57 · durable:0.54 · board:rising · trend:stable* AndroidWorld is an environment and benchmark for autonomous agents **Why it matters.** AndroidWorld is an open-source environment for building and benchmarking autonomous agents that interact with Android apps, featuring 116 hand-crafted tasks across 20 apps with dynamic variations for robust testing. It matters now because the growing interest in AI agents for mobile interfaces requires standardized, reproducible benchmarks to evaluate performance in real-world scenarios, especially as agent capabilities expand beyond web-based tasks. This tool bridges the gap by integrating with existing benchmarks like MiniWoB++ and supporting easy extensibility. _themes: agents · benchmarking · android · rl_ #### [openai/procgen](https://github.com/openai/procgen) *C++ · ★1,155 · MIT · production · score:0.75 · hot:0.55 · rising:0.61 · durable:0.70 · board:durable · trend:stable* Procgen Benchmark: Procedurally-Generated Game-Like Gym-Environments **Why it matters.** Procgen offers a set of procedurally-generated environments for OpenAI's Gym, designed to benchmark reinforcement learning agents on generalization and sample efficiency, which helps identify limitations in current RL methods. It matters now because robust evaluation of RL agents in randomized settings is essential for advancing AI research, especially as RL applications grow in complexity, though its maintenance status indicates it's a refined but not actively evolving tool. However, with no major updates recently, it may not address the latest RL challenges. _themes: rl · benchmarking · procedural-generation · environments_ #### [vibrantlabsai/ragas](https://github.com/vibrantlabsai/ragas) *Python · ★13,590 · Apache-2.0 · beta · score:0.85 · hot:0.54 · rising:0.59 · durable:0.74 · board:durable · trend:stable* Supercharge Your LLM Application Evaluations 🚀 **Why it matters.** Ragas is a Python library that provides tools for evaluating and optimizing Large Language Model (LLM) applications through objective metrics, automated test data generation, and integrations with frameworks like LangChain. It matters right now because the rapid adoption of LLMs in production requires efficient, data-driven evaluation to address reliability issues and build feedback loops, especially as developers face challenges in scaling LLM operations amid increasing regulatory and performance demands. _themes: llm · evaluation · testing · rag_ #### [google-research/lm-extraction-benchmark](https://github.com/google-research/lm-extraction-benchmark) *Python · ★304 · Apache-2.0 · experimental · score:0.70 · hot:0.54 · rising:0.53 · durable:0.53 · board:hot · trend:stable* **Why it matters.** This repository provides a benchmark for evaluating and improving targeted data extraction attacks on language models, focusing on recovering memorized training data through prefix-suffix matching. It matters now due to increasing privacy risks associated with large language models, as regulatory scrutiny and real-world incidents highlight the need for better defenses against data leakage in AI systems. _themes: privacy · extraction-attacks · llms · benchmark_ #### [google-research/task_adaptation](https://github.com/google-research/task_adaptation) *Python · ★348 · Apache-2.0 · experimental · score:0.60 · hot:0.53 · rising:0.53 · durable:0.51 · board:hot · trend:stable* **Why it matters.** This repository offers a benchmark for evaluating visual models on 19 diverse tasks to assess their transfer learning and adaptation capabilities, drawing from public datasets across various domains. It matters now because robust evaluation of model generalization is critical amid the proliferation of large vision models, though its lack of recent releases and modest 348 stars suggest limited ongoing maintenance and community engagement. _themes: vision · benchmarking · transfer-learning · evaluation_ #### [google-deepmind/clrs](https://github.com/google-deepmind/clrs) *Jupyter Notebook · ★524 · Apache-2.0 · beta · score:0.75 · hot:0.52 · rising:0.57 · durable:0.66 · board:durable · trend:stable* **Why it matters.** The CLRS repository provides a benchmark suite for evaluating machine learning models on classical algorithms from Cormen et al.'s textbook, implementing them as graphs to test algorithmic reasoning capabilities. It matters now as AI research increasingly focuses on symbolic reasoning and code generation, offering a standardized way to assess and improve models in these areas, which is crucial for advancing general AI intelligence. _themes: algorithms · benchmark · ml-reasoning · graphs_ #### [openai/mle-bench](https://github.com/openai/mle-bench) *Python · ★1,473 · NOASSERTION · experimental · score:0.75 · hot:0.51 · rising:0.54 · durable:0.61 · board:durable · trend:stable* MLE-bench is a benchmark for measuring how well AI agents perform at machine learning engineering **Why it matters.** MLE-bench is an open-source benchmark that evaluates AI agents on machine learning engineering tasks, providing tools for dataset construction, evaluation logic, and agent testing as detailed in a related research paper. It matters now because the rapid advancement of AI agents requires standardized metrics to assess their capabilities in practical ML workflows, helping researchers and developers identify strengths and weaknesses amid growing interest in automation. This enables better comparison and improvement of agents in a field that's increasingly critical for AI development. _themes: agents · benchmarking · ml-engineering · evaluation_ #### [huggingface/upskill](https://github.com/huggingface/upskill) *Python · ★481 · Apache-2.0 · beta · score:0.70 · hot:0.50 · rising:0.53 · durable:0.60 · board:durable · trend:stable* Generate and evaluate agent skills for code agents like Claude Code, Open Code, OpenAI Codex **Why it matters.** UPskill is a Python tool for generating and evaluating skills for AI code agents, using teacher models to create capabilities that student models can execute efficiently. It matters now because the growing adoption of AI agents requires cost-effective ways to enhance performance on complex tasks, helping developers optimize without relying on expensive models amid rising competition in AI tooling. _themes: agents · eval · code-agents · model-distillation_ #### [tatsu-lab/alpaca_eval](https://github.com/tatsu-lab/alpaca_eval) *Jupyter Notebook · ★1,972 · Apache-2.0 · beta · score:0.85 · hot:0.49 · rising:0.55 · durable:0.72 · board:durable · trend:down* An automatic evaluator for instruction-following language models. Human-validated, high-quality, cheap, and fast. **Why it matters.** AlpacaEval is an open-source tool for automatically evaluating instruction-following in language models by comparing them to baselines, achieving a high Spearman correlation of 0.98 with human judgments like those from ChatBot Arena, while being fast and costing under $10. It matters now because the LLM landscape demands efficient, cost-effective evaluation methods to benchmark models amid rapid advancements, reducing reliance on expensive human annotators and addressing issues like length bias in prior benchmarks. However, its reliance on proprietary models like GPT-4 for evaluation could limit accessibility and introduce dependencies. _themes: evaluation · llm · benchmarking · instruction-following_ #### [google-deepmind/superhuman](https://github.com/google-deepmind/superhuman) *TeX · ★702 · Apache-2.0 · beta · score:0.80 · hot:0.49 · rising:0.53 · durable:0.62 · board:durable · trend:stable* **Why it matters.** This repository from Google DeepMind's Superhuman Reasoning team hosts projects, datasets, and benchmarks for advanced AI mathematical reasoning, including tools like AlphaGeometry for solving geometry problems and Aletheia for iterative math research. It matters right now because recent achievements in AI solving International Mathematical Olympiad problems highlight progress in reasoning capabilities, potentially advancing AI applications in education, research, and problem-solving amid growing interest in AGI. _themes: math · reasoning · benchmarks · agents_ #### [anthropics/original_performance_takehome](https://github.com/anthropics/original_performance_takehome) *Python · ★3,782 · no-license · experimental · score:0.60 · hot:0.47 · rising:0.51 · durable:0.61 · board:durable · trend:stable* Anthropic's original performance take-home, now open for you to try! **Why it matters.** This repository provides an open-source coding challenge from Anthropic to optimize a simulated machine's performance, allowing users to benchmark their skills against AI models like Claude Opus 4.5 in code optimization tasks. It matters now because it highlights the narrowing gap between AI and human performance in programming, serving as a practical tool for developers to test and improve their abilities amid rapid AI advancements in automation. _themes: optimization · benchmarking · ai-eval · code-generation_ #### [openai/human-eval](https://github.com/openai/human-eval) *Python · ★3,205 · MIT · beta · score:0.80 · hot:0.47 · rising:0.54 · durable:0.65 · board:durable · trend:stable* Code for the paper "Evaluating Large Language Models Trained on Code" **Why it matters.** This repository provides a benchmark dataset and evaluation tools for assessing large language models on code generation tasks, as described in the OpenAI paper. It matters because standardized evaluations are essential for advancing AI research in coding, especially with the growing use of LLMs in software development, though it requires careful handling due to security risks from executing untrusted code. _themes: llm-eval · code-generation · benchmark · ai-research_ #### [allenai/olmes](https://github.com/allenai/olmes) *Python · ★363 · Apache-2.0 · beta · score:0.70 · hot:0.46 · rising:0.49 · durable:0.55 · board:durable · trend:stable* Reproducible, flexible LLM evaluations **Why it matters.** OLMES is an open-source system for reproducible and flexible evaluations of language models, building on EleutherAI's lm-evaluation-harness with enhancements like detailed instance-level data and custom metrics. It matters now because the rapid proliferation of LLMs demands standardized, trustworthy evaluation methods to enable fair comparisons and accelerate research, especially as organizations like AllenAI push for open science in AI. _themes: llm-evaluation · reproducibility · benchmarks · metrics_ #### [allenai/reward-bench](https://github.com/allenai/reward-bench) *Python · ★711 · Apache-2.0 · beta · score:0.70 · hot:0.46 · rising:0.52 · durable:0.69 · board:durable · trend:down* RewardBench: the first evaluation tool for reward models. **Why it matters.** RewardBench is an evaluation tool for assessing reward models used in preference learning and RLHF, providing scripts, datasets, and analysis tools to ensure fair and standardized testing. It matters right now because as AI models increasingly rely on RLHF for alignment and safety, robust benchmarks like this are essential for comparing model performance and advancing techniques like DPO amid growing concerns over AI ethics and reliability. _themes: rlhf · reward-models · evaluation · preference-learning_ #### [huggingface/evaluation-guidebook](https://github.com/huggingface/evaluation-guidebook) *Jupyter Notebook · ★2,092 · NOASSERTION · archived · score:0.60 · hot:0.46 · rising:0.46 · durable:0.57 · board:durable · trend:down* Sharing both practical insights and theoretical knowledge about LLM evaluation that we gathered while managing the Open LLM Leaderboard and designing lighteval! **Why it matters.** This repository provides a guidebook for evaluating Large Language Models, covering basics, practical tips, and theoretical insights based on Hugging Face's experience with benchmarks like the Open LLM Leaderboard. It matters for anyone building or assessing LLMs, as proper evaluation is essential for model reliability and improvement, but its relevance is limited since it is no longer maintained and points to an updated version elsewhere. _themes: evaluation · llm · metrics · tutorial_ #### [huggingface/Math-Verify](https://github.com/huggingface/Math-Verify) *Python · ★1,131 · Apache-2.0 · beta · score:0.70 · hot:0.45 · rising:0.51 · durable:0.69 · board:durable · trend:down* **Why it matters.** Math-Verify is a Python library for parsing and verifying mathematical expressions, specifically designed to evaluate the accuracy of Large Language Model outputs on mathematical tasks, using ANTLR for robust expression handling. It matters right now because as LLMs are increasingly applied to educational and scientific domains, precise evaluation tools like this help identify and improve model weaknesses in math, potentially leading to more reliable AI systems amid growing scrutiny of LLM benchmarks. However, its dependency on specific ANTLR versions may introduce compatibility issues, limiting immediate adoption. _themes: evaluation · math · llm · parsing_ #### [openai/simple-evals](https://github.com/openai/simple-evals) *Python · ★4,445 · MIT · archived · score:0.50 · hot:0.45 · rising:0.49 · durable:0.57 · board:durable · trend:stable* **Why it matters.** This repository provides a lightweight Python library for evaluating language models, focusing on benchmarks like HealthBench, BrowseComp, and SimpleQA, with reference implementations and results for OpenAI's models. It matters for transparency in model accuracy metrics, but its deprecation notice indicates limited future updates, making it less relevant for ongoing development compared to emerging evaluation tools. Researchers can still use it for historical reference, though alternatives are advisable for new projects. _themes: evaluation · benchmarks · llm · transparency_ #### [NVIDIA/RULER](https://github.com/NVIDIA/RULER) *Python · ★1,513 · Apache-2.0 · experimental · score:0.80 · hot:0.44 · rising:0.49 · durable:0.61 · board:durable · trend:down* This repo contains the source code for RULER: What’s the Real Context Size of Your Long-Context Language Models? **Why it matters.** RULER is a Python-based tool that generates synthetic benchmarks to evaluate the actual context sizes of long-context language models, going beyond simple recall to assess real performance on various tasks. It matters now because as models claim increasingly large context windows, verifying their effective capabilities is essential for reliable AI development, especially with the rapid advancement in applications requiring long-term memory. This repo provides code from a research paper to help benchmark 17 open-source models across multiple categories. _themes: evaluation · llm · benchmarking · long-context_ #### [anthropics/evals](https://github.com/anthropics/evals) *? · ★366 · CC-BY-4.0 · experimental · score:0.70 · hot:0.43 · rising:0.47 · durable:0.58 · board:durable · trend:down* **Why it matters.** This repository offers datasets generated by language models to evaluate behaviors like biases, sycophancy, and AI risks, as detailed in a related research paper. It matters now because AI safety and ethical evaluations are critical amid rapid model deployments, providing a practical tool for researchers to probe model flaws using AI-created data. However, its reliance on model-generated content raises questions about potential biases in the evaluations themselves. _themes: ai-safety · eval · datasets · bias_ #### [google-research/camel-prompt-injection](https://github.com/google-research/camel-prompt-injection) *Jupyter Notebook · ★314 · Apache-2.0 · experimental · score:0.60 · hot:0.42 · rising:0.46 · durable:0.58 · board:durable · trend:down* Code for the paper "Defeating Prompt Injections by Design" **Why it matters.** This repository provides code to reproduce experiments from the paper 'Defeating Prompt Injections by Design', implementing a defense mechanism called CaMeL to protect AI models from prompt injection attacks. It matters now because prompt injections remain a critical vulnerability in large language models, and while this experimental work offers insights for researchers, its unmaintained status and potential bugs limit immediate practical application in production environments. _themes: ai-security · prompt-injection · defenses · llms_ #### [google-deepmind/limit](https://github.com/google-deepmind/limit) *Jupyter Notebook · ★647 · Apache-2.0 · experimental · score:0.70 · hot:0.42 · rising:0.46 · durable:0.58 · board:durable · trend:down* On the Theoretical Limitations of Embedding-Based Retrieval **Why it matters.** This repository provides a dataset and code to evaluate the theoretical limitations of embedding-based retrieval systems, demonstrating that certain document-query combinations cannot be accurately retrieved regardless of embedding dimension. It highlights fundamental flaws in current single-vector embedding approaches, which is timely as AI applications increasingly depend on these for search and retrieval, potentially prompting researchers to explore alternative paradigms. However, its impact is theoretical and may not immediately translate to practical improvements without further validation. _themes: embeddings · retrieval · evaluation · limitations_ #### [openai/miniF2F](https://github.com/openai/miniF2F) *Objective-C++ · ★422 · no-license · experimental · score:0.65 · hot:0.40 · rising:0.44 · durable:0.52 · board:durable · trend:down* Formal to Formal Mathematics Benchmark **Why it matters.** MiniF2F is a benchmark that translates Olympiad-level and high-school math problems into formal systems like Lean and Metamath for evaluating automated theorem provers. It matters now because it supports research in AI-driven mathematical reasoning, especially with growing interest in formal verification, but its lack of updates since 2021 and absence of a license limit its practical adoption and community engagement. _themes: theorem-proving · benchmarks · formal-verification · ai-math_ #### [google-research/bleurt](https://github.com/google-research/bleurt) *Python · ★792 · Apache-2.0 · beta · score:0.70 · hot:0.40 · rising:0.44 · durable:0.55 · board:durable · trend:down* BLEURT is a metric for Natural Language Generation based on transfer learning. **Why it matters.** BLEURT is a transfer learning-based metric for evaluating Natural Language Generation by scoring how well a candidate sentence matches a reference in terms of fluency and meaning, using a BERT-derived model trained on ratings data. It matters now because reliable NLG evaluation is essential for improving and benchmarking language models amid rapid AI advancements, though its lack of recent updates may limit its applicability to newer frameworks. This repository provides the necessary code for implementation and fine-tuning, but users should verify its compatibility with current TensorFlow versions. _themes: nlg · evaluation · metrics · bert_ #### [google-research/robustness_metrics](https://github.com/google-research/robustness_metrics) *Jupyter Notebook · ★473 · Apache-2.0 · beta · score:0.60 · hot:0.34 · rising:0.39 · durable:0.51 · board:durable · trend:down* **Why it matters.** This repository provides a library for evaluating the robustness of classification models through metrics for out-of-distribution generalization, stability under perturbations, and uncertainty estimation, using datasets like ImageNetV2. It matters now because as AI models are deployed in real-world scenarios, ensuring their reliability against varied inputs is critical, especially amid growing scrutiny on AI safety and performance in non-ideal conditions; however, its lack of formal releases and dependency on specific TensorFlow versions may limit broader adoption. _themes: robustness · evaluation · uncertainty · metrics_ ### framework (70) #### [langchain-ai/langchain](https://github.com/langchain-ai/langchain) *Python · ★134,515 · MIT · production · score:0.90 · hot:0.96 · rising:0.95 · durable:0.92 · board:hot · trend:up* The agent engineering platform **Why it matters.** LangChain is a Python framework that simplifies building AI agents and LLM-powered applications by chaining components and integrating third-party services, making it easier to handle complex workflows. It matters right now because the rapid evolution of generative AI models like GPT and Gemini requires tools that abstract boilerplate code and ensure future-proof integrations, allowing developers to focus on innovation in a competitive AI landscape. _themes: agents · llm · rag · framework_ #### [run-llama/llama_index](https://github.com/run-llama/llama_index) *Python · ★48,813 · MIT · production · score:0.85 · hot:0.93 · rising:0.93 · durable:0.88 · board:hot · trend:up* LlamaIndex is the leading document agent and OCR platform **Why it matters.** LlamaIndex is an open-source framework for building applications that integrate LLMs with data sources, focusing on tasks like document parsing, OCR, indexing, and retrieval-augmented generation (RAG). It matters now because the growing adoption of LLMs in enterprise settings demands efficient tools for handling unstructured data, and LlamaIndex provides modular components to streamline this, though it relies heavily on external integrations which may introduce complexity. _themes: rag · agents · llm · framework_ #### [infiniflow/ragflow](https://github.com/infiniflow/ragflow) *Python · ★78,763 · Apache-2.0 · production · score:0.80 · hot:0.92 · rising:0.90 · durable:0.83 · board:hot · trend:up* RAGFlow is a leading open-source Retrieval-Augmented Generation (RAG) engine that fuses cutting-edge RAG with Agent capabilities to create a superior context layer for LLMs **Why it matters.** RAGFlow is an open-source engine that combines Retrieval-Augmented Generation (RAG) with agent capabilities to enhance context for large language models, enabling efficient workflows for building AI systems. It matters now because enterprises are increasingly adopting RAG for accurate, context-aware applications, and integrating agents automates complex tasks, addressing the growing need for scalable AI solutions in production environments. _themes: rag · agents · llm · workflow_ #### [langchain-ai/langgraph](https://github.com/langchain-ai/langgraph) *Python · ★30,036 · MIT · production · score:0.85 · hot:0.92 · rising:0.91 · durable:0.84 · board:hot · trend:up* Build resilient language agents as graphs. **Why it matters.** LangGraph is a Python framework for building resilient, stateful language agents using graph-based orchestration, enabling features like durable execution, human-in-the-loop interactions, and comprehensive memory management. It matters right now because AI agents are increasingly critical for enterprise applications, providing tools to handle long-running workflows amid the rapid adoption of LLMs, and it's backed by real-world use in companies like Klarna and Replit. _themes: agents · orchestration · llm · graphs_ #### [crewAIInc/crewAI](https://github.com/crewAIInc/crewAI) *Python · ★49,559 · MIT · production · score:0.80 · hot:0.90 · rising:0.93 · durable:0.85 · board:rising · trend:up* Framework for orchestrating role-playing, autonomous AI agents. By fostering collaborative intelligence, CrewAI empowers agents to work together seamlessly, tackling complex tasks. **Why it matters.** CrewAI is a Python framework for building and orchestrating autonomous AI agents that collaborate on complex tasks, emphasizing role-playing and event-driven control. It matters now because multi-agent systems are increasingly needed for enterprise AI automation, offering a lightweight, independent alternative to more bloated frameworks, though its claims of rapid adoption may be overstated without independent verification. _themes: agents · ai-agents · llms · orchestration_ #### [sgl-project/sglang](https://github.com/sgl-project/sglang) *Python · ★26,272 · Apache-2.0 · beta · score:0.85 · hot:0.89 · rising:0.87 · durable:0.78 · board:hot · trend:up* SGLang is a high-performance serving framework for large language models and multimodal models. **Why it matters.** SGLang is a Python-based framework designed for high-performance serving of large language models and multimodal models, focusing on optimized inference and support for various hardware like NVIDIA GPUs and TPUs. It matters now because it addresses the growing need for efficient AI deployment in production environments, with recent updates providing immediate support for emerging models and hardware advancements, enabling faster processing amid rapid AI innovation. _themes: inference · llm · serving · multimodal_ #### [openai/openai-agents-python](https://github.com/openai/openai-agents-python) *Python · ★22,950 · MIT · beta · score:0.80 · hot:0.88 · rising:0.88 · durable:0.83 · board:rising · trend:up* A lightweight, powerful framework for multi-agent workflows **Why it matters.** This repository offers a Python framework for building multi-agent workflows using LLMs, featuring tools for agent configuration, task handoffs, and safety guardrails, while supporting OpenAI APIs and others. It matters now because multi-agent systems are increasingly relevant for complex AI applications like automation and decision-making, but its early version and OpenAI-centric design may limit broader adoption until more stability and flexibility are proven. _themes: agents · llm · framework · tools_ #### [pytorch/pytorch](https://github.com/pytorch/pytorch) *Python · ★99,352 · NOASSERTION · production · score:1.00 · hot:0.85 · rising:0.90 · durable:0.78 · board:rising · trend:up* Tensors and Dynamic neural networks in Python with strong GPU acceleration **Why it matters.** PyTorch is an open-source library that provides tensor computations similar to NumPy with GPU acceleration and supports building dynamic neural networks using an autograd system. It matters right now because it's a foundational tool for AI research and development, powering advancements in areas like generative AI and large language models, while being actively maintained with regular updates. _themes: deep-learning · neural-networks · gpu-acceleration · autograd_ #### [stanfordnlp/dspy](https://github.com/stanfordnlp/dspy) *Python · ★33,923 · MIT · beta · score:0.80 · hot:0.85 · rising:0.88 · durable:0.78 · board:rising · trend:up* DSPy: The framework for programming—not prompting—language models **Why it matters.** DSPy is a Python framework that enables declarative programming of language models, allowing users to write modular code for AI systems like RAG pipelines and agents while automatically optimizing prompts and weights for better performance. It addresses the limitations of traditional prompting by providing a more structured, teachable approach, which is increasingly important as organizations seek reliable ways to integrate and scale LLMs in production. However, its effectiveness depends on the underlying model's capabilities, and it may require significant setup for complex applications. _themes: rag · agents · optimization · prompting_ #### [keras-team/keras](https://github.com/keras-team/keras) *Python · ★64,013 · Apache-2.0 · production · score:0.90 · hot:0.85 · rising:0.90 · durable:0.83 · board:rising · trend:up* Deep Learning for humans **Why it matters.** Keras 3 is a high-level deep learning framework that provides a unified API for building and training neural networks across multiple backends like JAX, TensorFlow, and PyTorch, simplifying model development for various tasks such as computer vision and NLP. It matters right now because it addresses the need for flexible, performant tools in an increasingly diverse AI ecosystem, enabling faster iteration and scaling while supporting emerging backends amid growing demands for efficient ML workflows. _themes: deep-learning · neural-networks · multi-backend · inference_ #### [microsoft/playwright](https://github.com/microsoft/playwright) *TypeScript · ★86,810 · Apache-2.0 · production · score:0.90 · hot:0.84 · rising:0.89 · durable:0.84 · board:rising · trend:up* Playwright is a framework for Web Testing and Automation. It allows testing Chromium, Firefox and WebKit with a single API. **Why it matters.** Playwright is a TypeScript-based framework for web testing and automation, allowing developers to write scripts that control Chromium, Firefox, and WebKit browsers with a unified API, which is useful for end-to-end testing and AI-driven tasks. It matters right now due to the increasing complexity of web applications and the demand for reliable cross-browser testing in CI/CD pipelines, as well as its growing use in AI agent development for automation. However, its reliance on specific browser engines and potential overkill for simple tasks should be considered. _themes: web-automation · end-to-end-testing · browser-testing · test-automation_ #### [alibaba/MNN](https://github.com/alibaba/MNN) *C++ · ★14,962 · Apache-2.0 · production · score:0.80 · hot:0.83 · rising:0.85 · durable:0.80 · board:rising · trend:up* MNN: A blazing-fast, lightweight inference engine battle-tested by Alibaba, powering high-performance on-device LLMs and Edge AI. **Why it matters.** MNN is a lightweight deep learning inference engine optimized for on-device performance, supporting efficient execution of models like LLMs on embedded devices such as ARM and Vulkan platforms. It matters now due to the increasing focus on edge AI and on-device processing for privacy and speed, with recent updates adding support for modern LLMs like Qwen series, making it relevant for real-world applications in apps and services. _themes: inference · edge-ai · llm · on-device_ #### [vllm-project/vllm-omni](https://github.com/vllm-project/vllm-omni) *Python · ★4,413 · Apache-2.0 · production · score:0.80 · hot:0.83 · rising:0.80 · durable:0.75 · board:hot · trend:up* A framework for efficient model inference with omni-modality models **Why it matters.** vllm-project/vllm-omni is a framework designed for efficient inference and serving of omni-modality models that handle multiple data types like audio, images, and videos, built on PyTorch for better performance in multimodal AI tasks. It matters right now because the rise of multimodal applications demands optimized inference to reduce costs and improve speed, as evidenced by its recent updates enhancing production readiness and broader model support amid growing AI deployment needs. _themes: inference · multimodal · model-serving · pytorch_ #### [microsoft/graphrag](https://github.com/microsoft/graphrag) *Python · ★32,347 · MIT · beta · score:0.85 · hot:0.83 · rising:0.85 · durable:0.83 · board:rising · trend:up* A modular graph-based Retrieval-Augmented Generation (RAG) system **Why it matters.** GraphRAG is a modular framework that uses knowledge graphs to enhance Retrieval-Augmented Generation (RAG) for large language models, enabling better extraction and reasoning from unstructured text, particularly private or narrative data. It matters now because RAG techniques are increasingly critical for improving LLM accuracy in enterprise applications, and graph-based approaches address limitations in traditional methods by providing structured context, though it comes with high computational costs that require careful management. _themes: rag · llm · graphs · retrieval_ #### [crestalnetwork/intentkit](https://github.com/crestalnetwork/intentkit) *Python · ★6,504 · MIT · beta · score:0.75 · hot:0.82 · rising:0.82 · durable:0.74 · board:rising · trend:up* IntentKit is an open-source, self-hosted cloud agent cluster that manages a collaborative team of AI agents for you. **Why it matters.** IntentKit is an open-source framework for managing a cluster of collaborative AI agents in a self-hosted cloud environment, emphasizing security and integrations like blockchain and social media. It matters now as businesses and developers seek scalable, low-maintenance AI solutions amid rising agentic AI adoption, though its reliance on self-hosting and niche features may pose setup challenges for non-experts. _themes: agents · ai-framework · blockchain · web3_ #### [tensorflow/tensorflow](https://github.com/tensorflow/tensorflow) *C++ · ★194,778 · Apache-2.0 · production · score:0.95 · hot:0.82 · rising:0.88 · durable:0.81 · board:rising · trend:up* An Open Source Machine Learning Framework for Everyone **Why it matters.** TensorFlow is an open-source framework for building and deploying machine learning models, particularly neural networks, with APIs in Python, C++, and other languages, supporting distributed training and production deployment. It remains relevant due to its extensive ecosystem and ongoing updates like v2.21.0, which address modern AI needs such as GPU acceleration and scalability, though it can be overly complex for simple tasks compared to lighter alternatives. _themes: deep-learning · neural-networks · machine-learning · distributed_ #### [BrainBlend-AI/atomic-agents](https://github.com/BrainBlend-AI/atomic-agents) *Python · ★5,826 · MIT · beta · score:0.70 · hot:0.81 · rising:0.80 · durable:0.75 · board:hot · trend:up* Building AI agents, atomically **Why it matters.** Atomic Agents is a lightweight, modular framework for building AI agents with single-purpose, reusable components, emphasizing composability and maintainability using tools like Instructor and Pydantic. It matters right now as the AI agent ecosystem expands rapidly with LLMs, providing developers a straightforward way to create reliable pipelines amid increasing complexity in agent-based applications, though its adoption is niche compared to more established alternatives. _themes: agents · llms · modular · composable_ #### [microsoft/agent-framework](https://github.com/microsoft/agent-framework) *Python · ★9,597 · MIT · beta · score:0.75 · hot:0.81 · rising:0.82 · durable:0.73 · board:rising · trend:up* A framework for building, orchestrating and deploying AI agents and multi-agent workflows with support for Python and .NET. **Why it matters.** Microsoft Agent Framework is an open-source tool for building, orchestrating, and deploying AI agents and multi-agent workflows, supporting both Python and .NET environments. It matters now because AI agents are gaining traction for automating complex tasks, and this framework offers a structured approach with Microsoft's backing, though its early release version suggests potential instability for production use. _themes: agents · multi-agent · orchestration · workflows_ #### [NVIDIA/warp](https://github.com/NVIDIA/warp) *Python · ★6,532 · Apache-2.0 · production · score:0.80 · hot:0.78 · rising:0.79 · durable:0.72 · board:rising · trend:up* A Python framework for GPU-accelerated simulation, robotics, and machine learning. **Why it matters.** Warp is a Python framework that JIT compiles regular Python functions into efficient GPU or CPU kernels for simulations, robotics, and machine learning, making it easier to handle high-performance computing tasks. It matters now because the demand for GPU-accelerated, differentiable programming is surging in AI research and robotics development, especially with NVIDIA's ecosystem providing ready access to CUDA hardware. This tool bridges simulation and ML workflows, enabling faster experimentation in fields like physics-based AI. _themes: gpu-acceleration · differentiable-programming · simulation · robotics_ #### [jingyaogong/minimind](https://github.com/jingyaogong/minimind) *Python · ★47,550 · Apache-2.0 · beta · score:0.75 · hot:0.77 · rising:0.83 · durable:0.80 · board:rising · trend:up* 🚀🚀 「大模型」2小时完全从0训练64M的小参数GPT！🌏 Train a 64M-parameter GPT from scratch in just 2h! **Why it matters.** This repo provides a from-scratch implementation to train a small 64M-parameter GPT model using PyTorch, covering the full LLM pipeline from data cleaning to advanced techniques like RLHF, all accessible on consumer hardware in under 2 hours for about $3; it matters now because it democratizes LLM education amid the AI hype, allowing beginners to understand core mechanics without relying on black-box libraries, fostering deeper learning and experimentation in a field dominated by large, inaccessible models. _themes: llm-training · from-scratch · pytorch · fine-tuning_ #### [NVIDIA/physicsnemo](https://github.com/NVIDIA/physicsnemo) *Python · ★2,697 · Apache-2.0 · production · score:0.80 · hot:0.77 · rising:0.79 · durable:0.73 · board:rising · trend:up* Open-source deep-learning framework for building, training, and fine-tuning deep learning models using state-of-the-art Physics-ML methods **Why it matters.** NVIDIA PhysicsNeMo is an open-source framework that enables building, training, and fine-tuning deep learning models incorporating physics-based methods for scientific and engineering applications, optimized for NVIDIA GPUs. It matters now due to the growing demand for AI in simulations and real-time predictions in fields like AI4Science, where integrating physics with ML can accelerate discoveries, and its v2.0 release improves accessibility amid advancing hardware capabilities. _themes: physics-ml · deep-learning · gpu-accelerated · fine-tuning_ #### [microsoft/react-native-windows](https://github.com/microsoft/react-native-windows) *C++ · ★17,236 · NOASSERTION · production · score:0.80 · hot:0.76 · rising:0.78 · durable:0.69 · board:rising · trend:up* A framework for building native Windows apps with React. **Why it matters.** React Native for Windows is a framework that extends React Native to build native applications for Windows platforms, including PCs, tablets, Xbox, and mixed reality devices, using JavaScript and React. It matters right now because it enables developers to leverage cross-platform code for Windows, which remains a key enterprise and gaming market, especially with Microsoft's continued investment in tools like this amid the shift to hybrid app development. _themes: react · cross-platform · windows-apps · native-ui_ #### [microsoft/autogen](https://github.com/microsoft/autogen) *Python · ★57,337 · CC-BY-4.0 · archived · score:0.60 · hot:0.76 · rising:0.77 · durable:0.74 · board:rising · trend:up* A programming framework for agentic AI **Why it matters.** AutoGen is a Python framework for building multi-agent AI systems that enable autonomous operations or human collaboration, primarily through conversational agents powered by LLMs. It matters now because, despite its popularity with over 57,000 stars, it's in maintenance mode with no new features, directing users to migrate to Microsoft Agent Framework for continued development and enterprise support, reflecting the rapid evolution in AI agent technologies. _themes: agents · framework · multi-agent · llm_ #### [microsoft/qlib](https://github.com/microsoft/qlib) *Python · ★40,975 · MIT · beta · score:0.80 · hot:0.75 · rising:0.81 · durable:0.79 · board:rising · trend:up* Qlib is an AI-oriented Quant investment platform that aims to use AI tech to empower Quant Research, from exploring ideas to implementing productions. Qlib supports diverse ML modeling paradigms, including supervised learning, market dynamics modeling, and RL, and is now equipped with https://github.com/microsoft/RD-Agent to automate R&D process. **Why it matters.** Qlib is an AI-oriented platform for quantitative investment that supports various machine learning paradigms like supervised learning, market dynamics modeling, and reinforcement learning, enabling users to transition from research ideas to production systems. It now integrates RD-Agent, an LLM-based tool for automating R&D processes such as factor mining and model optimization, which addresses the increasing demand for efficient AI-driven workflows in finance amid growing adoption of autonomous agents. This makes it particularly relevant as financial institutions seek to leverage AI for competitive edge in algorithmic trading. _themes: agents · finance · ml · optimization_ #### [NVIDIA/earth2studio](https://github.com/NVIDIA/earth2studio) *Python · ★776 · Apache-2.0 · beta · score:0.70 · hot:0.73 · rising:0.73 · durable:0.64 · board:rising · trend:up* Open-source deep-learning framework for exploring, building and deploying AI weather/climate workflows. **Why it matters.** Earth2Studio is an open-source Python framework from NVIDIA that enables users to quickly build, deploy, and experiment with AI-driven workflows for weather and climate modeling using deep learning. It matters now because advancing AI in climate science can improve predictions for real-world issues like extreme weather events, amid growing demands for accurate environmental forecasting and research tools. _themes: deep-learning · weather-forecasting · climate-science · inference_ #### [NVIDIA/TensorRT-Edge-LLM](https://github.com/NVIDIA/TensorRT-Edge-LLM) *C++ · ★349 · Apache-2.0 · beta · score:0.70 · hot:0.72 · rising:0.72 · durable:0.63 · board:hot · trend:up* High-performance, light-weight C++ LLM and VLM Inference Software for Physical AI **Why it matters.** TensorRT Edge-LLM is a C++ inference runtime for deploying large language models and vision-language models on NVIDIA edge devices like Jetson, focusing on high performance in resource-constrained environments. It matters now as edge AI applications grow for real-time processing in areas like autonomous systems, but its NVIDIA-specific nature limits versatility and broader adoption compared to more hardware-agnostic alternatives. The project is still maturing, with modest community interest indicated by 349 stars, suggesting it's not yet a go-to solution for all edge inference needs. _themes: inference · edge-ai · llm · vlm_ #### [microsoft/lisa](https://github.com/microsoft/lisa) *Python · ★324 · MIT · production · score:0.60 · hot:0.72 · rising:0.73 · durable:0.62 · board:rising · trend:up* LISA is developed and maintained by Microsoft, to empower Linux validation. **Why it matters.** LISA is a Python framework for automating end-to-end testing and validation of Linux distributions, focusing on kernel and compatibility issues across platforms like Azure and HyperV. It matters now because reliable Linux testing is critical for cloud deployments amid growing enterprise reliance on hybrid environments, though its Microsoft-centric design limits broader appeal. Adoption could help reduce testing overhead in large-scale operations, but its niche focus means it's not revolutionary. _themes: testing · automation · linux · cloud_ #### [microsoft/microsoft-ui-xaml](https://github.com/microsoft/microsoft-ui-xaml) *C++ · ★7,431 · MIT · production · score:0.80 · hot:0.72 · rising:0.76 · durable:0.71 · board:rising · trend:up* WinUI: a modern UI framework with a rich set of controls and styles to build dynamic and high-performing Windows applications. **Why it matters.** WinUI is a UI framework for building modern Windows applications using C++ or C#, providing Fluent Design controls and styles to ensure intuitive and performant user experiences. It matters now because Microsoft continues to evolve the Windows ecosystem, making WinUI essential for developers targeting Windows 10 and later to leverage new platform features and maintain compatibility. However, its relevance is primarily limited to Windows-specific development, which may not appeal broadly amid the rise of cross-platform tools. _themes: ui-framework · windows-dev · xaml · fluent-design_ #### [NVIDIA/bionemo-framework](https://github.com/NVIDIA/bionemo-framework) *Jupyter Notebook · ★728 · no-license · beta · score:0.70 · hot:0.69 · rising:0.69 · durable:0.63 · board:rising · trend:up* BioNeMo Framework: For building and adapting AI models in drug discovery at scale **Why it matters.** NVIDIA's BioNeMo Framework provides tools and libraries for accelerating AI model development in drug discovery, focusing on GPU-optimized workflows for biomolecular applications. It matters now because the pharmaceutical sector is rapidly integrating AI to reduce costs and speed up drug development, and this framework offers specialized, high-performance capabilities that could enhance efficiency in a competitive field. _themes: drug-discovery · gpu-acceleration · pytorch · fine-tuning_ #### [huggingface/candle](https://github.com/huggingface/candle) *Rust · ★20,043 · Apache-2.0 · beta · score:0.70 · hot:0.69 · rising:0.72 · durable:0.64 · board:rising · trend:up* Minimalist ML framework for Rust **Why it matters.** Candle is a minimalist machine learning framework for Rust that provides core tensor operations and GPU support, enabling efficient model inference and basic training. It matters now as Rust gains traction for performance-critical applications, offering a lightweight alternative to Python-based frameworks for developers seeking safety and speed, especially with Hugging Face's demos showcasing practical use cases like Whisper and LLaMA2. _themes: rust · ml-framework · inference · gpu_ #### [facebookresearch/ai4animationpy](https://github.com/facebookresearch/ai4animationpy) *Python · ★938 · NOASSERTION · experimental · score:0.70 · hot:0.67 · rising:0.66 · durable:0.60 · board:hot · trend:stable* A Python framework for AI-driven character animation using neural networks. **Why it matters.** AI4AnimationPy is a Python framework that enables AI-driven character animation by providing tools for motion capture processing, neural network training, and inference, all without relying on Unity. It matters now because it bridges the gap for researchers and developers working in Python ecosystems, potentially streamlining workflows in animation AI amid growing demands for efficient, engine-agnostic tools in fields like gaming and virtual reality. However, its lack of a formal release and unspecified license may hinder immediate adoption. _themes: animation · neural-networks · inference · pytorch_ #### [google-deepmind/open_spiel](https://github.com/google-deepmind/open_spiel) *C++ · ★5,156 · Apache-2.0 · production · score:0.70 · hot:0.66 · rising:0.70 · durable:0.76 · board:durable · trend:stable* OpenSpiel is a collection of environments and algorithms for research in general reinforcement learning and search/planning in games. **Why it matters.** OpenSpiel provides a robust collection of game environments and algorithms for reinforcement learning research, supporting a wide range of game types from single-agent to multi-agent scenarios with perfect and imperfect information. It matters now because reinforcement learning in games is crucial for advancing AI in strategic decision-making, and with growing interest in multi-agent systems for applications like AI safety and autonomous agents, this tool enables detailed analysis and experimentation in these areas. _themes: reinforcement-learning · multiagent · games · algorithms_ #### [facebookresearch/ReAgent](https://github.com/facebookresearch/ReAgent) *Python · ★3,694 · BSD-3-Clause · archived · score:0.40 · hot:0.64 · rising:0.63 · durable:0.50 · board:hot · trend:up* A platform for Reasoning systems (Reinforcement Learning, Contextual Bandits, etc.) **Why it matters.** ReAgent is an open-source platform for applied reinforcement learning, offering tools for training algorithms like DQN variants, data preprocessing, and model serving using PyTorch, but it has been archived and is no longer maintained. While it provides historical insights into Facebook's RL approaches and could be useful for educational purposes or legacy projects, its relevance is diminished due to lack of updates, and users should opt for more current alternatives to avoid potential issues. _themes: reinforcement-learning · bandits · pytorch · decision-making_ #### [facebookresearch/mmf](https://github.com/facebookresearch/mmf) *Python · ★5,626 · NOASSERTION · beta · score:0.70 · hot:0.64 · rising:0.64 · durable:0.66 · board:durable · trend:stable* A modular framework for vision & language multimodal research from Facebook AI Research (FAIR) **Why it matters.** MMF is a PyTorch-based framework for building and training models in vision and language multimodal research, offering modular components for datasets, models, and distributed training. It matters right now because multimodal AI is increasingly critical for tasks like visual question answering and hate speech detection, providing researchers with a practical tool from FAIR to accelerate development amid growing interest in integrated vision-language systems. _themes: multimodal · pytorch · vision-language · deep-learning_ #### [allenai/WildDet3D](https://github.com/allenai/WildDet3D) *Python · ★375 · NOASSERTION · experimental · score:0.60 · hot:0.64 · rising:0.62 · durable:0.56 · board:hot · trend:stable* Allen Institute for AI: WildDet3D: Scaling Promptable 3D Detection in the Wild **Why it matters.** WildDet3D is a research-oriented framework for scaling promptable 3D object detection in real-world, uncontrolled environments, enabling users to detect and manipulate 3D objects via natural language prompts. It matters right now because 3D vision technologies are critical for applications like autonomous driving and robotics, and adding promptability could make these systems more adaptable amid the growing integration of foundation models in AI workflows, though its lack of a formal release limits immediate applicability. _themes: 3d-detection · promptable-ai · computer-vision · scaling_ #### [wenet-e2e/wenet](https://github.com/wenet-e2e/wenet) *Python · ★5,089 · Apache-2.0 · production · score:0.80 · hot:0.64 · rising:0.67 · durable:0.77 · board:durable · trend:stable* Production First and Production Ready End-to-End Speech Recognition Toolkit **Why it matters.** WeNet is an end-to-end speech recognition toolkit built with PyTorch, offering production-ready solutions for accurate transcription using models like Conformer, Transformer, and Whisper variants, making it suitable for real-world applications. It matters now because speech recognition is increasingly critical for AI-driven interfaces, and WeNet's focus on ease of use and deployment addresses the growing need for accessible, high-performance tools amid advancements in multimodal AI. _themes: speech-recognition · asr · pytorch · production_ #### [facebookresearch/spider](https://github.com/facebookresearch/spider) *Python · ★417 · NOASSERTION · experimental · score:0.65 · hot:0.63 · rising:0.62 · durable:0.55 · board:hot · trend:stable* A general physic-based retargeting framework. **Why it matters.** SPIDER is a Python-based framework for physics-informed retargeting of human motions to diverse robots, such as dexterous hands and humanoids, using a pipeline from human videos to robot actions in simulation. It matters now because it addresses the growing need for scalable human-to-robot motion transfer in robotics research, potentially aiding in tasks like manipulation and control, though its experimental nature and lack of a formal release may hinder broader application. _themes: robotics · simulation · physics-based · retargeting_ #### [google-deepmind/simply](https://github.com/google-deepmind/simply) *Python · ★518 · Apache-2.0 · experimental · score:0.70 · hot:0.63 · rising:0.63 · durable:0.57 · board:hot · trend:stable* Minimal and scalable research codebase in JAX, designed for rapid iteration on frontier research in LLM and other autoregressive models. **Why it matters.** This repo provides a minimal, scalable codebase in JAX for rapid experimentation with LLMs and autoregressive models, emphasizing quick iteration for both human and AI-driven research. It matters now because the AI field is advancing quickly, and tools that enable fast, low-overhead development can accelerate innovation in frontier LLM research, especially with the rising interest in automated AI agents. _themes: jax · llm · research · agents_ #### [facebookresearch/map-anything](https://github.com/facebookresearch/map-anything) *Python · ★3,221 · Apache-2.0 · beta · score:0.75 · hot:0.62 · rising:0.66 · durable:0.74 · board:durable · trend:stable* MapAnything: Universal Feed-Forward Metric 3D Reconstruction **Why it matters.** MapAnything is an open-source framework for universal metric 3D reconstruction, using a transformer model to regress 3D geometry from inputs like images or depth, supporting over 12 tasks such as SfM and depth estimation in a unified way. It matters now because 3D reconstruction is increasingly critical for applications in robotics and AR/VR, and this tool simplifies experimentation by integrating multiple models into a modular stack, though its research focus means it may require refinement for production use. _themes: 3d-reconstruction · inference · depth-estimation · multi-view-stereo_ #### [huggingface/open-r1](https://github.com/huggingface/open-r1) *Python · ★25,997 · Apache-2.0 · experimental · score:0.80 · hot:0.62 · rising:0.65 · durable:0.69 · board:durable · trend:stable* Fully open reproduction of DeepSeek-R1 **Why it matters.** This repository offers an open-source reproduction of the DeepSeek-R1 model, including scripts for training, fine-tuning, and synthetic data generation, aiming to make advanced AI pipelines accessible for community collaboration. It matters now as it promotes transparency and innovation in AI research amid growing demands for open alternatives to proprietary models, potentially accelerating progress in model distillation and synthetic data techniques. _themes: fine-tuning · synthetic-data · model-training · open-source_ #### [The-Pocket/PocketFlow](https://github.com/The-Pocket/PocketFlow) *Python · ★10,435 · MIT · experimental · score:0.60 · hot:0.60 · rising:0.59 · durable:0.66 · board:durable · trend:stable* Pocket Flow: 100-line LLM framework. Let Agents build Agents! **Why it matters.** PocketFlow is a minimalist 100-line Python framework for building and managing LLM-based agents that can create other agents, focusing on simplicity with features like multi-agents, workflows, and RAG. It matters now because agentic AI is gaining traction for complex applications, and this lightweight approach reduces dependencies, making it easier for developers to experiment without vendor lock-in; however, its lack of formal releases raises questions about stability and scalability in real-world scenarios. _themes: agents · rag · llm · workflow_ #### [AkaliKong/MiniOneRec](https://github.com/AkaliKong/MiniOneRec) *Python · ★1,460 · Apache-2.0 · beta · score:0.60 · hot:0.59 · rising:0.61 · durable:0.62 · board:durable · trend:stable* Minimal reproduction of OneRec **Why it matters.** MiniOneRec is an open-source framework that provides a minimal reproduction of the OneRec system, offering an end-to-end workflow for generative recommendation including SID construction, supervised fine-tuning, and reinforcement learning. It matters right now because it enables researchers to experiment with scaling generative AI for recommendations using LLMs, though it suffers from reproducibility issues due to dependency problems and bugs, as noted in recent announcements. _themes: generative-ai · recommendation · llm · fine-tuning_ #### [openai/chatkit-js](https://github.com/openai/chatkit-js) *TypeScript · ★1,900 · Apache-2.0 · production · score:0.70 · hot:0.58 · rising:0.61 · durable:0.69 · board:durable · trend:stable* **Why it matters.** ChatKit is a TypeScript framework from OpenAI that provides a ready-to-use, customizable chat interface for building AI-powered conversations, including features like response streaming and tool integration. It matters now because the surge in AI chat applications demands efficient development tools, allowing developers to integrate advanced chat experiences quickly without building core components from scratch, amid growing competition in generative AI interfaces. _themes: chat · agents · inference · ui_ #### [EasyJailbreak/EasyJailbreak](https://github.com/EasyJailbreak/EasyJailbreak) *Python · ★838 · GPL-3.0 · beta · score:0.70 · hot:0.56 · rising:0.57 · durable:0.65 · board:durable · trend:stable* An easy-to-use Python framework to generate adversarial jailbreak prompts. **Why it matters.** EasyJailbreak is a Python framework that streamlines the creation of adversarial prompts to bypass safety mechanisms in large language models, breaking the process into modular steps for easier experimentation and evaluation. It matters now because LLM security vulnerabilities are a growing concern amid widespread AI adoption, and this tool helps researchers systematically test and improve defenses against jailbreaks. _themes: jailbreak · llm-security · adversarial-attacks · optimization_ #### [google-research/t5x](https://github.com/google-research/t5x) *Python · ★2,962 · Apache-2.0 · beta · score:0.70 · hot:0.54 · rising:0.56 · durable:0.59 · board:durable · trend:stable* **Why it matters.** T5X is a framework for high-performance training, evaluation, and inference of sequence models like language models, built on JAX and Flax as an updated version of the original T5 codebase. It matters now because it supports scalable, research-oriented workflows amid the growing demand for efficient large-scale AI experiments, especially with the shift towards JAX for better performance and flexibility in model development. _themes: jax · flax · sequence-models · training_ #### [microsoft/BitNet](https://github.com/microsoft/BitNet) *Python · ★38,412 · MIT · beta · score:0.80 · hot:0.53 · rising:0.61 · durable:0.74 · board:durable · trend:stable* Official inference framework for 1-bit LLMs **Why it matters.** BitNet is Microsoft's official inference framework for 1-bit Large Language Models (LLMs), providing optimized kernels for fast and lossless inference on CPUs and GPUs, with significant speedups and energy reductions that enable running large models on resource-constrained devices. It matters now because the AI industry is increasingly focused on efficiency and sustainability, making this framework timely for deploying LLMs in edge computing, mobile apps, and environments with limited power, potentially reducing barriers to advanced AI adoption amid growing environmental concerns. _themes: inference · quantization · efficiency · llms_ #### [google-deepmind/xmanager](https://github.com/google-deepmind/xmanager) *Python · ★905 · Apache-2.0 · beta · score:0.70 · hot:0.52 · rising:0.55 · durable:0.63 · board:durable · trend:stable* A platform for managing machine learning experiments **Why it matters.** XManager is a framework for packaging, running, and tracking machine learning experiments, supporting local environments and Google Cloud Platform via Python scripts. It matters now because ML experiment management is essential for reproducibility and scalability in research workflows, helping teams handle increasing complexity without reinventing basic infrastructure. _themes: mlops · experiment-management · orchestration · cloud-computing_ #### [xai-org/x-algorithm](https://github.com/xai-org/x-algorithm) *Rust · ★16,342 · Apache-2.0 · production · score:0.80 · hot:0.52 · rising:0.62 · durable:0.77 · board:durable · trend:stable* Algorithm powering the For You feed on X **Why it matters.** This repository contains the algorithm for X's For You feed, which uses a Grok-based transformer model to rank posts from followed accounts and broader discoveries based on user engagement history, eliminating most hand-engineered features for a purely ML-driven approach. It matters because it offers a rare glimpse into a production-scale social media recommendation system, potentially useful for developers building similar systems, though its heavy reliance on X's proprietary data and infrastructure limits immediate applicability. As an open-source release from xAI, it could influence ML practices in content recommendation but requires significant adaptation for other use cases. _themes: recommendation · transformer · ranking · ml_ #### [google-deepmind/mujoco_mpc](https://github.com/google-deepmind/mujoco_mpc) *C++ · ★1,605 · Apache-2.0 · beta · score:0.70 · hot:0.52 · rising:0.56 · durable:0.65 · board:durable · trend:stable* Real-time behaviour synthesis with MuJoCo, using Predictive Control **Why it matters.** MuJoCo MPC is a framework for real-time predictive control in robotics simulations using MuJoCo, allowing users to author and solve complex tasks like quadruped locomotion and manipulation via methods such as iLQG and Predictive Sampling. It matters now because advancements in simulation-based control are critical for developing efficient AI-driven robotics, especially with growing interest in embodied AI and real-world applications from organizations like Google DeepMind. However, its early release stage and specific focus limit immediate broad adoption. _themes: mpc · robotics · simulation · control_ #### [google-deepmind/lab](https://github.com/google-deepmind/lab) *C · ★7,350 · NOASSERTION · production · score:0.60 · hot:0.51 · rising:0.59 · durable:0.70 · board:durable · trend:stable* A customisable 3D platform for agent-based AI research **Why it matters.** DeepMind Lab is a customizable 3D environment based on Quake III Arena for testing AI agents in navigation and puzzle-solving tasks, primarily for deep reinforcement learning research. It serves as a benchmark for AI evaluation but hasn't been updated since 2020, potentially limiting its relevance amid rapid advancements in newer simulation frameworks. Its established use in academic research makes it valuable for historical comparisons, though alternatives may offer more modern features. _themes: reinforcement-learning · agents · simulation · deep-learning_ #### [facebookresearch/hydra](https://github.com/facebookresearch/hydra) *Python · ★10,323 · MIT · production · score:0.85 · hot:0.49 · rising:0.58 · durable:0.72 · board:durable · trend:stable* Hydra is a framework for elegantly configuring complex applications **Why it matters.** Hydra is a Python framework that enables elegant configuration management for complex applications, allowing users to handle overrides, compositions, and hierarchical configs with ease. It matters right now because configuration complexity is a growing pain point in AI/ML workflows, where experiments often involve numerous parameters, and Hydra's integration with tools like PyTorch helps streamline development and reproducibility. Its ecosystem of extensions further enhances its utility for modern ML projects. _themes: config · ml-workflows · python · experimentation_ #### [huggingface/ratchet](https://github.com/huggingface/ratchet) *Rust · ★756 · MIT · experimental · score:0.40 · hot:0.48 · rising:0.47 · durable:0.49 · board:durable · trend:stable* A cross-platform browser ML framework. **Why it matters.** Ratchet is a Rust-based framework for running GPU-accelerated ML inference in web browsers and natively, focusing on models like Whisper and Phi with features like quantization and lazy computation. It matters now because it addresses the need for lightweight, cross-platform AI integration in web apps, but its early development stage means it's not yet reliable for production use. _themes: inference · webgpu · quantization · cross-platform_ #### [deepseek-ai/smallpond](https://github.com/deepseek-ai/smallpond) *Python · ★4,947 · MIT · beta · score:0.70 · hot:0.47 · rising:0.54 · durable:0.67 · board:durable · trend:stable* A lightweight data processing framework built on DuckDB and 3FS. **Why it matters.** Smallpond is a lightweight data processing framework that uses DuckDB for high-performance SQL queries and 3FS for scalable storage, allowing users to handle petabyte-scale datasets with simple Python operations. It matters now because the growing demand for efficient big data processing in AI and analytics requires tools that are easy to use and performant without managing complex infrastructure, as evidenced by its strong benchmark results. However, the lack of official releases raises questions about its stability and readiness for production. _themes: data-processing · duckdb · scalability · big-data_ #### [allenai/ai2thor](https://github.com/allenai/ai2thor) *C# · ★1,710 · Apache-2.0 · production · score:0.80 · hot:0.47 · rising:0.54 · durable:0.67 · board:durable · trend:stable* An open-source platform for Visual AI. **Why it matters.** AI2-THOR is an open-source platform that provides simulated environments for visual AI research, featuring interactive scenes, objects, agents, and various sensory data to support tasks in embodied AI and robotics. It matters right now because embodied AI simulations are essential for advancing computer vision and real-world interactions, especially amid growing interest in robotics and AI safety, with its extensive features aiding in bridging simulation to real-world applications. _themes: simulation · computer-vision · embodied-ai · robotics_ #### [facebookresearch/home-robot](https://github.com/facebookresearch/home-robot) *Python · ★1,199 · MIT · beta · score:0.70 · hot:0.46 · rising:0.51 · durable:0.65 · board:durable · trend:down* Mobile manipulation research tools for roboticists **Why it matters.** HomeRobot is an open-source toolkit for mobile manipulation research, enabling low-cost robots to perform tasks like exploring environments and manipulating objects in unknown settings via the OVMM challenge. It matters now because it supports the ongoing CVPR 2024 challenge, fostering innovation in robotics AI amid growing interest in practical, affordable robotic applications, though its focus on specific hardware limits broader adoption. _themes: robotics · manipulation · exploration · ai-challenge_ #### [google-research/big_vision](https://github.com/google-research/big_vision) *Jupyter Notebook · ★3,422 · Apache-2.0 · experimental · score:0.80 · hot:0.46 · rising:0.50 · durable:0.62 · board:durable · trend:stable* Official codebase used to develop Vision Transformer, SigLIP, MLP-Mixer, LiT and more. **Why it matters.** Big Vision is a codebase for training large-scale vision models like Vision Transformer and SigLIP using Jax and Flax on TPUs or GPUs, focusing on scalable and reproducible experiments. It matters now because it provides a starting point for vision research with state-of-the-art models, though its maintenance is limited to internal Google use, making it less reliable for external adoption. _themes: vision · transformer · training · jax_ #### [huggingface/picotron](https://github.com/huggingface/picotron) *Python · ★2,153 · Apache-2.0 · experimental · score:0.70 · hot:0.45 · rising:0.50 · durable:0.62 · board:durable · trend:down* Minimalistic 4D-parallelism distributed training framework for education purpose **Why it matters.** Picotron is a simplified framework for distributed training of large language models using 4D parallelism, focused on educational purposes to teach techniques like data, tensor, pipeline, and context parallelism. It matters now because accessible tools for learning distributed training are increasingly needed amid growing AI complexity, though its performance is not yet optimal and it's still in development. _themes: distributed-training · parallelism · education · llm_ #### [google-research/circuit_training](https://github.com/google-research/circuit_training) *Python · ★1,643 · Apache-2.0 · production · score:0.80 · hot:0.45 · rising:0.51 · durable:0.63 · board:durable · trend:stable* **Why it matters.** AlphaChip is an open-source framework that uses distributed deep reinforcement learning to automate chip floorplan generation, replicating a methodology from a 2021 Nature paper to optimize semiconductor design processes. It matters now because the growing complexity of AI hardware demands efficient design tools, and this framework demonstrates practical RL applications in engineering, potentially accelerating innovation in chip development amid increasing demand for custom AI chips. _themes: reinforcement-learning · deep-learning · chip-design · automation_ #### [facebookresearch/metaquery](https://github.com/facebookresearch/metaquery) *Python · ★317 · NOASSERTION · experimental · score:0.60 · hot:0.44 · rising:0.46 · durable:0.57 · board:durable · trend:down* Official Implementation of Paper Transfer between Modalities with MetaQueries **Why it matters.** MetaQuery is a research implementation for transferring knowledge between different data modalities, such as text and images, using meta-learning techniques as described in the associated paper. It matters now because multimodal AI is advancing rapidly, and efficient cross-modal transfer can reduce training costs and improve model generalization in applications like vision-language tasks, though its impact is still niche and unproven at scale. _themes: meta-learning · multimodal · transfer-learning · fine-tuning_ #### [allenai/ScienceWorld](https://github.com/allenai/ScienceWorld) *Scala · ★351 · Apache-2.0 · beta · score:0.70 · hot:0.44 · rising:0.48 · durable:0.62 · board:durable · trend:down* ScienceWorld is a text-based virtual environment centered around accomplishing tasks from the standardized elementary science curriculum. **Why it matters.** ScienceWorld is a text-based simulation environment designed to evaluate AI agents on elementary science tasks, drawing from standardized curricula to test reasoning and language capabilities. It matters now because it provides a benchmark for assessing whether language models can perform at a basic educational level, amid growing interest in AI's practical applications in education and reinforcement learning, especially as models like GPT variants advance in natural language understanding. _themes: reinforcement-learning · language-models · text-games · evaluation_ #### [stanford-crfm/mistral](https://github.com/stanford-crfm/mistral) *Python · ★579 · Apache-2.0 · beta · score:0.60 · hot:0.42 · rising:0.47 · durable:0.60 · board:durable · trend:down* Mistral: A strong, northwesterly wind: Framework for transparent and accessible large-scale language model training, built with Hugging Face 🤗 Transformers. **Why it matters.** Mistral is a framework designed to facilitate transparent and accessible training of large-scale language models using Hugging Face Transformers, including tools for dataset integration, distributed training on platforms like GCP, and evaluation scripts. It addresses the growing need for reproducible AI research amid increasing scrutiny on model development practices, though its specific focus on certain setups may limit broader appeal. However, as AI transparency becomes a priority, it offers practical utilities for those building or fine-tuning models. _themes: fine-tuning · distributed-training · evaluation · transformers_ #### [allenai/lumos](https://github.com/allenai/lumos) *Python · ★476 · MIT · experimental · score:0.70 · hot:0.42 · rising:0.45 · durable:0.60 · board:durable · trend:down* Code and data for "Lumos: Learning Agents with Unified Data, Modular Design, and Open-Source LLMs" **Why it matters.** Lumos is a framework for building language agents that use modular components for planning, grounding, and execution, leveraging open-source LLMs like LLaMA-2 and a unified dataset for training on complex interactive tasks. It achieves performance comparable to proprietary models like GPT-4 on tasks such as web QA and math, which matters now as it promotes accessible, open-source alternatives amid growing interest in agentic AI, potentially accelerating research and reducing reliance on closed ecosystems. However, its lack of a formal release and reliance on experimental data generation may limit immediate applicability. _themes: agents · llms · modular-design · training-data_ #### [NVIDIA/GR00T-Dreams](https://github.com/NVIDIA/GR00T-Dreams) *Jupyter Notebook · ★524 · Apache-2.0 · experimental · score:0.70 · hot:0.42 · rising:0.45 · durable:0.57 · board:durable · trend:down* DreamGen: Nvidia GEAR Lab's initiative to solve the robotics data problem using world models **Why it matters.** This repository provides a pipeline for generating synthetic trajectory data for robots using NVIDIA's Cosmos world models, allowing robots to learn new tasks from a single image and language instructions without real-world data collection. It matters now because robotics is facing a data scarcity issue that hinders scalable training, and this approach could accelerate development, though it's limited to NVIDIA's ecosystem and remains unproven at scale. _themes: robotics · synthetic-data · world-models · fine-tuning_ #### [facebookresearch/deepconf](https://github.com/facebookresearch/deepconf) *Python · ★385 · MIT · experimental · score:0.50 · hot:0.42 · rising:0.44 · durable:0.55 · board:durable · trend:down* DeepConf: Deep Think with Confidence **Why it matters.** DeepConf is a Python framework that builds on vLLM to enable parallel inference for reasoning tasks like math and coding, incorporating voting strategies and confidence-based early stopping to enhance output reliability. It matters for addressing LLM inaccuracies in critical applications, but its experimental nature, lack of a formal release, and modest 384 stars indicate it's not yet proven or widely adopted, potentially limiting its immediate utility. Overall, it's a niche extension for researchers refining LLM workflows rather than a transformative tool. _themes: inference · llm · parallel · voting_ #### [google-deepmind/streetlearn](https://github.com/google-deepmind/streetlearn) *C++ · ★320 · Apache-2.0 · experimental · score:0.60 · hot:0.42 · rising:0.44 · durable:0.53 · board:durable · trend:down* A C++/Python implementation of the StreetLearn environment based on images from Street View, as well as a TensorFlow implementation of goal-driven navigation agents solving the task published in “Learning to Navigate in Cities Without a Map”, NeurIPS 2018 **Why it matters.** StreetLearn provides a C++ and Python environment for training AI agents to navigate real-world street environments using Google Street View images, based on a 2018 NeurIPS paper, and includes TensorFlow implementations for goal-driven navigation tasks. It matters for research in reinforcement learning and urban navigation, but its age and lack of recent updates mean it may not fully address current advancements in scalable RL frameworks or modern datasets. However, it remains a solid resource for studying map-free navigation in realistic settings. _themes: reinforcement-learning · navigation · computer-vision · deep-learning_ #### [allenai/unified-io-2](https://github.com/allenai/unified-io-2) *Python · ★645 · Apache-2.0 · beta · score:0.70 · hot:0.41 · rising:0.47 · durable:0.62 · board:durable · trend:down* **Why it matters.** This repository provides code for Unified-IO 2, a multimodal model framework based on T5X that supports training, inference, and demos using JAX and PyTorch, with additional components like an audio tokenizer. It matters for researchers exploring unified AI across modalities, but its reliance on older dependencies and lack of a formal release may introduce compatibility issues, making it less immediately practical for production use. The recent updates in early 2024 highlight ongoing advancements in multimodal AI, though the untested GPU setups limit broader accessibility. _themes: multimodal · jax · inference · training_ #### [EleutherAI/DALLE-mtf](https://github.com/EleutherAI/DALLE-mtf) *Python · ★431 · MIT · experimental · score:0.60 · hot:0.41 · rising:0.43 · durable:0.57 · board:durable · trend:down* Open-AI's DALL-E for large scale training in mesh-tensorflow. **Why it matters.** This repository provides an open-source implementation of OpenAI's DALL-E using Mesh-Tensorflow, enabling large-scale training of text-to-image models on TPUs. It matters now because it democratizes access to advanced multimodal AI research, allowing developers and researchers to experiment with generative models amid the rapid growth of AI applications, though it's still in early stages and requires significant setup. _themes: text-to-image · transformers · multimodal · large-scale-training_ #### [allenai/allenact](https://github.com/allenai/allenact) *Python · ★380 · NOASSERTION · beta · score:0.70 · hot:0.41 · rising:0.44 · durable:0.58 · board:durable · trend:down* An open source framework for research in Embodied-AI from AI2. **Why it matters.** AllenAct is an open-source framework for Embodied-AI research, offering modular tools for environments, tasks, and algorithms like PPO in areas such as computer vision and reinforcement learning. It matters now because Embodied-AI is advancing rapidly in robotics and simulation, enabling researchers to build and test agents that interact with real-world-like settings, though its niche focus limits broader appeal. Adoption is supported by AI2's backing and documentation, but the NOASSERTION license may raise concerns for enterprise use. _themes: embodied-ai · reinforcement-learning · computer-vision · deep-learning_ #### [allenai/Holodeck](https://github.com/allenai/Holodeck) *Python · ★538 · Apache-2.0 · experimental · score:0.65 · hot:0.40 · rising:0.43 · durable:0.53 · board:durable · trend:down* CVPR 2024: Language Guided Generation of 3D Embodied AI Environments. **Why it matters.** Holodeck is a research tool for generating 3D environments in embodied AI using language inputs, leveraging large language models and AI2-THOR for text-to-3D creation; it's based on a CVPR 2024 paper and focuses on simulation for AI agents. It matters now because the intersection of generative AI and 3D worlds is growing in importance for robotics and virtual environments research, but its experimental nature and platform dependencies may hinder broader adoption. _themes: generative-ai · llms · text-to-3d · embodied-ai_ #### [google-research/recsim](https://github.com/google-research/recsim) *Python · ★783 · Apache-2.0 · archived · score:0.50 · hot:0.39 · rising:0.41 · durable:0.49 · board:durable · trend:down* A Configurable Recommender Systems Simulation Platform **Why it matters.** RecSim is a simulation platform for recommender systems that enables researchers to model sequential user interactions and test reinforcement learning algorithms in controlled environments. It matters for advancing RL techniques in recommendations, but its last update was in 2019, making it potentially outdated compared to current tools that incorporate recent AI advancements like transformer-based models. _themes: reinforcement-learning · simulation · recommender-system · tensorflow_ ### infra (29) #### [mudler/LocalAI](https://github.com/mudler/LocalAI) *Go · ★45,704 · MIT · production · score:0.85 · hot:0.90 · rising:0.92 · durable:0.86 · board:rising · trend:up* LocalAI is the open-source AI engine. Run any model - LLMs, vision, voice, image, video - on any hardware. No GPU required. **Why it matters.** LocalAI is an open-source engine that enables running various AI models like LLMs, vision, and voice processing on local hardware without requiring a GPU, emphasizing privacy and compatibility with popular APIs. It matters now because growing privacy concerns and the high costs of cloud-based AI make local alternatives essential for developers and organizations seeking control over their data and infrastructure, though its broad backend support may introduce complexity in setup and optimization. _themes: inference · agents · privacy · multi-modal_ #### [chroma-core/chroma](https://github.com/chroma-core/chroma) *Rust · ★27,586 · Apache-2.0 · production · score:0.85 · hot:0.86 · rising:0.88 · durable:0.79 · board:rising · trend:up* Data infrastructure for AI **Why it matters.** Chroma is an open-source vector database written in Rust that enables efficient storage, indexing, and querying of vector embeddings for AI applications, particularly for similarity search and retrieval-augmented generation. It matters now because the rapid adoption of LLMs and AI agents demands scalable infrastructure for handling high-dimensional data, and Chroma provides a straightforward API that simplifies prototyping and deployment, though it faces competition from more established databases. However, its focus on ease-of-use and integration with Python makes it a solid choice for developers building AI features, but it may lack advanced enterprise features out of the box. _themes: rag · agents · vector-db · search_ #### [qdrant/qdrant](https://github.com/qdrant/qdrant) *Rust · ★30,571 · Apache-2.0 · production · score:0.80 · hot:0.84 · rising:0.87 · durable:0.78 · board:rising · trend:up* Qdrant - High-performance, massive-scale Vector Database and Vector Search Engine for the next generation of AI. Also available in the cloud https://cloud.qdrant.io/ **Why it matters.** Qdrant is a high-performance vector database and search engine written in Rust, designed for storing and querying vectors with payloads, supporting applications like semantic search, recommendations, and hybrid filtering. It matters now because the growing demand for AI-driven similarity search in areas like large language models and computer vision requires scalable, efficient solutions, and Qdrant provides a reliable, open-source alternative to proprietary services with strong performance benchmarks. _themes: vector-search · ai-search · embeddings · recommender_ #### [weaviate/weaviate](https://github.com/weaviate/weaviate) *Go · ★16,059 · BSD-3-Clause · production · score:0.85 · hot:0.84 · rising:0.84 · durable:0.73 · board:rising · trend:up* Weaviate is an open-source vector database that stores both objects and vectors, allowing for the combination of vector search with structured filtering with the fault tolerance and scalability of a cloud-native database​. **Why it matters.** Weaviate is an open-source vector database that enables efficient storage and querying of vectors with structured data, supporting features like hybrid search and RAG for applications such as semantic search and recommendation systems. It matters now because the growing demand for AI-driven search in production environments requires scalable, fault-tolerant solutions, but its value depends on specific use cases where vector search needs to integrate with traditional databases. However, users should critically assess its performance against alternatives for real-world scalability and cost. _themes: rag · vector-search · hybrid-search_ #### [alibaba/OpenSandbox](https://github.com/alibaba/OpenSandbox) *Python · ★10,097 · Apache-2.0 · production · score:0.85 · hot:0.83 · rising:0.84 · durable:0.80 · board:rising · trend:up* Secure, Fast, and Extensible Sandbox runtime for AI agents. **Why it matters.** OpenSandbox provides a secure, extensible sandbox runtime for AI agents, supporting multi-language SDKs and runtimes like Docker and Kubernetes to enable safe execution for tasks such as code interpretation and agent evaluation. It matters now because the growing adoption of AI agents in production environments demands robust security and isolation to mitigate risks like code vulnerabilities, while offering scalability for enterprise-level deployments. _themes: agents · security · sandbox · ai-infra_ #### [e2b-dev/E2B](https://github.com/e2b-dev/E2B) *Python · ★11,781 · Apache-2.0 · beta · score:0.80 · hot:0.82 · rising:0.81 · durable:0.78 · board:hot · trend:up* Open-source, secure environment with real-world tools for enterprise-grade agents. **Why it matters.** E2B provides a secure, cloud-based sandbox for executing AI-generated code, allowing developers to run and manage isolated environments for AI agents safely. It matters now as the proliferation of AI agents in enterprises demands robust security measures to mitigate risks from untrusted code, making it essential for building reliable and compliant AI applications. _themes: agents · code-interpreter · llm · security_ #### [apache/airflow](https://github.com/apache/airflow) *Python · ★45,092 · Apache-2.0 · production · score:0.85 · hot:0.82 · rising:0.86 · durable:0.74 · board:rising · trend:up* Apache Airflow - A platform to programmatically author, schedule, and monitor workflows **Why it matters.** Apache Airflow is an open-source platform for programmatically defining, scheduling, and monitoring workflows, primarily used for data pipelines and ETL processes. It matters right now because reliable workflow orchestration is essential for scaling data engineering tasks in production environments, especially amid growing data volumes in AI and ML projects, though it can be complex to configure and maintain for beginners. Its Python-based approach makes it a practical choice for teams needing automation beyond simple scripting. _themes: orchestration · data-pipelines · etl · scheduling_ #### [trycua/cua](https://github.com/trycua/cua) *HTML · ★13,510 · MIT · beta · score:0.80 · hot:0.81 · rising:0.81 · durable:0.75 · board:rising · trend:up* Open-source infrastructure for Computer-Use Agents. Sandboxes, SDKs, and benchmarks to train and evaluate AI agents that can control full desktops (macOS, Linux, Windows). **Why it matters.** Cua provides open-source infrastructure, including sandboxes, SDKs, and benchmarks, for training and evaluating AI agents that can control full desktops on macOS, Linux, and Windows, enabling safe and isolated interactions like screen capture and mouse/keyboard operations. It matters now because the demand for autonomous AI agents in automation is growing, but reliable, cross-platform tools for development and testing are scarce, making Cua a practical solution for advancing agent reliability without the risks of direct hardware access. _themes: agents · desktop-automation · virtualization · ai-infra_ #### [NVIDIA/OpenShell](https://github.com/NVIDIA/OpenShell) *Rust · ★5,097 · Apache-2.0 · experimental · score:0.70 · hot:0.78 · rising:0.78 · durable:0.71 · board:hot · trend:up* OpenShell is the safe, private runtime for autonomous AI agents. **Why it matters.** OpenShell provides a secure, sandboxed runtime for executing autonomous AI agents, using YAML policies to prevent unauthorized access and data leaks. It matters now because the proliferation of AI agents raises significant security concerns in enterprise and personal use, and NVIDIA's involvement could drive adoption of safer agent frameworks. However, as it's still in alpha with limitations like single-player mode, it's an early but promising step toward addressing these risks. _themes: agents · security · sandbox · runtime_ #### [NVIDIA/aistore](https://github.com/NVIDIA/aistore) *Go · ★1,821 · MIT · production · score:0.75 · hot:0.78 · rising:0.80 · durable:0.74 · board:rising · trend:up* AIStore: scalable storage for AI applications **Why it matters.** AIStore is a distributed storage system built for AI applications, providing scalable, high-performance storage with features like linear scalability, multi-cloud access, and Kubernetes integration. It matters now as AI workloads demand efficient data management for ML training and processing, especially with the rise of large-scale models and multi-cloud environments, where traditional storage often falls short in performance and flexibility. However, its adoption is niche, primarily benefiting those already in distributed systems rather than offering groundbreaking innovation. _themes: distributed-storage · ml-training · multi-cloud · high-performance_ #### [NVIDIA/gpu-operator](https://github.com/NVIDIA/gpu-operator) *Go · ★2,650 · Apache-2.0 · production · score:0.85 · hot:0.78 · rising:0.79 · durable:0.70 · board:rising · trend:up* NVIDIA GPU Operator creates, configures, and manages GPUs in Kubernetes **Why it matters.** The NVIDIA GPU Operator automates the deployment, configuration, and management of NVIDIA GPUs within Kubernetes clusters, handling components like drivers, device plugins, and monitoring tools to simplify hardware integration. This is particularly relevant now as AI and machine learning workloads increasingly rely on scalable GPU resources, and with Kubernetes' widespread adoption, it reduces operational errors and eases cluster scaling in dynamic environments like cloud or on-prem setups. _themes: gpu · kubernetes · cuda · monitoring_ #### [microsoft/azurelinux](https://github.com/microsoft/azurelinux) *RPM Spec · ★4,600 · MIT · production · score:0.80 · hot:0.77 · rising:0.80 · durable:0.71 · board:rising · trend:up* Linux OS for Azure 1P services and edge appliances **Why it matters.** Azure Linux is a lightweight, customizable Linux distribution developed by Microsoft specifically for Azure cloud services and edge appliances, featuring a small core set of packages that can be extended for various workloads. It matters right now because the growing demand for efficient, resource-constrained operating systems in edge computing and cloud environments makes this open-source offering from Microsoft a practical option for developers and operators seeking to optimize their infrastructure, while also reflecting Microsoft's ongoing commitment to Linux in hybrid cloud scenarios. _themes: linux · cloud · edge · opensource_ #### [microsoft/TypeScript-Website](https://github.com/microsoft/TypeScript-Website) *TypeScript · ★2,500 · CC-BY-4.0 · production · score:0.75 · hot:0.76 · rising:0.78 · durable:0.69 · board:rising · trend:up* The Website and web infrastructure for learning TypeScript **Why it matters.** This repository hosts the official website for TypeScript, including its documentation, tutorials, and web infrastructure, serving as the primary resource for developers learning the language. It matters now because TypeScript is integral to modern JavaScript development, and updates to the site ensure users stay informed on evolving features, especially with recent releases like TypeScript 6. _themes: typescript · documentation · web-infra · education_ #### [microsoft/WSL2-Linux-Kernel](https://github.com/microsoft/WSL2-Linux-Kernel) *C · ★10,301 · NOASSERTION · production · score:0.80 · hot:0.75 · rising:0.78 · durable:0.74 · board:rising · trend:up* The source for the Linux kernel used in Windows Subsystem for Linux 2 (WSL2) **Why it matters.** This repository provides the source code for the customized Linux kernel used in Windows Subsystem for Linux 2 (WSL2), enabling Windows users to run a Linux environment with better performance than its predecessor. It matters because it supports cross-platform development workflows, but its relevance is limited to Microsoft ecosystems and may lag behind upstream kernel updates, potentially introducing compatibility issues for advanced users. _themes: kernel · virtualization · cross-platform · development_ #### [NVIDIA/k8s-device-plugin](https://github.com/NVIDIA/k8s-device-plugin) *Go · ★3,723 · Apache-2.0 · production · score:0.80 · hot:0.75 · rising:0.78 · durable:0.69 · board:rising · trend:up* NVIDIA device plugin for Kubernetes **Why it matters.** This repository provides a device plugin for Kubernetes that enables the allocation and management of NVIDIA GPUs to pods, allowing for GPU-accelerated workloads in containerized environments. It matters right now because the growing adoption of AI and machine learning in production requires efficient GPU resource handling in orchestrated systems, though it's limited to NVIDIA hardware and demands careful setup to prevent compatibility issues. _themes: gpu · kubernetes · orchestration · acceleration_ #### [microsoft/WSL](https://github.com/microsoft/WSL) *C++ · ★31,907 · MIT · production · score:0.90 · hot:0.72 · rising:0.78 · durable:0.74 · board:rising · trend:up* Windows Subsystem for Linux **Why it matters.** WSL enables running unmodified Linux command-line tools and applications directly on Windows, eliminating the need for virtual machines or dual booting. It matters now because it supports cross-platform development in a Windows-dominated enterprise landscape, where tools like Docker and cloud services increasingly require Linux compatibility, though it still faces limitations in performance and full Linux feature parity. _themes: virtualization · cross-platform · linux · windows_ #### [NVIDIA/open-gpu-kernel-modules](https://github.com/NVIDIA/open-gpu-kernel-modules) *C · ★16,909 · NOASSERTION · production · score:0.85 · hot:0.70 · rising:0.73 · durable:0.75 · board:durable · trend:up* NVIDIA Linux open GPU kernel module source **Why it matters.** This repository provides the source code for NVIDIA's open GPU kernel modules for Linux, allowing users to build and install custom versions for supported architectures, but it requires specific firmware and user-space drivers, limiting its standalone utility. It matters because it enhances compatibility and customization for NVIDIA GPUs in Linux environments, which is relevant for AI and high-performance computing, though the NOASSERTION license raises questions about true openness and integration ease. Critics note that it's not a full open-source driver, potentially fragmenting the ecosystem. _themes: gpu · kernel · linux · driver_ #### [punkpeye/awesome-mcp-servers](https://github.com/punkpeye/awesome-mcp-servers) *? · ★85,073 · MIT · beta · score:0.75 · hot:0.69 · rising:0.71 · durable:0.72 · board:durable · trend:up* A collection of MCP servers. **Why it matters.** This repository curates a list of servers implementing the Model Context Protocol (MCP), which enables AI models to securely access external resources like files, databases, and APIs for enhanced contextual interactions. It matters right now as AI applications increasingly require standardized protocols for real-time data integration to improve accuracy and functionality, especially in production environments. However, its value is limited to the quality and maintenance of the listed servers, with no recent releases potentially indicating stagnation. _themes: ai-protocols · servers · integration · contextual-ai_ #### [skyzh/tiny-llm](https://github.com/skyzh/tiny-llm) *Python · ★4,099 · Apache-2.0 · experimental · score:0.70 · hot:0.67 · rising:0.66 · durable:0.65 · board:hot · trend:stable* A course of learning LLM inference serving on Apple Silicon for systems engineers: build a tiny vLLM + Qwen. **Why it matters.** This repository provides a hands-on course for systems engineers to build a simplified LLM inference server using MLX on Apple Silicon, focusing on core components like attention and KV caching with the Qwen model. It matters because it offers accessible learning for those without NVIDIA GPUs, amid increasing interest in efficient local LLM serving, though it's limited to educational purposes and may not be optimized for production use. The project emphasizes from-scratch implementation to deepen understanding of underlying techniques, but its incomplete roadmap and lack of formal releases mean it's still evolving. _themes: llm · inference · serving · education_ #### [Kong/kong](https://github.com/Kong/kong) *Lua · ★43,220 · Apache-2.0 · production · score:0.85 · hot:0.66 · rising:0.71 · durable:0.82 · board:durable · trend:stable* 🦍 The API and AI Gateway **Why it matters.** Kong is an open-source API gateway that handles routing, security, and management for traditional APIs as well as AI traffic, including LLMs and MCP, through its extensible plugin architecture. It matters right now because the rapid adoption of AI services demands scalable, secure gateways to manage complex traffic, ensure semantic security, and integrate with cloud-native environments like Kubernetes, making it essential for modern application development amid the AI boom. _themes: api-gateway · ai-gateway · llm · kubernetes_ #### [microsoft/mcp-gateway](https://github.com/microsoft/mcp-gateway) *C# · ★590 · MIT · beta · score:0.70 · hot:0.65 · rising:0.64 · durable:0.62 · board:hot · trend:stable* MCP Gateway is a reverse proxy and management layer for MCP servers, enabling scalable, session-aware stateful routing and lifecycle management of MCP servers in Kubernetes environments. **Why it matters.** MCP Gateway is a reverse proxy and management layer for Model Context Protocol (MCP) servers, handling scalable, session-aware routing and lifecycle management in Kubernetes environments, which is crucial for AI workloads. It matters now because the growing complexity of AI applications, especially with LLMs, demands reliable infrastructure for session management and server scaling to avoid downtime in production. This tool from Microsoft integrates with Azure, addressing real-world needs for enterprise-grade AI deployment but lacks a formal release, potentially limiting immediate adoption. _themes: ai · gateway · kubernetes · routing_ #### [NVIDIA/dcgm-exporter](https://github.com/NVIDIA/dcgm-exporter) *Go · ★1,685 · Apache-2.0 · production · score:0.80 · hot:0.64 · rising:0.66 · durable:0.68 · board:durable · trend:stable* NVIDIA GPU metrics exporter for Prometheus leveraging DCGM **Why it matters.** DCGM-Exporter is a tool that collects and exposes NVIDIA GPU metrics to Prometheus using DCGM, allowing for real-time monitoring of GPU health, utilization, and performance in data centers and Kubernetes setups. It matters now because the growing demand for GPU resources in AI/ML and cloud computing requires reliable monitoring to optimize workloads and prevent downtime, especially as organizations scale their infrastructure. However, it's specialized for NVIDIA hardware, so its relevance depends on specific ecosystem needs. _themes: monitoring · gpu · prometheus · kubernetes_ #### [NVIDIA/dgx-spark-playbooks](https://github.com/NVIDIA/dgx-spark-playbooks) *Jupyter Notebook · ★732 · Apache-2.0 · beta · score:0.70 · hot:0.62 · rising:0.64 · durable:0.59 · board:rising · trend:stable* Collection of step-by-step playbooks for setting up AI/ML workloads on NVIDIA DGX Spark devices with Blackwell architecture. **Why it matters.** This repository provides a collection of step-by-step playbooks for installing, configuring, and running AI/ML workloads on NVIDIA's DGX Spark devices with Blackwell architecture, focusing on practical setups for frameworks and inference. It matters now because Blackwell hardware is newly released, offering advanced AI performance, and these guides help users quickly adopt it for real-world tasks in a competitive AI landscape where hardware optimization is key. _themes: inference · fine-tuning · hardware-setup · ai-frameworks_ #### [deepseek-ai/3FS](https://github.com/deepseek-ai/3FS) *C++ · ★9,808 · MIT · beta · score:0.80 · hot:0.58 · rising:0.62 · durable:0.69 · board:durable · trend:stable* A high-performance distributed file system designed to address the challenges of AI training and inference workloads. **Why it matters.** 3FS is a high-performance distributed file system optimized for AI training and inference, using SSDs and RDMA networks to provide strong consistency and efficient data access without requiring custom APIs. It matters now because the growing demands of large-scale AI workloads necessitate scalable, locality-oblivious storage that simplifies development and reduces costs, especially as organizations grapple with data management in distributed environments. _themes: distributed-fs · ai-training · inference · high-performance_ #### [openai/codex-universal](https://github.com/openai/codex-universal) *Dockerfile · ★872 · no-license · beta · score:0.60 · hot:0.53 · rising:0.55 · durable:0.61 · board:durable · trend:stable* Base docker image used in Codex environments **Why it matters.** This repository provides a base Docker image that approximates the OpenAI Codex environment for local development and debugging, allowing users to customize and test setups without relying on the cloud. It matters because it facilitates offline experimentation with AI-assisted coding tools, which can accelerate development cycles amid growing adoption of AI platforms, though it's not a perfect replica and may introduce inconsistencies. _themes: docker · ai-env · development · debugging_ #### [NVIDIA/Cosmos](https://github.com/NVIDIA/Cosmos) *? · ★8,100 · no-license · archived · score:0.10 · hot:0.48 · rising:0.51 · durable:0.59 · board:durable · trend:stable* New repo collection for NVIDIA Cosmos: https://github.com/nvidia-cosmos **Why it matters.** This repository was for NVIDIA's Cosmos project, likely a collection of tools or frameworks related to AI and computing, but it has been deprecated and is no longer maintained, directing users to an archived branch. It matters primarily for historical reference to the initial CES 2025 release, though it's not relevant for current development or adoption. _themes: ai · nvidia · archived · infra_ #### [deepseek-ai/open-infra-index](https://github.com/deepseek-ai/open-infra-index) *? · ★7,974 · CC0-1.0 · production · score:0.80 · hot:0.48 · rising:0.55 · durable:0.74 · board:durable · trend:stable* Production-tested AI infrastructure tools for efficient AGI development and community-driven innovation **Why it matters.** This repository acts as an index for DeepSeek AI's open-sourced AI infrastructure tools, such as optimized GPU kernels for inference, which are production-tested to enhance efficiency in AGI development. It matters right now because it promotes community-driven innovation by sharing battle-tested components that address scaling challenges in hardware for AI, helping developers and researchers accelerate progress in a rapidly evolving AI landscape. _themes: inference · gpu-kernels · ai-infra · open-source_ #### [microsoft/AI-System](https://github.com/microsoft/AI-System) *Python · ★4,258 · CC-BY-4.0 · beta · score:0.70 · hot:0.47 · rising:0.53 · durable:0.65 · board:durable · trend:stable* System for AI Education Resource. **Why it matters.** This repository provides educational resources for a course on AI systems, focusing on the design and optimization of computer systems that support AI, including deep learning lifecycle and hands-on experiments. It matters now because the rapid advancement of AI relies on strong system foundations, and this fills a gap in AI education by emphasizing systems thinking, which is increasingly critical for developing efficient AI applications amid growing computational demands. _themes: ai-education · systems · deep-learning · optimization_ #### [google-research/mint](https://github.com/google-research/mint) *Python · ★558 · Apache-2.0 · experimental · score:0.65 · hot:0.38 · rising:0.41 · durable:0.52 · board:durable · trend:down* Multi-modal Content Creation Model Training Infrastructure including the FACT model (AI Choreographer) implementation. **Why it matters.** This repo offers the implementation and training infrastructure for the AI Choreographer model, which generates 3D dance movements conditioned on music using the AIST++ dataset, as presented in a 2021 ICCV paper. It matters now for researchers exploring multi-modal generative AI in creative domains, but its age, lack of releases, and complex setup make it less accessible or immediately applicable compared to newer advancements. _themes: multi-modal · generation · dance · training_ ### library (316) #### [vllm-project/vllm](https://github.com/vllm-project/vllm) *Python · ★77,729 · Apache-2.0 · production · score:0.90 · hot:0.92 · rising:0.91 · durable:0.85 · board:hot · trend:up* A high-throughput and memory-efficient inference and serving engine for LLMs **Why it matters.** vLLM is a library designed for high-throughput and memory-efficient inference and serving of large language models, featuring optimizations like PagedAttention and continuous batching to handle resource-intensive tasks effectively. It matters right now because the growing demand for scalable AI deployments in production environments requires tools that reduce costs and improve performance, making vLLM a practical solution for real-world applications amid the rapid expansion of generative AI. _themes: inference · llm-serving · quantization · optimization_ #### [huggingface/transformers](https://github.com/huggingface/transformers) *Python · ★159,759 · Apache-2.0 · production · score:0.95 · hot:0.91 · rising:0.94 · durable:0.87 · board:rising · trend:up* 🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training. **Why it matters.** Hugging Face Transformers is a Python library that provides easy-to-use implementations of transformer models for natural language processing, computer vision, audio, and multimodal tasks, supporting both inference and training with pre-trained models. It matters right now because it lowers the barrier to adopting state-of-the-art AI techniques amid the AI boom, enabling rapid prototyping and deployment, though it requires users to manage frequent updates and potential compatibility issues with evolving frameworks like PyTorch. _themes: nlp · deep-learning · inference · fine-tuning_ #### [huggingface/peft](https://github.com/huggingface/peft) *Python · ★20,988 · Apache-2.0 · production · score:0.90 · hot:0.89 · rising:0.90 · durable:0.86 · board:rising · trend:up* 🤗 PEFT: State-of-the-art Parameter-Efficient Fine-Tuning. **Why it matters.** Hugging Face's PEFT library enables parameter-efficient fine-tuning of large pretrained models by updating only a small subset of parameters, significantly reducing computational and storage costs while maintaining performance comparable to full fine-tuning. This matters now because the growing scale of AI models makes traditional fine-tuning impractical for many users, and PEFT democratizes access to advanced model adaptation amid rising demands for efficient AI development and deployment. _themes: fine-tuning · llm · lora · efficiency_ #### [ggml-org/llama.cpp](https://github.com/ggml-org/llama.cpp) *C++ · ★105,751 · MIT · production · score:0.90 · hot:0.88 · rising:0.94 · durable:0.84 · board:rising · trend:up* LLM inference in C/C++ **Why it matters.** ggml-org/llama.cpp is a C++ library for efficient inference of large language models, leveraging the GGML tensor library to enable CPU/GPU execution with low resource requirements, making it suitable for local deployments and edge devices. It matters now due to growing demand for lightweight AI inference amid privacy concerns and hardware constraints, with recent updates enhancing integration with tools like Hugging Face and adding support for new models and multimodal features, though it may lack comprehensive documentation for beginners. _themes: inference · llm · c++ · quantization_ #### [NVIDIA/TensorRT-LLM](https://github.com/NVIDIA/TensorRT-LLM) *Python · ★13,451 · NOASSERTION · production · score:0.90 · hot:0.88 · rising:0.86 · durable:0.76 · board:hot · trend:up* TensorRT LLM provides users with an easy-to-use Python API to define Large Language Models (LLMs) and supports state-of-the-art optimizations to perform inference efficiently on NVIDIA GPUs. TensorRT LLM also contains components to create Python and C++ runtimes that orchestrate the inference execution in a performant way. **Why it matters.** TensorRT-LLM is a library from NVIDIA that offers a Python API for defining and optimizing large language models for efficient inference on NVIDIA GPUs, including support for advanced features like distributed parallelism and sparse attention. It matters now because the growing demand for high-performance LLM deployment in production environments requires hardware-specific optimizations to handle large-scale models effectively, especially with the latest GPU advancements like Blackwell. This tool helps reduce inference latency and costs, making it essential for real-world AI applications. _themes: inference · llm · gpu-optimization · pytorch_ #### [milvus-io/milvus](https://github.com/milvus-io/milvus) *Go · ★43,924 · Apache-2.0 · production · score:0.90 · hot:0.88 · rising:0.89 · durable:0.80 · board:rising · trend:up* Milvus is a high-performance, cloud-native vector database built for scalable vector ANN search **Why it matters.** Milvus is an open-source vector database designed for high-performance approximate nearest neighbor searches on large-scale datasets, particularly for AI applications involving embeddings from text, images, and multi-modal data. It matters right now because the growing adoption of large language models, RAG pipelines, and generative AI demands efficient, scalable solutions for vector similarity searches, and Milvus's distributed architecture supports real-time updates and horizontal scaling to meet these needs. _themes: vector-search · anns · rag · llm_ #### [HKUDS/LightRAG](https://github.com/HKUDS/LightRAG) *Python · ★33,824 · MIT · production · score:0.85 · hot:0.87 · rising:0.89 · durable:0.84 · board:rising · trend:up* [EMNLP2025] "LightRAG: Simple and Fast Retrieval-Augmented Generation" **Why it matters.** LightRAG is a Python library for implementing Retrieval-Augmented Generation (RAG) systems that emphasize simplicity and speed, integrating features like knowledge graphs, reranking, and scalability for efficient handling of large datasets. It matters now because RAG is increasingly vital for improving LLM accuracy in real-world applications, but its README's future-dated updates (e.g., 2026 features) raise skepticism about the timeline and practicality of these enhancements. _themes: rag · llm · knowledge-graph · scalability_ #### [mem0ai/mem0](https://github.com/mem0ai/mem0) *Python · ★53,516 · Apache-2.0 · production · score:0.85 · hot:0.87 · rising:0.89 · durable:0.85 · board:rising · trend:up* Universal memory layer for AI Agents **Why it matters.** Mem0 provides a universal memory layer for AI agents, allowing them to store, manage, and retrieve long-term memories to improve context and personalization in interactions. It matters now because the rapid growth of AI agents and LLMs highlights the need for efficient memory management, and this repo's latest updates demonstrate substantial benchmark improvements, making it a practical solution for building more capable applications. However, its focus on single-pass retrieval and entity linking may limit flexibility in complex, multi-step scenarios. _themes: agents · rag · memory · llm_ #### [MemPalace/mempalace](https://github.com/MemPalace/mempalace) *Python · ★48,057 · MIT · production · score:0.80 · hot:0.87 · rising:0.90 · durable:0.84 · board:rising · trend:up* The best-benchmarked open-source AI memory system. And it's free. **Why it matters.** MemPalace is an open-source Python library for local AI memory management, storing conversation history as verbatim text with semantic search and a pluggable backend like ChromaDB, allowing scoped queries without summarization or external APIs. It matters now because it addresses privacy concerns in LLM applications by keeping data local, and its strong benchmarks on tasks like LongMemEval make it relevant for developers building efficient, offline AI systems amid rising demand for Retrieval-Augmented Generation tools. _themes: memory · rag · llm · semantic-search_ #### [browser-use/browser-use](https://github.com/browser-use/browser-use) *Python · ★88,609 · MIT · beta · score:0.80 · hot:0.86 · rising:0.89 · durable:0.83 · board:rising · trend:up* 🌐 Make websites accessible for AI agents. Automate tasks online with ease. **Why it matters.** Browser-use is a Python library that facilitates browser automation for AI agents using Playwright, allowing them to interact with websites for tasks like data extraction and navigation. It integrates LLMs to enable autonomous agent workflows, which is increasingly relevant amid the rapid adoption of AI agents for automation, though its reliance on external APIs and cloud services may introduce dependencies and privacy concerns. This tool addresses a growing need for streamlined web access in AI development, but users should evaluate its stability given the beta-level releases. _themes: agents · browser-automation · llm · python_ #### [ultralytics/ultralytics](https://github.com/ultralytics/ultralytics) *Python · ★56,152 · AGPL-3.0 · production · score:0.90 · hot:0.86 · rising:0.90 · durable:0.84 · board:rising · trend:up* Ultralytics YOLO 🚀 **Why it matters.** Ultralytics/ultralytics is a Python library providing state-of-the-art YOLO models for computer vision tasks like object detection, instance segmentation, and pose estimation, built on PyTorch for ease of use and high performance. It matters right now because YOLO remains a cornerstone in real-time applications such as autonomous driving and surveillance, with this repo offering frequent updates, multilingual documentation, and community support to make advanced CV accessible amid growing AI demands. Its AGPL-3.0 license and enterprise options address both open-source and commercial needs in a competitive landscape. _themes: object-detection · computer-vision · deep-learning · inference_ #### [ScrapeGraphAI/Scrapegraph-ai](https://github.com/ScrapeGraphAI/Scrapegraph-ai) *Python · ★23,336 · MIT · production · score:0.80 · hot:0.85 · rising:0.87 · durable:0.82 · board:rising · trend:up* Python scraper based on AI **Why it matters.** ScrapeGraphAI is a Python library that leverages large language models and graph logic to simplify web scraping and data extraction from various sources like websites and documents, allowing users to specify what data they want in natural language. It matters now because the growing adoption of AI for automation, especially in RAG pipelines, makes efficient scraping crucial for developers building AI-driven applications, potentially reducing the complexity of traditional scraping methods. However, its reliance on LLMs could introduce variability and ethical concerns in data handling. _themes: llm · web-scraping · rag · data-extraction_ #### [Unstructured-IO/unstructured](https://github.com/Unstructured-IO/unstructured) *HTML · ★14,530 · Apache-2.0 · production · score:0.85 · hot:0.84 · rising:0.86 · durable:0.79 · board:rising · trend:up* Convert documents to structured data effortlessly. Unstructured is open-source ETL solution for transforming complex documents into clean, structured formats for language models. Visit our website to learn more about our enterprise grade Platform product for production grade workflows, partitioning, enrichments, chunking and embedding. **Why it matters.** Unstructured is an open-source library that processes and converts unstructured documents like PDFs, HTML, and Word files into structured data formats ready for language models, streamlining ETL workflows for AI applications. It matters right now because the growing demand for efficient document handling in LLMs and RAG systems addresses key bottlenecks in data preparation, enabling faster development of AI-driven tools. However, its reliance on a separate enterprise platform for advanced features may limit accessibility for some users. _themes: rag · document-processing · llm · etl_ #### [PaddlePaddle/PaddleOCR](https://github.com/PaddlePaddle/PaddleOCR) *Python · ★75,931 · Apache-2.0 · production · score:0.85 · hot:0.84 · rising:0.88 · durable:0.83 · board:rising · trend:up* Turn any PDF or image document into structured data for your AI. A powerful, lightweight OCR toolkit that bridges the gap between images/PDFs and LLMs. Supports 100+ languages. **Why it matters.** PaddleOCR is an open-source toolkit that extracts text and structures from PDFs and images, converting them into structured formats like JSON or Markdown for easy integration with LLMs, supporting over 100 languages with high accuracy. It matters now because the growing demand for RAG and agentic AI applications requires efficient document parsing to handle real-world challenges like warped or skewed documents, making it a practical solution for bridging visual data and AI workflows. Its production-ready efficiency and widespread adoption by projects like Dify position it as a reliable tool in the expanding multimodal AI landscape. _themes: ocr · document-parsing · rag · inference_ #### [axolotl-ai-cloud/axolotl](https://github.com/axolotl-ai-cloud/axolotl) *Python · ★11,740 · Apache-2.0 · beta · score:0.80 · hot:0.83 · rising:0.84 · durable:0.75 · board:rising · trend:up* Go ahead and axolotl questions **Why it matters.** Axolotl is a Python library designed for fine-tuning large language models, offering features like model quantization, LoRA support, and integration with various architectures to optimize training efficiency. It matters right now because the proliferation of LLMs requires accessible tools for customization, and its recent updates address resource constraints and emerging model types, making it relevant for efficient AI development amid hardware limitations. However, its rapid update cycle suggests ongoing instability, so users should verify compatibility before adoption. _themes: fine-tuning · llm · quantization · inference_ #### [hiyouga/LlamaFactory](https://github.com/hiyouga/LlamaFactory) *Python · ★70,476 · Apache-2.0 · production · score:0.90 · hot:0.83 · rising:0.87 · durable:0.86 · board:rising · trend:up* Unified Efficient Fine-Tuning of 100+ LLMs & VLMs (ACL 2024) **Why it matters.** LlamaFactory offers a unified framework for efficiently fine-tuning over 100 large language models and vision-language models, featuring zero-code CLI and Web UI tools that simplify the process. It matters now because fine-tuning is essential for customizing AI models in real-world applications, and this repo's popularity, with over 70k stars and adoption by companies like Amazon and NVIDIA, indicates its practical utility, though its effectiveness depends on specific use cases and hardware. _themes: fine-tuning · llm · quantization · peft_ #### [facebookresearch/faiss](https://github.com/facebookresearch/faiss) *C++ · ★39,801 · MIT · production · score:0.90 · hot:0.83 · rising:0.89 · durable:0.80 · board:rising · trend:up* A library for efficient similarity search and clustering of dense vectors. **Why it matters.** Faiss is a C++ library with Python bindings for efficient similarity search and clustering of dense vectors, supporting operations on large datasets that may not fit in RAM and including GPU-accelerated algorithms. It matters now because vector search is essential for modern AI applications like recommendation systems and large-scale embeddings, where performance and scalability are critical amid growing data volumes. However, while it's widely used, users should be aware of its complexity and the need for careful tuning to achieve optimal results. _themes: vector-search · clustering · gpu-acceleration · scalability_ #### [pytorch/ao](https://github.com/pytorch/ao) *Python · ★2,792 · NOASSERTION · beta · score:0.80 · hot:0.83 · rising:0.80 · durable:0.69 · board:hot · trend:up* PyTorch native quantization and sparsity for training and inference **Why it matters.** TorchAO is a PyTorch library that enables native quantization and sparsity techniques for optimizing model training and inference, such as using float8 for faster training and int4 for reduced memory inference on large models like Llama. It matters now because the AI community is grappling with the computational demands of scaling large language models, and recent updates demonstrate practical speedups and integrations that address hardware efficiency challenges amid advancing GPU technologies. _themes: quantization · sparsity · inference · training_ #### [huggingface/pytorch-image-models](https://github.com/huggingface/pytorch-image-models) *Python · ★36,667 · Apache-2.0 · production · score:0.90 · hot:0.83 · rising:0.87 · durable:0.84 · board:rising · trend:up* The largest collection of PyTorch image encoders / backbones. Including train, eval, inference, export scripts, and pretrained weights -- ResNet, ResNeXT, EfficientNet, NFNet, Vision Transformer (ViT), MobileNetV4, MobileNet-V3 & V2, RegNet, DPN, CSPNet, Swin Transformer, MaxViT, CoAtNet, ConvNeXt, and more **Why it matters.** This repository provides a large collection of PyTorch-based image models, including architectures like ResNet, EfficientNet, and Vision Transformers, along with scripts for training, evaluation, inference, and pretrained weights, making it a go-to resource for computer vision tasks. It matters now because it supports the ongoing demand for efficient, pre-built models in AI development, especially as transformer-based and mobile-optimized models gain prominence in research and applications, though its maintenance appears steady but not cutting-edge with the latest release from early 2026. _themes: pytorch · image-classification · pretrained-models · computer-vision_ #### [huggingface/trl](https://github.com/huggingface/trl) *Python · ★18,141 · Apache-2.0 · production · score:0.90 · hot:0.83 · rising:0.87 · durable:0.75 · board:rising · trend:up* Train transformer language models with reinforcement learning. **Why it matters.** TRL is a library for fine-tuning transformer language models using reinforcement learning techniques like Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO), enabling better alignment and performance on specific tasks. It matters right now because the AI field is emphasizing safer, more controllable models amid growing concerns about ethics and deployment, and TRL's integration with Hugging Face tools makes it accessible for scaling experiments on limited hardware. _themes: fine-tuning · reinforcement-learning · transformers · scaling_ #### [ggml-org/whisper.cpp](https://github.com/ggml-org/whisper.cpp) *C++ · ★48,767 · MIT · production · score:0.90 · hot:0.83 · rising:0.85 · durable:0.81 · board:rising · trend:up* Port of OpenAI's Whisper model in C/C++ **Why it matters.** Whisper.cpp is a C/C++ port of OpenAI's Whisper model for efficient speech-to-text inference, optimized for various hardware like ARM, x86, and GPUs without external dependencies. It matters now because it enables lightweight, high-performance ASR deployment on edge devices and diverse platforms, addressing real-world needs in mobile, IoT, and embedded applications amid growing demand for on-device AI. However, its focus on inference alone means it lacks training capabilities, potentially limiting its appeal for full-cycle development. _themes: inference · speech-recognition · optimization · cross-platform_ #### [microsoft/garnet](https://github.com/microsoft/garnet) *C# · ★11,797 · MIT · production · score:0.80 · hot:0.82 · rising:0.85 · durable:0.80 · board:rising · trend:up* Garnet is a remote cache-store from Microsoft Research that offers strong performance (throughput and latency), scalability, storage, recovery, cluster sharding, key migration, and replication features. Garnet can work with existing Redis clients. **Why it matters.** Garnet is a remote cache-store developed by Microsoft that uses the RESP protocol for compatibility with existing Redis clients, offering high throughput, low latency, and features like scalability and persistence. It matters now because as cloud-based applications demand efficient caching to handle growing data loads and reduce costs, Garnet provides a performant alternative that could integrate seamlessly into current setups, potentially improving performance in high-traffic scenarios. _themes: caching · scalability · low-latency · key-value_ #### [huggingface/diffusers](https://github.com/huggingface/diffusers) *Python · ★33,383 · Apache-2.0 · production · score:0.90 · hot:0.81 · rising:0.86 · durable:0.78 · board:rising · trend:up* 🤗 Diffusers: State-of-the-art diffusion models for image, video, and audio generation in PyTorch. **Why it matters.** Hugging Face Diffusers is a Python library that provides modular tools for working with diffusion models in PyTorch, enabling easy inference and training for generating images, videos, and audio. It stands out for its focus on usability and customizability, making it accessible for building generative AI applications. This matters now because diffusion models are central to the ongoing advancements in AI generation, with increasing adoption in creative and research fields, though it prioritizes simplicity over raw performance optimizations. _themes: diffusion · image-generation · pytorch · generative-models_ #### [gradio-app/gradio](https://github.com/gradio-app/gradio) *Python · ★42,371 · Apache-2.0 · production · score:0.85 · hot:0.81 · rising:0.85 · durable:0.78 · board:rising · trend:up* Build and share delightful machine learning apps, all in Python. 🌟 Star to support our work! **Why it matters.** Gradio is a Python library that simplifies building and sharing interactive web interfaces for machine learning models or any Python functions with minimal code, requiring no frontend expertise. It matters now because the rapid growth of AI development demands quick, accessible ways to demo and deploy models, fostering collaboration and real-world application in an era of widespread AI adoption. This tool lowers barriers for developers, enabling faster iteration and sharing in projects involving generative AI and beyond. _themes: machine-learning · ui · deployment · data-visualization_ #### [huggingface/lerobot](https://github.com/huggingface/lerobot) *Python · ★23,350 · Apache-2.0 · beta · score:0.75 · hot:0.80 · rising:0.84 · durable:0.73 · board:rising · trend:up* 🤗 LeRobot: Making AI for Robotics more accessible with end-to-end learning **Why it matters.** LeRobot is a library from Hugging Face that provides tools, datasets, and models for end-to-end robotics AI development in PyTorch, focusing on hardware-agnostic interfaces and standardized data formats to simplify real-world applications. It matters right now because it lowers barriers for developers and researchers in robotics, potentially accelerating innovation amid growing interest in physical AI, though its real-world transfer success depends on further validation and broader hardware support. _themes: robotics · datasets · pytorch · end-to-end_ #### [google-deepmind/mujoco](https://github.com/google-deepmind/mujoco) *C++ · ★12,933 · Apache-2.0 · production · score:0.90 · hot:0.80 · rising:0.84 · durable:0.76 · board:rising · trend:up* Multi-Joint dynamics with Contact. A general purpose physics simulator. **Why it matters.** MuJoCo is a general-purpose physics engine for simulating multi-joint dynamics and contacts, primarily used in robotics, biomechanics, and machine learning for accurate environment modeling. It matters right now because physics-based simulations are critical for advancing AI research in areas like reinforcement learning and embodied agents, especially with the growing demand for realistic training environments in robotics. Recent updates, such as version 3.7.0, continue to improve performance and integration, making it a reliable tool amid rapid developments in AI hardware and simulation needs. _themes: physics · robotics · simulation · ml_ #### [streamlit/streamlit](https://github.com/streamlit/streamlit) *Python · ★44,275 · Apache-2.0 · production · score:0.90 · hot:0.80 · rising:0.84 · durable:0.75 · board:rising · trend:up* Streamlit — A faster way to build and share data apps. **Why it matters.** Streamlit is a Python library that simplifies building interactive web applications for data analysis, visualization, and machine learning by converting scripts into shareable apps with minimal code. It matters now because it enables rapid prototyping and deployment in the AI and data science boom, allowing developers to iterate quickly without deep web expertise, amid growing demands for accessible tools in collaborative environments. _themes: data-viz · ml-prototyping · python-apps · interactive-tools_ #### [huggingface/datasets](https://github.com/huggingface/datasets) *Python · ★21,420 · Apache-2.0 · production · score:0.90 · hot:0.80 · rising:0.83 · durable:0.77 · board:rising · trend:up* 🤗 The largest hub of ready-to-use datasets for AI models with fast, easy-to-use and efficient data manipulation tools **Why it matters.** Hugging Face Datasets is a Python library that provides easy access to a large collection of public datasets and tools for efficient data loading and preprocessing, supporting formats like CSV, JSON, and integration with frameworks such as PyTorch and TensorFlow. It matters now because the growing demand for diverse, ready-to-use datasets in AI research and development helps accelerate model training, especially amid the proliferation of large language models and multimodal applications. However, its reliance on the Hugging Face ecosystem might limit flexibility for users outside that framework. _themes: datasets · data-processing · machine-learning · huggingface_ #### [OpenBB-finance/OpenBB](https://github.com/OpenBB-finance/OpenBB) *Python · ★66,099 · NOASSERTION · production · score:0.80 · hot:0.80 · rising:0.85 · durable:0.79 · board:rising · trend:up* Financial data platform for analysts, quants and AI agents. **Why it matters.** OpenBB is an open-source platform that integrates financial data from various sources, allowing users to access and manipulate data in Python, AI agents, and other tools with a 'connect once, consume everywhere' approach. It matters right now because the financial industry is rapidly adopting AI for analysis and decision-making, and OpenBB provides a practical foundation for quants and developers to handle complex data integrations amid rising demands in crypto, equities, and quantitative finance. However, its unspecified license could pose risks for enterprise adoption. _themes: finance · ai · data-integration · machine-learning_ #### [microsoft/onnxruntime](https://github.com/microsoft/onnxruntime) *C++ · ★19,895 · MIT · production · score:0.90 · hot:0.79 · rising:0.84 · durable:0.75 · board:rising · trend:up* ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator **Why it matters.** ONNX Runtime is an open-source library that accelerates machine learning inference and training for ONNX models, supporting frameworks like PyTorch, TensorFlow, and scikit-learn while optimizing for various hardware. It matters now because efficient model deployment is critical for scaling AI in production, especially with increasing demands for cost-effective inference on diverse devices, though its performance gains depend on specific hardware and model compatibility. Adoption is driven by the need for interoperability in a fragmented ML ecosystem, but users should verify optimizations against alternatives. _themes: inference · hardware-acceleration · onnx · training_ #### [0xPlaygrounds/rig](https://github.com/0xPlaygrounds/rig) *Rust · ★6,965 · MIT · beta · score:0.60 · hot:0.79 · rising:0.79 · durable:0.71 · board:hot · trend:up* ⚙️🦀 Build modular and scalable LLM Applications in Rust **Why it matters.** Rig is a Rust library designed for building modular and scalable applications powered by large language models, offering features like agentic workflows, integrations with over 20 model providers, and OpenTelemetry compatibility. It matters now amid the growing demand for efficient AI tools, as Rust's performance and safety could appeal to developers seeking production-grade LLM apps, but its active development with frequent breaking changes makes it risky for immediate adoption. _themes: agents · llm · rust · inference_ #### [microsoft/markitdown](https://github.com/microsoft/markitdown) *Python · ★112,498 · MIT · beta · score:0.70 · hot:0.79 · rising:0.84 · durable:0.82 · board:rising · trend:up* Python tool for converting files and office documents to Markdown. **Why it matters.** MarkItDown is a Python library that converts files and office documents to Markdown, focusing on preserving structure for LLM and text analysis pipelines, making it useful for integrating documents into AI workflows. It matters now because the growing adoption of LLMs requires reliable tools for document preprocessing, especially with integrations like LangChain and AutoGen, but its early version includes breaking changes that could disrupt existing implementations. _themes: markdown · document-conversion · llm · ai-tools_ #### [huggingface/transformers.js](https://github.com/huggingface/transformers.js) *JavaScript · ★15,876 · Apache-2.0 · production · score:0.80 · hot:0.79 · rising:0.82 · durable:0.78 · board:rising · trend:up* State-of-the-art Machine Learning for the web. Run 🤗 Transformers directly in your browser, with no need for a server! **Why it matters.** Hugging Face's transformers.js library allows running pre-trained machine learning models directly in web browsers using JavaScript, supporting tasks in NLP, computer vision, audio, and multimodal areas without requiring a server, which leverages ONNX Runtime for execution. This matters now because it enables edge computing for AI applications, reducing latency and enhancing privacy in web apps, amid growing demands for client-side processing in an era of increasing web-based AI integration. _themes: inference · browser · transformers · onnx_ #### [NVIDIA/Megatron-LM](https://github.com/NVIDIA/Megatron-LM) *Python · ★16,097 · NOASSERTION · beta · score:0.85 · hot:0.79 · rising:0.80 · durable:0.70 · board:rising · trend:up* Ongoing research training transformer models at scale **Why it matters.** Megatron-LM is a NVIDIA repository providing tools for training large transformer models at scale, including distributed parallelism strategies and GPU-optimized building blocks for custom frameworks. It matters right now because the rapid advancement of AI requires efficient handling of massive models, and it offers practical solutions for researchers and engineers dealing with distributed training challenges amid growing computational demands. _themes: transformers · distributed-training · llms · parallelism_ #### [huggingface/huggingface_hub](https://github.com/huggingface/huggingface_hub) *Python · ★3,530 · Apache-2.0 · production · score:0.90 · hot:0.79 · rising:0.81 · durable:0.72 · board:rising · trend:up* The official Python client for the Hugging Face Hub. **Why it matters.** The huggingface_hub library is a Python client that enables users to interact with the Hugging Face Hub, allowing them to download, upload, and manage machine learning models and datasets seamlessly. It matters now because the proliferation of open-source AI models demands easy access and collaboration tools, and this library integrates with a leading platform that hosts thousands of pre-trained models, facilitating rapid development in AI projects amid growing adoption of generative AI. _themes: machine-learning · pretrained-models · inference · model-hub_ #### [NVIDIA/nccl](https://github.com/NVIDIA/nccl) *C++ · ★4,630 · NOASSERTION · production · score:0.85 · hot:0.78 · rising:0.80 · durable:0.68 · board:rising · trend:up* Optimized primitives for collective multi-GPU communication **Why it matters.** NCCL offers optimized C++ primitives for collective multi-GPU communication operations like all-reduce and broadcast, which are essential for efficient distributed training in deep learning workflows. It matters now because the growing scale of AI models demands high-bandwidth GPU communication to avoid bottlenecks in multi-node setups, making it a foundational tool for scalable training on NVIDIA hardware. However, its NVIDIA-specific nature limits broader applicability compared to more platform-agnostic alternatives. _themes: distributed-computing · gpu · deep-learning · cuda_ #### [google-deepmind/dm_control](https://github.com/google-deepmind/dm_control) *Python · ★4,541 · Apache-2.0 · production · score:0.85 · hot:0.78 · rising:0.82 · durable:0.71 · board:rising · trend:up* Google DeepMind's software stack for physics-based simulation and Reinforcement Learning environments, using MuJoCo. **Why it matters.** This repository provides Google DeepMind's software stack for physics-based simulation and reinforcement learning environments using the MuJoCo engine, including Python bindings, RL task suites, and tools for environment composition. It matters now because it supports advanced research in robotics and AI control systems, where high-fidelity simulations are essential for training agents in complex physical interactions, especially amid growing interest in embodied AI and real-world applications. _themes: reinforcement-learning · physics-simulation · deep-learning · mujoco_ #### [huggingface/sentence-transformers](https://github.com/huggingface/sentence-transformers) *Python · ★18,564 · Apache-2.0 · production · score:0.90 · hot:0.78 · rising:0.82 · durable:0.74 · board:rising · trend:up* State-of-the-Art Text Embeddings **Why it matters.** This repository provides a Python library for computing state-of-the-art text embeddings using pre-trained Sentence Transformer models, enabling tasks like semantic search, similarity scoring, and reranking. It matters right now because high-quality embeddings are essential for modern AI applications such as retrieval-augmented generation and large-scale information retrieval, especially with the growing emphasis on efficient NLP in enterprise and research settings amid advancing benchmarks like MTEB. _themes: embeddings · inference · fine-tuning · semantic-search_ #### [NVIDIA/kvpress](https://github.com/NVIDIA/kvpress) *Python · ★1,043 · Apache-2.0 · beta · score:0.80 · hot:0.78 · rising:0.76 · durable:0.74 · board:hot · trend:up* LLM KV cache compression made easy **Why it matters.** KVPress is a library from NVIDIA that provides methods for compressing the key-value cache in transformer-based LLMs, addressing the memory overhead of long-context processing, which is a common bottleneck in deploying large models. This matters now because as LLMs handle increasingly longer sequences for applications like extended conversations or RAG, efficient memory usage becomes critical for scalability on limited hardware, potentially enabling broader adoption without requiring massive infrastructure. However, its impact depends on real-world performance gains, as not all compression methods may preserve accuracy effectively. _themes: inference · kv-cache · llm · compression_ #### [NVIDIA/NeMo-Agent-Toolkit](https://github.com/NVIDIA/NeMo-Agent-Toolkit) *Python · ★2,204 · Apache-2.0 · beta · score:0.75 · hot:0.78 · rising:0.80 · durable:0.69 · board:rising · trend:up* The NVIDIA NeMo Agent toolkit is an open-source library for efficiently connecting and optimizing teams of AI agents. **Why it matters.** The NVIDIA NeMo Agent Toolkit is an open-source library that enables efficient connection, optimization, and observability for teams of AI agents, supporting features like performance primitives, runtime intelligence, and integrations with frameworks such as LangChain. It matters now because the growing adoption of AI agents in enterprise applications demands better speed, accuracy, and scalability, and this toolkit provides practical tools for instrumentation and deployment amid increasing competition in agent orchestration. _themes: agents · optimization · observability · inference_ #### [NVIDIA/TransformerEngine](https://github.com/NVIDIA/TransformerEngine) *Python · ★3,280 · Apache-2.0 · production · score:0.85 · hot:0.78 · rising:0.79 · durable:0.73 · board:rising · trend:up* A library for accelerating Transformer models on NVIDIA GPUs, including using 8-bit and 4-bit floating point (FP8 and FP4) precision on Hopper, Ada and Blackwell GPUs, to provide better performance with lower memory utilization in both training and inference. **Why it matters.** NVIDIA TransformerEngine is a library that optimizes Transformer models for NVIDIA GPUs by supporting lower precision formats like FP8 and FP4, which reduces memory usage and improves speed for training and inference. It matters now because the demand for efficient AI computations is growing amid resource constraints, and with integrations for PyTorch and JAX, it directly addresses performance bottlenecks in large-scale model development on compatible hardware like Hopper and Blackwell GPUs. _themes: deep-learning · gpu · mixed-precision · acceleration_ #### [tesseract-ocr/tesseract](https://github.com/tesseract-ocr/tesseract) *C++ · ★73,608 · Apache-2.0 · production · score:0.90 · hot:0.77 · rising:0.85 · durable:0.81 · board:rising · trend:up* Tesseract Open Source OCR Engine (main repository) **Why it matters.** Tesseract is an open-source OCR engine that converts images of text into editable and searchable data using both legacy pattern recognition and a modern LSTM-based neural network, supporting over 100 languages. It matters now because accurate OCR is crucial for automating document processing in industries like finance and healthcare, especially with the growing demand for AI-driven data extraction tools. However, its performance can vary by language and image quality, requiring users to fine-tune or integrate it carefully. _themes: ocr · machine-learning · lstm · image-processing_ #### [huggingface/accelerate](https://github.com/huggingface/accelerate) *Python · ★9,613 · Apache-2.0 · production · score:0.85 · hot:0.77 · rising:0.82 · durable:0.78 · board:rising · trend:up* 🚀 A simple way to launch, train, and use PyTorch models on almost any device and distributed configuration, automatic mixed precision (including fp8), and easy-to-configure FSDP and DeepSpeed support **Why it matters.** Hugging Face Accelerate is a library that simplifies running PyTorch training scripts on various devices, including multi-GPUs, TPUs, and with mixed precision, by handling boilerplate code with minimal changes. It matters now because the increasing scale of AI models demands efficient distributed training, making it easier for developers to leverage advanced hardware without extensive rework. This tool from Hugging Face aligns with the growing ecosystem of PyTorch-based workflows. _themes: pytorch · distributed · mixed-precision · training_ #### [samchon/nestia](https://github.com/samchon/nestia) *TypeScript · ★2,146 · MIT · production · score:0.75 · hot:0.77 · rising:0.73 · durable:0.69 · board:hot · trend:up* NestJS Helper + AI Chatbot Development **Why it matters.** Nestia is a collection of helper libraries for NestJS that provide performance-optimized decorators, SDK generation, and tools for API documentation and testing, while also integrating AI features like LLM function calling; it claims significant speed improvements over alternatives, making it relevant for developers seeking efficient backend development in TypeScript ecosystems. This matters now as the demand for fast, scalable APIs grows with AI applications, potentially reducing development time and improving performance in production environments. _themes: agents · api · typescript · validation_ #### [rasbt/LLMs-from-scratch](https://github.com/rasbt/LLMs-from-scratch) *Jupyter Notebook · ★91,080 · NOASSERTION · experimental · score:0.90 · hot:0.77 · rising:0.78 · durable:0.75 · board:rising · trend:up* Implement a ChatGPT-like LLM in PyTorch from scratch, step by step **Why it matters.** This repository provides a step-by-step implementation of a ChatGPT-like large language model using PyTorch, including code for building, pretraining, and finetuning from scratch, based on an accompanying book for educational purposes. It matters right now because the rapid advancement of AI and LLMs has created a need for accessible resources that demystify complex models, enabling beginners and researchers to understand and experiment with foundational technologies amid growing interest in generative AI. _themes: from-scratch · llm · pytorch · fine-tuning_ #### [NVIDIA/Model-Optimizer](https://github.com/NVIDIA/Model-Optimizer) *Python · ★2,514 · Apache-2.0 · beta · score:0.75 · hot:0.77 · rising:0.78 · durable:0.67 · board:rising · trend:up* A unified library of SOTA model optimization techniques like quantization, pruning, distillation, speculative decoding, etc. It compresses deep learning models for downstream deployment frameworks like TensorRT-LLM, TensorRT, vLLM, etc. to optimize inference speed. **Why it matters.** NVIDIA Model Optimizer is a library that applies state-of-the-art techniques like quantization, pruning, and distillation to compress deep learning models, making them faster for inference on frameworks such as TensorRT and vLLM. It matters now because the growing demand for efficient AI deployment on edge devices and in production environments requires optimized models to reduce latency and computational costs, though its NVIDIA-centric focus may limit broader applicability. However, as a relatively niche tool, it primarily benefits users already in the NVIDIA ecosystem rather than offering universal solutions. _themes: inference · quantization · pruning · optimization_ #### [NVIDIA/cutlass](https://github.com/NVIDIA/cutlass) *C++ · ★9,602 · NOASSERTION · production · score:0.90 · hot:0.76 · rising:0.79 · durable:0.70 · board:rising · trend:up* CUDA Templates and Python DSLs for High-Performance Linear Algebra **Why it matters.** CUTLASS provides C++ templates and Python DSLs for optimizing high-performance linear algebra operations like matrix multiplication on NVIDIA GPUs using CUDA, supporting a wide range of data types and architectures. It matters right now because efficient GPU computations are critical for accelerating deep learning workloads amid growing demands for faster AI training and inference, especially as NVIDIA hardware evolves with new features. However, its focus on low-level optimizations may require expertise, limiting immediate adoption for non-specialists. _themes: cuda · gpu · linear-algebra · deep-learning_ #### [NVIDIA/cccl](https://github.com/NVIDIA/cccl) *C++ · ★2,285 · NOASSERTION · production · score:0.75 · hot:0.76 · rising:0.76 · durable:0.66 · board:hot · trend:up* CUDA Core Compute Libraries **Why it matters.** NVIDIA/cccl unifies essential CUDA C++ libraries like Thrust, CUB, and libcudacxx into a single repository, providing developers with high-quality abstractions for efficient GPU programming and safer code. This consolidation streamlines development for CUDA-based projects, which is increasingly relevant amid the surge in AI and high-performance computing demands, as it reduces fragmentation and enhances productivity for GPU-accelerated applications. _themes: cuda · gpu-computing · c++ · hpc_ #### [ultralytics/yolov5](https://github.com/ultralytics/yolov5) *Python · ★57,243 · AGPL-3.0 · production · score:0.90 · hot:0.76 · rising:0.83 · durable:0.84 · board:durable · trend:up* YOLOv5 🚀 in PyTorch > ONNX > CoreML > TFLite **Why it matters.** Ultralytics/yolov5 is an open-source implementation of the YOLOv5 object detection model using PyTorch, supporting tasks like object detection, segmentation, and classification with exports to formats such as ONNX and TFLite for deployment across devices. It matters right now because object detection is essential in applications like autonomous vehicles and mobile vision, and this repo provides a user-friendly, high-performance solution that's actively maintained and widely adopted by developers and researchers. Its extensive community support and regular updates make it a practical choice for integrating state-of-the-art vision AI into real-world projects. _themes: object-detection · pytorch · inference · fine-tuning_ #### [mistralai/mistral-inference](https://github.com/mistralai/mistral-inference) *Jupyter Notebook · ★10,775 · Apache-2.0 · production · score:0.85 · hot:0.76 · rising:0.80 · durable:0.79 · board:rising · trend:up* Official inference library for Mistral models **Why it matters.** Mistral-inference is an official library that provides minimal code for running inference on Mistral's large language models, optimized for GPU hardware. It matters now because Mistral models are increasingly adopted for their efficiency and open-source accessibility in AI applications, especially as developers seek cost-effective alternatives to proprietary solutions amid growing demand for on-device or edge AI inference. _themes: llm · inference · gpu · models_ #### [NVIDIA/cuopt](https://github.com/NVIDIA/cuopt) *Cuda · ★824 · Apache-2.0 · production · score:0.70 · hot:0.76 · rising:0.77 · durable:0.68 · board:rising · trend:up* GPU accelerated decision optimization **Why it matters.** NVIDIA cuOpt is a GPU-accelerated library for solving optimization problems like linear programming, mixed integer programming, quadratic programming, and vehicle routing, enabling faster processing of large-scale datasets that traditional CPUs struggle with. It matters right now because industries such as logistics, supply chain, and AI-driven planning require real-time optimization, and GPU acceleration addresses the growing need for high-performance computing in an era of big data and edge applications. This makes it particularly relevant for applications where speed and scalability are critical. _themes: optimization · gpu · linear-programming · routing_ #### [NVIDIA/tilus](https://github.com/NVIDIA/tilus) *Python · ★475 · Apache-2.0 · beta · score:0.70 · hot:0.75 · rising:0.74 · durable:0.68 · board:hot · trend:up* Tilus is a tile-level kernel programming language with explicit control over shared memory and registers. **Why it matters.** Tilus is a domain-specific language for GPU kernel programming that provides thread-block-level control over shared memory, registers, and tensors, enabling explicit optimization for NVIDIA GPUs. It matters now because it supports the latest hardware like Blackwell and Hopper, allowing developers to build high-performance kernels that rival cuBLAS, which is crucial amid the growing demand for efficient AI and machine learning computations on GPUs. _themes: cuda · kernels · optimization · tensors_ #### [allenai/open-instruct](https://github.com/allenai/open-instruct) *Python · ★3,696 · Apache-2.0 · beta · score:0.70 · hot:0.75 · rising:0.78 · durable:0.72 · board:rising · trend:up* AllenAI's post-training codebase **Why it matters.** This repository from AllenAI provides code and resources for instruction-tuning and post-training pre-trained language models using publicly available datasets, including techniques like DPO and RLVR, which help in aligning models with human preferences. It matters now because fine-tuning LLMs for better instruction-following is essential amid rapid advancements in AI safety and efficiency, though the codebase has unmaintained evaluation components and focuses on research-oriented outputs rather than polished production tools. _themes: fine-tuning · instruction-tuning · RL · language-models_ #### [huggingface/text-embeddings-inference](https://github.com/huggingface/text-embeddings-inference) *Rust · ★4,706 · Apache-2.0 · production · score:0.80 · hot:0.75 · rising:0.77 · durable:0.72 · board:rising · trend:up* A blazing fast inference solution for text embeddings models **Why it matters.** Text Embeddings Inference (TEI) is a Rust-based toolkit for deploying and serving text embeddings and sequence classification models from Hugging Face, focusing on high-performance inference with features like dynamic batching and hardware optimization. It addresses the growing need for efficient embedding generation in AI applications, such as semantic search and retrieval, by providing fast boot times and small Docker images for scalable deployments. However, its reliance on specific hardware like NVIDIA GPUs may limit broader accessibility compared to more versatile alternatives. _themes: inference · embeddings · optimization · rust_ #### [NVIDIA/torch-harmonics](https://github.com/NVIDIA/torch-harmonics) *Jupyter Notebook · ★660 · NOASSERTION · beta · score:0.70 · hot:0.75 · rising:0.73 · durable:0.63 · board:hot · trend:up* Differentiable signal processing on the sphere for PyTorch **Why it matters.** Torch-harmonics is a PyTorch library that provides differentiable operations for signal processing on spherical data, including spherical harmonic transforms and convolutions, which are essential for applications like PDE solvers and neural operators in scientific computing. It matters now because the growing integration of AI in fields such as climate modeling and astrophysics demands efficient, differentiable tools for handling spherical geometries, potentially accelerating research and development in these areas. _themes: pytorch · signal-processing · spherical-harmonics · differentiable-computing_ #### [google-deepmind/mujoco_warp](https://github.com/google-deepmind/mujoco_warp) *Python · ★1,186 · Apache-2.0 · beta · score:0.75 · hot:0.75 · rising:0.76 · durable:0.66 · board:rising · trend:up* GPU-optimized version of the MuJoCo physics simulator, designed for NVIDIA hardware. **Why it matters.** MuJoCo Warp is a GPU-optimized version of the MuJoCo physics simulator, specifically designed for NVIDIA hardware to accelerate simulations in robotics and AI research. It matters right now because the growing demand for fast, scalable physics engines in machine learning workflows, particularly for training robotic policies, highlights the need for hardware-specific optimizations amid advancing GPU technologies. However, its niche focus on NVIDIA platforms limits broader applicability compared to more general simulators. _themes: simulation · gpu-acceleration · physics · robotics_ #### [scikit-learn/scikit-learn](https://github.com/scikit-learn/scikit-learn) *Python · ★65,868 · BSD-3-Clause · production · score:0.95 · hot:0.74 · rising:0.82 · durable:0.77 · board:rising · trend:up* scikit-learn: machine learning in Python **Why it matters.** Scikit-learn is a mature Python library that offers a wide range of machine learning algorithms for tasks like classification, regression, and clustering, making it essential for data analysis and model prototyping. It matters right now as it remains a go-to tool for traditional ML workflows in production environments, providing stability and ease of use amid the evolving AI landscape dominated by deep learning frameworks. Its extensive ecosystem supports rapid development and integration into larger projects. _themes: machine-learning · data-analysis · statistics · python_ #### [NVIDIA/DALI](https://github.com/NVIDIA/DALI) *C++ · ★5,672 · Apache-2.0 · production · score:0.85 · hot:0.74 · rising:0.77 · durable:0.71 · board:rising · trend:up* A GPU-accelerated library containing highly optimized building blocks and an execution engine for data processing to accelerate deep learning training and inference applications. **Why it matters.** NVIDIA DALI is a GPU-accelerated library that optimizes data loading and preprocessing for deep learning, handling tasks like image and audio augmentation to reduce CPU bottlenecks and speed up training and inference. It matters now because as AI models and datasets scale, efficient data pipelines are essential to avoid performance limitations, and DALI provides a framework-agnostic solution that integrates with TensorFlow, PyTorch, and others. However, its benefits are most pronounced in GPU-heavy workflows, so adoption depends on specific hardware and use cases. _themes: gpu-acceleration · data-processing · deep-learning · image-augmentation_ #### [huggingface/smolagents](https://github.com/huggingface/smolagents) *Python · ★26,732 · Apache-2.0 · production · score:0.85 · hot:0.74 · rising:0.81 · durable:0.80 · board:rising · trend:up* 🤗 smolagents: a barebones library for agents that think in code. **Why it matters.** Smolagents is a minimal library for building AI agents that generate and execute code, emphasizing simplicity and security through sandboxed environments. It integrates with various LLMs, tools, and modalities, making it easier to create agents for tasks like automation without heavy abstractions. This matters now as AI agents are gaining traction for real-world applications, but concerns around code safety and ease of use make a lightweight, secure option like this relevant for developers exploring agent-based systems. _themes: agents · code-execution · llm-integration · tools_ #### [google-deepmind/concordia](https://github.com/google-deepmind/concordia) *Python · ★1,359 · Apache-2.0 · beta · score:0.70 · hot:0.74 · rising:0.75 · durable:0.67 · board:rising · trend:up* A library for generative social simulation **Why it matters.** Concordia is a Python library from Google DeepMind that facilitates generative agent-based simulations, where entities interact in simulated environments via a Game Master inspired by tabletop RPGs, requiring an LLM for natural language processing. It matters now because it addresses growing needs in AI safety, social science research, and ethics by providing tools to model complex social interactions, though its reliance on external LLMs may introduce variability and dependency issues. This makes it relevant amid increasing interest in multi-agent systems for real-world applications like synthetic data generation. _themes: agents · simulation · multi-agent · llm_ #### [facebookresearch/spdl](https://github.com/facebookresearch/spdl) *Python · ★381 · BSD-2-Clause · beta · score:0.65 · hot:0.74 · rising:0.73 · durable:0.64 · board:hot · trend:up* Scalable and Performant Data Loading **Why it matters.** SPDL is a library from Facebook Research that focuses on creating scalable and performant data loading pipelines for processing array data in machine learning and deep learning workflows. It addresses common bottlenecks in data handling, which is increasingly critical as models and datasets grow larger, potentially improving training efficiency. However, its early release version and research-oriented nature mean its practical benefits and stability require further real-world testing. _themes: data-loading · performance · scalability · ml_ #### [BlinkDL/RWKV-LM](https://github.com/BlinkDL/RWKV-LM) *Python · ★14,479 · Apache-2.0 · beta · score:0.75 · hot:0.74 · rising:0.76 · durable:0.75 · board:rising · trend:up* RWKV (pronounced RwaKuv) is an RNN with great LLM performance, which can also be directly trained like a GPT transformer (parallelizable). We are at RWKV-7 "Goose". So it's combining the best of RNN and transformer - great performance, linear time, constant space (no kv-cache), fast training, infinite ctx_len, and free sentence embedding. **Why it matters.** RWKV-LM implements an RNN architecture that achieves transformer-like performance for language models while offering advantages in training efficiency, such as parallelization, linear time complexity, and constant memory usage without KV-caches. It matters now because the growing demand for scalable AI models highlights the need for alternatives that reduce computational costs and enable infinite context lengths, potentially addressing limitations in current transformer-based systems; however, its claims require rigorous independent verification against established benchmarks. _themes: rnn · language-model · efficient-inference · linear-attention_ #### [NVIDIA/TileGym](https://github.com/NVIDIA/TileGym) *Python · ★705 · NOASSERTION · beta · score:0.70 · hot:0.74 · rising:0.73 · durable:0.66 · board:hot · trend:up* Helpful kernel tutorials and examples for tile-based GPU programming **Why it matters.** TileGym provides tutorials and examples for CUDA Tile-based GPU programming, focusing on building efficient kernels for large language models like Llama 3.1 and DeepSeek V2. It matters now because optimizing GPU performance is critical amid the growing demands of AI workloads, and NVIDIA's tools help developers leverage specific hardware for better efficiency in real-world applications. _themes: cuda · gpu-optimization · llm · kernels_ #### [facebookresearch/fairchem](https://github.com/facebookresearch/fairchem) *Python · ★2,065 · NOASSERTION · beta · score:0.70 · hot:0.74 · rising:0.74 · durable:0.63 · board:rising · trend:up* FAIR Chemistry's library of machine learning methods for chemistry **Why it matters.** fairchem is a library from FAIR that provides machine learning methods, data, models, and demos for chemistry applications, particularly in materials science and quantum chemistry, serving as a centralized resource for researchers. It matters now because the field of ML in chemistry is advancing rapidly for applications like drug discovery and material design, but its breaking changes from version 1 to 2 and compatibility warnings could hinder adoption, requiring users to carefully manage updates and data inconsistencies. _themes: ml-for-chemistry · materials-science · quantum-chemistry · deep-learning_ #### [allenai/OLMo-core](https://github.com/allenai/OLMo-core) *Python · ★1,160 · Apache-2.0 · beta · score:0.70 · hot:0.73 · rising:0.74 · durable:0.65 · board:rising · trend:up* PyTorch building blocks for the OLMo ecosystem **Why it matters.** OLMo-core provides PyTorch-based building blocks for the OLMo ecosystem, focusing on optimized components for language model training and inference, including support for advanced attention mechanisms and hardware-specific optimizations. It matters right now because the AI community is pushing for more efficient large language models amid hardware constraints, but its reliance on multiple external dependencies could complicate adoption for non-experts. However, it supports ongoing research in scalable AI by offering pre-configured tools tested on high-end hardware. _themes: pytorch · llm · optimization · training_ #### [NVIDIA/NeMo-Retriever](https://github.com/NVIDIA/NeMo-Retriever) *Python · ★2,905 · Apache-2.0 · production · score:0.80 · hot:0.73 · rising:0.77 · durable:0.71 · board:rising · trend:up* NeMo Retriever Library is a scalable, performance-oriented document content and metadata extraction microservice. NeMo Retriever extraction uses specialized NVIDIA NIM microservices to find, contextualize, and extract text, tables, charts and images that you can use in downstream generative applications. **Why it matters.** NeMo Retriever is a library for extracting and contextualizing text, tables, charts, and images from documents, then computing embeddings and storing them in vector databases like Milvus for use in generative AI applications. It matters right now because the growing demand for efficient, scalable document processing is essential for building retrieval-augmented generation (RAG) systems, especially with the rise of multimodal AI workflows that require robust data ingestion. _themes: rag · document-extraction · embeddings · multimodal_ #### [huggingface/tokenizers](https://github.com/huggingface/tokenizers) *Rust · ★10,647 · Apache-2.0 · production · score:0.90 · hot:0.73 · rising:0.79 · durable:0.77 · board:rising · trend:up* 💥 Fast State-of-the-Art Tokenizers optimized for Research and Production **Why it matters.** Hugging Face's tokenizers library provides high-performance implementations of state-of-the-art tokenizers in Rust, enabling fast vocabulary training and text processing for NLP tasks like those in BERT and GPT models. It matters now because the growing demand for efficient NLP pipelines in research and production requires tools that handle large-scale data quickly, reducing computation time and costs. This library's versatility and speed make it essential for optimizing workflows amid the rapid advancement of language models. _themes: nlp · tokenization · performance · rust_ #### [huggingface/kernels](https://github.com/huggingface/kernels) *Python · ★604 · Apache-2.0 · beta · score:0.70 · hot:0.73 · rising:0.73 · durable:0.63 · board:hot · trend:up* Build compute kernels and load them from the Hub. **Why it matters.** This repo enables building and loading compute kernels from Hugging Face Hub, allowing for portable, version-compatible optimizations primarily for PyTorch and CUDA workloads, which helps avoid traditional packaging limitations. It matters now as AI models demand efficient computations for better performance, but its niche focus on Hugging Face's ecosystem means adoption is mostly relevant for users already integrated with their tools, potentially limiting broader appeal. _themes: pytorch · optimization · inference · cuda_ #### [microsoft/msphpsql](https://github.com/microsoft/msphpsql) *PHP · ★1,868 · MIT · production · score:0.65 · hot:0.73 · rising:0.77 · durable:0.69 · board:rising · trend:up* Microsoft Drivers for PHP for SQL Server **Why it matters.** This repository provides Microsoft's official PHP extensions for connecting to SQL Server databases, including SQLSRV for procedural access and PDO_SQLSRV for PDO-based interactions, relying on the ODBC driver for low-level operations. It matters now because the latest release supports PHP 8.3 and later, addressing compatibility needs for developers maintaining or building applications in Microsoft ecosystems, though it has limitations and ongoing improvements are planned. _themes: php · sql-server · database · driver_ #### [NVIDIA/spark-rapids](https://github.com/NVIDIA/spark-rapids) *Scala · ★973 · Apache-2.0 · production · score:0.70 · hot:0.73 · rising:0.75 · durable:0.65 · board:rising · trend:up* Spark RAPIDS plugin - accelerate Apache Spark with GPUs **Why it matters.** The NVIDIA Spark RAPIDS plugin accelerates Apache Spark workloads by leveraging GPUs through the RAPIDS libraries, enabling faster processing for big data tasks while aiming for compatibility with Spark's results. It matters now as organizations seek to optimize data pipelines for cost and performance in AI/ML and analytics, though its adoption depends on GPU availability and specific workload needs, with potential limitations in non-GPU environments. _themes: big-data · gpu-acceleration · spark · data-processing_ #### [facebookresearch/xformers](https://github.com/facebookresearch/xformers) *Python · ★10,426 · NOASSERTION · experimental · score:0.80 · hot:0.73 · rising:0.74 · durable:0.65 · board:rising · trend:up* Hackable and optimized Transformers building blocks, supporting a composable construction. **Why it matters.** xFormers provides customizable and optimized building blocks for Transformers, enabling faster research iterations with custom CUDA kernels and composable components not yet in mainstream libraries like PyTorch. It matters now because Transformers are foundational in AI, and its focus on efficiency addresses the growing need for memory and speed optimizations in large-scale models, though its experimental status means it may require troubleshooting. _themes: transformers · optimization · inference · research_ #### [huggingface/xet-core](https://github.com/huggingface/xet-core) *Rust · ★470 · Apache-2.0 · production · score:0.70 · hot:0.73 · rising:0.73 · durable:0.66 · board:rising · trend:up* xet client tech, used in huggingface_hub **Why it matters.** xet-core is a Rust library that provides client technology for Xet storage, enabling efficient chunk-based deduplication, caching, and file handling specifically for Hugging Face Hub. It matters now because the growing size of AI models and datasets demands optimized storage solutions to reduce transfer times and costs, directly benefiting developers working within the Hugging Face ecosystem by improving performance and compatibility with existing tools. _themes: deduplication · storage · caching · rust_ #### [microsoft/multilspy](https://github.com/microsoft/multilspy) *Python · ★567 · MIT · experimental · score:0.60 · hot:0.73 · rising:0.69 · durable:0.59 · board:hot · trend:up* multilspy is a lsp client library in Python intended to be used to build applications around language servers. **Why it matters.** Multilspy is a Python library that serves as a client for the Language Server Protocol, enabling developers to integrate static analysis from various language servers into applications, particularly for AI-driven code generation as explored in a recent NeurIPS paper. It matters now because the growing use of large language models for coding tasks demands tools that ensure code accuracy and reduce errors, like those addressed by Monitor-Guided Decoding, amid the rapid adoption of AI in software development. _themes: code-generation · static-analysis · ai-for-code · lsp_ #### [NVIDIA/libnvidia-container](https://github.com/NVIDIA/libnvidia-container) *C · ★1,097 · Apache-2.0 · production · score:0.80 · hot:0.73 · rising:0.75 · durable:0.68 · board:rising · trend:up* NVIDIA container runtime library **Why it matters.** libnvidia-container is a C library and CLI tool that automatically configures Linux containers to utilize NVIDIA hardware, such as GPUs, by leveraging kernel primitives and remaining independent of specific container runtimes. It matters now because the growing adoption of AI, machine learning, and GPU-accelerated workloads in containerized environments demands reliable hardware integration for performance and efficiency in production deployments. _themes: containers · gpu · nvidia · linux_ #### [huggingface/speech-to-speech](https://github.com/huggingface/speech-to-speech) *Python · ★4,669 · Apache-2.0 · beta · score:0.75 · hot:0.73 · rising:0.76 · durable:0.72 · board:rising · trend:up* Build local voice agents with open-source models **Why it matters.** This repository provides a modular pipeline for building local voice agents using open-source models, including voice activity detection, speech-to-text, language models, and text-to-speech components, all integrated via Hugging Face's Transformers library. It matters now because the growing demand for privacy-focused, offline AI assistants requires accessible tools for rapid prototyping and deployment, especially amid advancements in speech AI that make such applications more feasible for developers. _themes: agents · speech · inference · open-source_ #### [NVIDIA/MatX](https://github.com/NVIDIA/MatX) *C++ · ★1,413 · BSD-3-Clause · production · score:0.70 · hot:0.72 · rising:0.74 · durable:0.68 · board:rising · trend:up* An efficient C++20 GPU numerical computing library with Python-like syntax **Why it matters.** MatX is a C++ library for GPU-accelerated numerical computing that mimics Python-like syntax, allowing developers to achieve high performance on NVIDIA hardware without extensive manual optimizations. It matters now as GPU computing demands grow in AI and high-performance computing, offering substantial speed improvements over tools like NumPy and CuPy, which can enhance productivity in data-intensive applications. _themes: gpu-computing · hpc · numerical · acceleration_ #### [google-deepmind/distrax](https://github.com/google-deepmind/distrax) *Python · ★632 · Apache-2.0 · beta · score:0.70 · hot:0.72 · rising:0.71 · durable:0.61 · board:hot · trend:up* **Why it matters.** Distrax is a lightweight JAX-native library that provides probability distributions and bijectors, serving as a simplified reimplementation of parts of TensorFlow Probability with a focus on readability and extensibility for custom ML applications. It matters now because JAX is increasingly adopted in research for its performance in areas like reinforcement learning, and Distrax fills a niche by offering an easy-to-extend alternative that enhances workflow efficiency without replicating TFP's full scope. _themes: jax · probability · distributions · bijectors_ #### [NVIDIA/cudnn-frontend](https://github.com/NVIDIA/cudnn-frontend) *Python · ★720 · MIT · production · score:0.70 · hot:0.72 · rising:0.73 · durable:0.64 · board:rising · trend:up* cudnn_frontend provides a c++ wrapper for the cudnn backend API and samples on how to use it **Why it matters.** cuDNN Frontend is a C++ header-only library with a Python interface that wraps NVIDIA's cuDNN backend API, providing access to high-performance GPU kernels for deep learning operations like matrix multiplications and activations. It matters now because NVIDIA is open-sourcing these kernels, enabling developers to inspect, modify, and optimize them for custom AI workloads, which is increasingly important amid the demand for efficient hardware acceleration in large-scale models and the shift towards FP8 precision. _themes: gpu · deep-learning · kernels · optimization_ #### [microsoft/monaco-editor](https://github.com/microsoft/monaco-editor) *JavaScript · ★45,888 · MIT · production · score:0.90 · hot:0.72 · rising:0.79 · durable:0.78 · board:rising · trend:up* A browser based code editor **Why it matters.** Monaco Editor is a browser-based code editor that offers advanced features like syntax highlighting, autocompletion, and debugging, essentially bringing VS Code's capabilities to web applications. It matters right now because it's a go-to choice for developers integrating rich code editing into tools like online IDEs or playgrounds, with active maintenance and broad adoption despite some deprecated features like AMD support. _themes: editor · browser · typescript · code-editing_ #### [facebookresearch/momentum](https://github.com/facebookresearch/momentum) *C++ · ★329 · MIT · beta · score:0.60 · hot:0.72 · rising:0.71 · durable:0.60 · board:hot · trend:up* A library for human kinematic motion and numerical optimization solvers to apply human motion **Why it matters.** Momentum is a library providing algorithms for human kinematic motion and numerical optimization solvers, primarily implemented in C++ with Python interfaces, aimed at applications like robotics and animation. It matters right now as advancements in AI and simulation demand efficient human motion tools, but its early release (v0.1.109) and moderate 329 stars indicate it's still niche and not yet widely adopted or proven in production. _themes: kinematics · optimization · simulation · robotics_ #### [anthropics/buffa](https://github.com/anthropics/buffa) *Rust · ★603 · Apache-2.0 · beta · score:0.70 · hot:0.72 · rising:0.71 · durable:0.64 · board:hot · trend:up* Rust implementation of protobuf with editions support, JSON serialization, and zero-copy views **Why it matters.** Buffa is a Rust library implementing Protocol Buffers with support for editions, JSON serialization, and zero-copy views, addressing a gap in the Rust ecosystem for a modern, editions-aware alternative. It matters because it passes full conformance tests and offers features like linear-time serialization and unknown field preservation, which could improve efficiency for developers handling Protobuf in production systems, though its early release version suggests it's still maturing. _themes: serialization · protobuf · rust · zero-copy_ #### [google-deepmind/optax](https://github.com/google-deepmind/optax) *Python · ★2,238 · Apache-2.0 · production · score:0.80 · hot:0.72 · rising:0.74 · durable:0.68 · board:rising · trend:up* Optax is a gradient processing and optimization library for JAX. **Why it matters.** Optax is a library for gradient processing and optimization specifically designed for JAX, offering composable building blocks to create custom optimizers. It matters now because JAX is increasingly used in ML research for its performance and flexibility, and Optax addresses the need for efficient, well-tested optimization tools that can accelerate development of new algorithms. However, its utility is limited to JAX ecosystems, making it less broadly applicable compared to more general frameworks. _themes: optimization · jax · machine-learning · gradients_ #### [microsoft/STL](https://github.com/microsoft/STL) *C++ · ★11,022 · NOASSERTION · production · score:0.85 · hot:0.71 · rising:0.76 · durable:0.70 · board:rising · trend:up* MSVC's implementation of the C++ Standard Library. **Why it matters.** This repository hosts Microsoft's implementation of the C++ Standard Library, which is a core component used in MSVC and Visual Studio for C++ development. It matters now because it's undergoing a GitHub migration, allowing for community contributions and transparency in bug fixes and feature additions, which could improve C++ ecosystem reliability amid ongoing language evolution. _themes: c++ · library · open-source · standards_ #### [GeeeekExplorer/nano-vllm](https://github.com/GeeeekExplorer/nano-vllm) *Python · ★12,999 · MIT · experimental · score:0.75 · hot:0.71 · rising:0.70 · durable:0.70 · board:hot · trend:up* Nano vLLM **Why it matters.** Nano-vLLM is a lightweight, from-scratch implementation of vLLM for efficient offline inference of large language models, featuring optimizations like tensor parallelism and CUDA graphs in a concise 1,200-line Python codebase. It matters now as it provides a readable alternative for developers working on resource-constrained hardware, potentially lowering barriers to LLM deployment, but its lack of formal releases and unproven stability compared to the original vLLM raise concerns about reliability and maintenance. _themes: inference · llm · optimization · pytorch_ #### [cfug/dio](https://github.com/cfug/dio) *Dart · ★12,814 · MIT · production · score:0.85 · hot:0.71 · rising:0.75 · durable:0.78 · board:durable · trend:up* A powerful HTTP client for Dart and Flutter, which supports global settings, Interceptors, FormData, aborting and canceling a request, files uploading and downloading, requests timeout, custom adapters, etc. **Why it matters.** Dio is an HTTP client library for Dart and Flutter that offers features like request interception, cancellation, timeouts, and file handling, streamlining network operations in applications. It matters now because reliable HTTP clients are critical for modern mobile and web apps dealing with APIs, and Dio's extensive adoption in the Flutter ecosystem provides a robust, customizable alternative to basic options, though it requires careful handling of breaking changes in updates. _themes: http · networking · interceptors · cancellable_ #### [deepseek-ai/DeepGEMM](https://github.com/deepseek-ai/DeepGEMM) *Cuda · ★6,652 · MIT · production · score:0.80 · hot:0.71 · rising:0.77 · durable:0.78 · board:durable · trend:up* DeepGEMM: clean and efficient FP8 GEMM kernels with fine-grained scaling **Why it matters.** DeepGEMM is a CUDA library providing high-performance GEMM kernels optimized for low-precision formats like FP8, FP4, and BF16, which are crucial for accelerating computations in large language models. It matters now because the AI industry is shifting towards efficient hardware utilization for training and inference, and DeepGEMM offers JIT-compiled kernels that match or exceed specialized libraries, making it valuable for optimizing GPU workloads amid growing demands for energy-efficient AI. Additionally, its features like fused MoE and MQA scoring address specific needs in modern model architectures. _themes: cuda · gemm · fp8 · optimization_ #### [facebookresearch/AugLy](https://github.com/facebookresearch/AugLy) *Python · ★5,078 · NOASSERTION · production · score:0.75 · hot:0.71 · rising:0.75 · durable:0.73 · board:rising · trend:up* A data augmentations library for audio, image, text, and video. **Why it matters.** AugLy is a Python library providing data augmentations for audio, image, text, and video modalities, with over 100 functions to enhance dataset diversity and test model robustness, particularly for real-world internet scenarios like social media interactions. It matters now because as AI applications increasingly deal with multimodal data in areas like content moderation and copyright detection, these augmentations help address gaps in model generalization, though its focus on platform-specific examples may limit broader applicability. _themes: data-augmentation · multimodal · robustness · ai-training_ #### [NVIDIA/nvshmem](https://github.com/NVIDIA/nvshmem) *C++ · ★510 · NOASSERTION · production · score:0.75 · hot:0.71 · rising:0.70 · durable:0.65 · board:hot · trend:up* NVIDIA NVSHMEM is a parallel programming interface for NVIDIA GPUs based on OpenSHMEM. NVSHMEM can significantly reduce multi-process communication and coordination overheads by allowing programmers to perform one-sided communication from within CUDA kernels and on CUDA streams. **Why it matters.** NVSHMEM is a library that extends OpenSHMEM for NVIDIA GPUs, enabling efficient one-sided communication directly from CUDA kernels to reduce overhead in multi-GPU setups. It matters now because distributed computing demands for AI training and HPC are growing, and this tool helps optimize performance by simplifying inter-GPU data access in large-scale environments. _themes: gpu-communication · parallel-programming · cuda · deep-learning_ #### [PaddlePaddle/PaddleSpeech](https://github.com/PaddlePaddle/PaddleSpeech) *Python · ★12,593 · Apache-2.0 · production · score:0.85 · hot:0.70 · rising:0.75 · durable:0.73 · board:rising · trend:up* Easy-to-use Speech Toolkit including Self-Supervised Learning model, SOTA/Streaming ASR with punctuation, Streaming TTS with text frontend, Speaker Verification System, End-to-End Speech Translation and Keyword Spotting. Won NAACL2022 Best Demo Award. **Why it matters.** PaddleSpeech is an open-source toolkit for speech and audio processing, offering features like state-of-the-art ASR, TTS, speech translation, and speaker verification, all built on PaddlePaddle for easy deployment and research. It matters now because speech AI is rapidly advancing in applications like virtual assistants and real-time translation, and this toolkit provides accessible, high-performance tools that have won awards, making it a practical choice for developers and researchers amid growing demand for multilingual and streaming capabilities. _themes: speech-recognition · tts · asr · self-supervised-learning_ #### [google-deepmind/mujoco_playground](https://github.com/google-deepmind/mujoco_playground) *Python · ★1,885 · Apache-2.0 · beta · score:0.70 · hot:0.70 · rising:0.72 · durable:0.66 · board:rising · trend:up* An open-source library for GPU-accelerated robot learning and sim-to-real transfer. **Why it matters.** MuJoCo Playground is a library providing GPU-accelerated simulation environments for robot learning tasks like locomotion and manipulation, built on DeepMind's MuJoCo MJX to support sim-to-real transfer. It matters now as robotics research increasingly relies on efficient simulations for training AI agents, especially with the growing emphasis on scalable hardware like GPUs, but it remains niche and dependent on the MuJoCo ecosystem. Its active development and integration with tools like JAX make it useful for accelerating experiments, though it may not address broader accessibility needs. _themes: robotics · simulation · reinforcement-learning · gpu_ #### [facebookresearch/fairseq2](https://github.com/facebookresearch/fairseq2) *Python · ★1,129 · MIT · beta · score:0.75 · hot:0.70 · rising:0.70 · durable:0.65 · board:rising · trend:up* FAIR Sequence Modeling Toolkit 2 **Why it matters.** fairseq2 is a Python-based toolkit for sequence modeling using PyTorch, designed to help researchers train custom models for content generation tasks, serving as a modular reboot of the original fairseq. It matters now because it supports recent advancements in AI research, as evidenced by its use in FAIR papers on topics like multilingual speech recognition and reinforcement learning for LLMs, offering a cleaner API amid the growing demand for flexible tools in generative AI. However, its differences from predecessors may require users to adapt existing workflows, potentially limiting immediate adoption. _themes: sequence-modeling · pytorch · deep-learning · generation_ #### [EleutherAI/gpt-neox](https://github.com/EleutherAI/gpt-neox) *Python · ★7,420 · Apache-2.0 · production · score:0.85 · hot:0.70 · rising:0.76 · durable:0.77 · board:durable · trend:up* An implementation of model parallel autoregressive transformers on GPUs, based on the Megatron and DeepSpeed libraries **Why it matters.** GPT-NeoX is a library for training large-scale autoregressive transformers on GPUs, leveraging Megatron and DeepSpeed for model parallelism, which supports handling models with billions of parameters. It matters now because the growing need for efficient large-language model training in research and industry requires accessible tools that work across diverse hardware, though its complexity makes it suitable only for advanced users dealing with high-scale AI experiments. _themes: transformers · language-models · large-scale-training · model-parallelism_ #### [google-research/google-research](https://github.com/google-research/google-research) *Jupyter Notebook · ★37,749 · Apache-2.0 · experimental · score:0.80 · hot:0.70 · rising:0.74 · durable:0.64 · board:rising · trend:up* Google Research **Why it matters.** This repository serves as a collection of code, datasets, and tools from Google Research's AI and machine learning projects, allowing users to access and experiment with implementations of various research papers. It matters because it provides valuable resources for advancing AI development, though the lack of a unified structure and varying project maturity can make it challenging to navigate. As an open-source hub from a leading organization, it remains relevant for staying updated on cutting-edge techniques amid rapid AI evolution. _themes: ai · machine-learning · research · datasets_ #### [huggingface/optimum](https://github.com/huggingface/optimum) *Python · ★3,359 · Apache-2.0 · production · score:0.80 · hot:0.70 · rising:0.72 · durable:0.72 · board:durable · trend:up* 🚀 Accelerate inference and training of 🤗 Transformers, Diffusers, TIMM and Sentence Transformers with easy to use hardware optimization tools **Why it matters.** Hugging Face Optimum is a library that extends Transformers, Diffusers, TIMM, and Sentence Transformers to optimize and accelerate model inference and training on specific hardware like Intel, NVIDIA, and others using tools for quantization and ONNX conversion. It matters now because the increasing size and complexity of AI models demand efficient hardware utilization to reduce costs and enable faster deployment, especially in resource-constrained environments where performance gains can directly impact real-world applications. _themes: inference · optimization · quantization · training_ #### [deepseek-ai/DeepEP](https://github.com/deepseek-ai/DeepEP) *Cuda · ★9,136 · MIT · beta · score:0.80 · hot:0.70 · rising:0.75 · durable:0.73 · board:rising · trend:up* DeepEP: an efficient expert-parallel communication library **Why it matters.** DeepEP is a communication library designed to optimize expert parallelism in Mixture-of-Experts models by providing high-throughput GPU kernels for all-to-all operations, supporting low-precision formats and hardware-specific features like NVLink and RDMA. It matters now because the growing adoption of MoE architectures in large-scale AI training demands efficient communication to reduce bottlenecks, and its optimizations align with current hardware trends, though it may have implementation differences from the referenced paper that users should verify. _themes: moe · distributed-training · gpu-kernels · inference_ #### [NVIDIA/cuEquivariance](https://github.com/NVIDIA/cuEquivariance) *Python · ★387 · no-license · beta · score:0.70 · hot:0.69 · rising:0.69 · durable:0.60 · board:hot · trend:up* cuEquivariance is a math library that is a collective of low-level primitives and tensor ops to accelerate widely-used models, like DiffDock, MACE, Allegro and NEQUIP, based on equivariant neural networks. Also includes kernels for accelerated structure prediction. **Why it matters.** cuEquivariance is a NVIDIA library providing optimized CUDA kernels and APIs for building equivariant neural networks, which accelerate models in scientific computing like molecular simulations by respecting spatial symmetries. It matters now because equivariant approaches improve data efficiency in AI for physics and drug discovery, but as a beta product, it may have stability issues and is limited to GPU environments, making it suitable only for targeted research workflows. _themes: equivariance · cuda · neural-networks · acceleration_ #### [pgvector/pgvector](https://github.com/pgvector/pgvector) *C · ★20,942 · NOASSERTION · production · score:0.85 · hot:0.69 · rising:0.73 · durable:0.71 · board:rising · trend:up* Open-source vector similarity search for Postgres **Why it matters.** pgvector is an extension for PostgreSQL that enables vector similarity search, allowing users to perform exact and approximate nearest neighbor searches on vectors stored directly in the database, supporting various data types and distance metrics. It matters now because the growing demand for AI-driven applications, such as recommendation systems and semantic search, requires efficient vector handling integrated with reliable databases, and pgvector provides this without the need for separate vector databases, leveraging PostgreSQL's strengths for scalability and ACID compliance. _themes: vector-search · nearest-neighbor · database · ai_ #### [OptimalScale/LMFlow](https://github.com/OptimalScale/LMFlow) *Python · ★8,484 · Apache-2.0 · production · score:0.80 · hot:0.69 · rising:0.72 · durable:0.77 · board:durable · trend:stable* An Extensible Toolkit for Finetuning and Inference of Large Foundation Models. Large Models for All. **Why it matters.** LMFlow is an extensible toolkit built with PyTorch for finetuning and inference of large language models, making it easier to adapt pretrained models for specific tasks without needing extensive custom code. It matters now due to the growing demand for accessible LLM customization amid advancements in AI, as evidenced by its recent updates, award at NAACL 2024, and support for emerging model architectures like Hymba. _themes: fine-tuning · inference · language-model · pytorch_ #### [google-deepmind/torax](https://github.com/google-deepmind/torax) *Python · ★661 · NOASSERTION · beta · score:0.70 · hot:0.69 · rising:0.69 · durable:0.61 · board:rising · trend:up* TORAX: Tokamak transport simulation in JAX **Why it matters.** TORAX is a differentiable simulator for tokamak core transport using JAX, enabling fast simulations for fusion research tasks like trajectory optimization and controller design. It matters now because the push for practical fusion energy requires advanced computational tools, and TORAX's auto-differentiation features facilitate efficient modeling and integration with machine learning, potentially accelerating progress in sustainable energy development amid global research efforts. _themes: jax · simulation · optimization · physics_ #### [google-deepmind/gemma](https://github.com/google-deepmind/gemma) *Python · ★4,963 · Apache-2.0 · production · score:0.70 · hot:0.68 · rising:0.74 · durable:0.71 · board:rising · trend:up* Gemma open-weight LLM library, from Google DeepMind **Why it matters.** Gemma is an open-weights Large Language Model library from Google DeepMind, built with JAX for inference and fine-tuning of their models, including multi-modal capabilities. It matters now as it provides accessible tools for experimenting with advanced AI models in a competitive landscape, though its real-world impact depends on the quality and ease of integration compared to established alternatives. _themes: fine-tuning · inference · multi-modal · jax_ #### [huggingface/optimum-intel](https://github.com/huggingface/optimum-intel) *Jupyter Notebook · ★566 · Apache-2.0 · production · score:0.70 · hot:0.68 · rising:0.70 · durable:0.67 · board:rising · trend:up* 🤗 Optimum Intel: Accelerate inference with Intel optimization tools **Why it matters.** Hugging Face Optimum Intel serves as an interface between popular ML libraries like Transformers and Diffusers and Intel's OpenVINO toolkit, enabling optimized inference through techniques such as quantization, pruning, and distillation on Intel hardware. It matters now because the demand for efficient AI deployment on edge devices and diverse architectures is growing, helping developers reduce latency and resource usage in real-world applications amid increasing model sizes and computational constraints. _themes: inference · optimization · quantization · intel-hardware_ #### [facebookresearch/detectron2](https://github.com/facebookresearch/detectron2) *Python · ★34,332 · Apache-2.0 · production · score:0.90 · hot:0.68 · rising:0.76 · durable:0.80 · board:durable · trend:up* Detectron2 is a platform for object detection, segmentation and other visual recognition tasks. **Why it matters.** Detectron2 is a PyTorch-based library for object detection, segmentation, and other visual recognition tasks, serving as a successor to earlier Facebook tools with improved performance and features like panoptic segmentation and ViTDet. It matters now because it supports both research and production workflows in computer vision, which is increasingly critical for applications in AI-driven fields like autonomous systems and image analysis, though it requires solid PyTorch knowledge to use effectively. _themes: computer-vision · object-detection · segmentation · pytorch_ #### [huggingface/datatrove](https://github.com/huggingface/datatrove) *Python · ★2,995 · Apache-2.0 · beta · score:0.70 · hot:0.68 · rising:0.70 · durable:0.67 · board:rising · trend:up* Freeing data processing from scripting madness by providing a set of platform-agnostic customizable pipeline processing blocks. **Why it matters.** DataTrove is a Python library that provides modular, customizable pipelines for processing, filtering, and deduplicating large-scale text data, making it easier to handle workloads like LLM training without custom scripting. It matters now because the growing demand for high-quality datasets in AI requires efficient, scalable tools, and this addresses that by being platform-agnostic and low-memory, though its 0.9.0 release suggests it's still maturing and may have rough edges. _themes: data-processing · scalability · text-filtering · llm-training_ #### [huggingface/setfit](https://github.com/huggingface/setfit) *Jupyter Notebook · ★2,716 · Apache-2.0 · production · score:0.85 · hot:0.68 · rising:0.71 · durable:0.71 · board:rising · trend:stable* Efficient few-shot learning with Sentence Transformers **Why it matters.** SetFit is a library that enables efficient few-shot learning by fine-tuning Sentence Transformers with minimal labeled data, eliminating the need for prompts and achieving high accuracy quickly. It matters right now because few-shot learning addresses data scarcity in NLP tasks, offering a faster alternative to traditional fine-tuning methods amid growing demands for resource-efficient AI solutions, especially in multilingual applications. _themes: few-shot-learning · nlp · fine-tuning · efficiency_ #### [microsoft/BotFramework-WebChat](https://github.com/microsoft/BotFramework-WebChat) *HTML · ★1,770 · MIT · production · score:0.70 · hot:0.67 · rising:0.73 · durable:0.66 · board:rising · trend:up* A highly-customizable web-based client for Azure Bot Services. **Why it matters.** BotFramework-WebChat is an open-source web component that provides a customizable interface for building chatbots using Azure Bot Services, allowing developers to integrate conversational AI into web applications. It matters because conversational interfaces remain essential for customer engagement and automation, but its relevance is tempered by Microsoft's archiving of the core Bot Framework SDK in late 2025, potentially signaling a shift away from this ecosystem for new projects. Developers with existing implementations may still find it valuable, though alternatives are worth considering for long-term viability. _themes: conversational-ai · chatbot · web-ui · adaptive-cards_ #### [google-deepmind/kfac-jax](https://github.com/google-deepmind/kfac-jax) *Python · ★321 · Apache-2.0 · beta · score:0.60 · hot:0.67 · rising:0.67 · durable:0.59 · board:rising · trend:up* Second Order Optimization and Curvature Estimation with K-FAC in JAX. **Why it matters.** KFAC-JAX is a library for implementing second-order optimization using the K-FAC method in JAX, which approximates curvature to potentially speed up neural network training. It matters now because advanced optimization techniques are increasingly needed for efficient large-scale model training in research, but its early version (v0.0.8) and limited adoption (320 stars) indicate it's not yet a mature solution for production use. _themes: optimization · jax · second-order · curvature_ #### [microsoft/SandDance](https://github.com/microsoft/SandDance) *TypeScript · ★7,119 · MIT · production · score:0.60 · hot:0.67 · rising:0.72 · durable:0.70 · board:rising · trend:up* Visually explore, understand, and present your data. **Why it matters.** SandDance is an open-source TypeScript library for interactive data visualization, allowing users to explore and present datasets through unit visualizations and animated transitions, with integrations into tools like Power BI. It matters now because it offers modular, embeddable components for custom applications in a data-driven world, though its v3 status and Microsoft focus might limit broader adoption compared to more versatile alternatives. _themes: data-visualization · interactive-viz · data-exploration · embedding_ #### [facebookresearch/vrs](https://github.com/facebookresearch/vrs) *C++ · ★418 · Apache-2.0 · production · score:0.65 · hot:0.67 · rising:0.71 · durable:0.67 · board:rising · trend:stable* VRS is a file format optimized to record & playback streams of sensor data, such as images, audio samples, and any other discrete sensors (IMU, temperature, etc), stored in per-device streams of timestamped records. **Why it matters.** VRS is a C++ library implementing a file format for efficiently recording and playing back timestamped streams of sensor data, such as images, audio, and IMU readings, which is particularly useful for AR/VR development and sensor fusion tasks. It matters now because the growing adoption of wearable devices and autonomous systems demands optimized data handling for prototyping and analysis, as demonstrated by its use in Meta's Quest and Aria projects, though its niche focus limits broader appeal. _themes: sensor-data · file-format · streaming · efficiency_ #### [microsoft/wil](https://github.com/microsoft/wil) *C++ · ★2,907 · MIT · production · score:0.75 · hot:0.67 · rising:0.71 · durable:0.67 · board:rising · trend:up* Windows Implementation Library **Why it matters.** Microsoft's WIL is a header-only C++ library that offers type-safe wrappers and helpers for common Windows API patterns, such as resource management with RAII, Win32 functions, registry operations, and network APIs, making Windows development safer and more efficient. It matters right now because Windows remains a dominant platform for enterprise and desktop applications, and WIL helps developers avoid error-prone boilerplate code amid the ongoing shift to modern C++ standards, potentially reducing bugs in production systems. This is particularly relevant as organizations continue to maintain and update legacy Windows software. _themes: c++ · windows · raii · api-wrappers_ #### [facebookresearch/Pearl](https://github.com/facebookresearch/Pearl) *Jupyter Notebook · ★2,992 · MIT · beta · score:0.70 · hot:0.66 · rising:0.69 · durable:0.65 · board:rising · trend:up* A Production-ready Reinforcement Learning AI Agent Library brought by the Applied Reinforcement Learning team at Meta. **Why it matters.** Pearl is an open-source library for building production-ready reinforcement learning agents that handle complex environments with limited observability and sparse feedback, developed by Meta's Applied Reinforcement Learning team. It matters now because reinforcement learning is increasingly applied in real-world AI systems for decision-making, and this library provides accessible tools for researchers and practitioners to prototype and deploy RL agents amid growing interest in AI agents. However, as a beta release without a formal latest version, its stability for immediate production use remains unproven. _themes: rl · agents · production · decision-making_ #### [NVIDIA/nvbench](https://github.com/NVIDIA/nvbench) *Cuda · ★854 · Apache-2.0 · beta · score:0.70 · hot:0.66 · rising:0.67 · durable:0.63 · board:rising · trend:stable* CUDA Kernel Benchmarking Library **Why it matters.** NVBench is a C++ library from NVIDIA for benchmarking CUDA kernels, offering features like parameter sweeps, throughput calculations, and runtime customization to measure GPU performance accurately. It matters now because optimizing CUDA code is critical for AI, machine learning, and high-performance computing workloads where GPU efficiency directly impacts results. However, as it's still a work-in-progress, it may not be ready for widespread adoption yet. _themes: benchmark · cuda · gpu · performance_ #### [google-deepmind/deepmind-research](https://github.com/google-deepmind/deepmind-research) *Jupyter Notebook · ★14,844 · Apache-2.0 · experimental · score:0.85 · hot:0.66 · rising:0.70 · durable:0.64 · board:rising · trend:up* This repository contains implementations and illustrative code to accompany DeepMind publications **Why it matters.** This repository provides code implementations and illustrative examples for DeepMind's research publications, primarily in AI and machine learning, allowing researchers to reproduce and build upon their work. It matters now because it offers access to high-quality, peer-reviewed code from a leading AI lab, which can accelerate innovation in fields like reinforcement learning and neural networks, though its lack of formal releases may limit practical adoption. However, the collection is somewhat fragmented, as it points to other repos rather than containing comprehensive implementations. _themes: reinforcement-learning · deep-learning · research · open-source_ #### [microsoft/LMOps](https://github.com/microsoft/LMOps) *Python · ★4,349 · MIT · experimental · score:0.70 · hot:0.66 · rising:0.67 · durable:0.61 · board:rising · trend:up* General technology for enabling AI capabilities w/ LLMs and MLLMs **Why it matters.** Microsoft LMOps is a research-oriented repository focusing on technologies for enhancing AI capabilities with large language models (LLMs), including prompt optimization, longer context handling, and alignment techniques, primarily through a collection of papers and code snippets. It matters right now because the rapid growth of generative AI demands better tools for efficient LLM deployment and customization, addressing key challenges in prompt engineering and model performance amid increasing real-world applications. _themes: prompt-engineering · llm-optimization · alignment · inference_ #### [huggingface/deep-rl-class](https://github.com/huggingface/deep-rl-class) *MDX · ★4,849 · Apache-2.0 · archived · score:0.80 · hot:0.66 · rising:0.67 · durable:0.58 · board:rising · trend:up* This repo contains the Hugging Face Deep Reinforcement Learning Course. **Why it matters.** This repository provides a comprehensive course on deep reinforcement learning from Hugging Face, including theory explanations and practical notebooks for hands-on exercises. It matters because reinforcement learning remains a key area in AI development, offering accessible resources for learners, though its low-maintenance status means some features are outdated and it may not reflect the latest advancements. _themes: reinforcement-learning · deep-learning · education · hands-on_ #### [NVIDIA/cutile-python](https://github.com/NVIDIA/cutile-python) *Python · ★2,025 · NOASSERTION · experimental · score:0.70 · hot:0.66 · rising:0.66 · durable:0.59 · board:rising · trend:stable* cuTile is a programming model for writing parallel kernels for NVIDIA GPUs **Why it matters.** cuTile Python is a programming model that enables writing parallel kernels for NVIDIA GPUs using a tile-based approach in Python, simplifying GPU programming for tasks like vector addition. It matters right now because the growing demand for efficient parallel computing in AI and machine learning workloads requires easier access to GPU acceleration, and NVIDIA's tools could streamline development for performance-critical applications. _themes: gpu · parallel-kernels · tile-programming · python_ #### [NVIDIA/makani](https://github.com/NVIDIA/makani) *Python · ★372 · NOASSERTION · experimental · score:0.60 · hot:0.66 · rising:0.66 · durable:0.59 · board:rising · trend:stable* Massively parallel training of machine-learning based weather and climate models **Why it matters.** Makani is a PyTorch library for massively parallel training of machine-learning-based weather and climate models, supporting features like data parallelism and asynchronous loading on multiple GPUs. It matters now because climate change demands faster and more accurate predictive models, and this tool aids researchers in scaling experiments, though its research-focused nature means it's not yet ready for widespread production use. _themes: parallel-training · weather-ml · climate-modeling · pytorch_ #### [microsoft/vs-streamjsonrpc](https://github.com/microsoft/vs-streamjsonrpc) *C# · ★915 · NOASSERTION · production · score:0.70 · hot:0.66 · rising:0.68 · durable:0.60 · board:rising · trend:up* The StreamJsonRpc library offers JSON-RPC 2.0 over any .NET Stream, WebSocket, or Pipe. With bonus support for request cancellation, client proxy generation, and more. **Why it matters.** StreamJsonRpc is a .NET library that implements JSON-RPC 2.0 for communication over streams, WebSockets, or pipes, enabling efficient remote procedure calls in distributed systems. It matters now because it extends the standard with features like request cancellation and dynamic proxy generation, which are crucial for modern .NET applications dealing with real-time data exchange and scalable architectures, especially in tools like Visual Studio. _themes: rpc · json · dotnet · serialization_ #### [google-deepmind/mujoco_menagerie](https://github.com/google-deepmind/mujoco_menagerie) *Python · ★3,311 · NOASSERTION · beta · score:0.70 · hot:0.65 · rising:0.66 · durable:0.60 · board:rising · trend:up* A collection of high-quality models for the MuJoCo physics engine, curated by Google DeepMind. **Why it matters.** This repository provides a curated collection of high-quality models for the MuJoCo physics engine, aimed at ensuring reliable simulations for robotics research without the need for users to debug common issues. It matters now because accurate physical modeling is critical for advancing AI-driven robotics experiments, especially amid growing interest in reinforcement learning, though the lack of a specified license and formal releases could hinder broader adoption. _themes: simulation · robotics · physics-engine · models_ #### [NVIDIA/CUDALibrarySamples](https://github.com/NVIDIA/CUDALibrarySamples) *Cuda · ★2,372 · Apache-2.0 · production · score:0.75 · hot:0.64 · rising:0.67 · durable:0.62 · board:rising · trend:up* CUDA Library Samples **Why it matters.** This repository provides sample code for NVIDIA's CUDA libraries, demonstrating GPU-accelerated operations for tasks like linear algebra, image processing, and data compression, which helps developers integrate these libraries into their applications. It matters right now because the demand for high-performance computing in AI, scientific simulations, and edge computing is surging, making efficient GPU utilization essential for optimizing performance and reducing computation times. However, its value is niche, primarily for those already invested in NVIDIA ecosystems, as alternatives exist from other hardware vendors. _themes: gpu · acceleration · linear-algebra · image-processing_ #### [huggingface/swift-transformers](https://github.com/huggingface/swift-transformers) *Swift · ★1,305 · Apache-2.0 · beta · score:0.60 · hot:0.64 · rising:0.66 · durable:0.67 · board:durable · trend:stable* Swift Package to implement a transformers-like API in Swift **Why it matters.** Swift-transformers is a library that ports Hugging Face's transformers API to Swift, offering utilities for tokenization, model downloads, and chat templating to integrate language models into Swift applications. It matters now because the increasing focus on on-device AI for Apple platforms requires efficient, native tools, reducing reliance on Python backends and enabling faster development for iOS and macOS apps, though it's still limited compared to the original Python library in terms of features and ecosystem support. _themes: inference · tokenization · llm_ #### [NVIDIA/stdexec](https://github.com/NVIDIA/stdexec) *C++ · ★2,312 · Apache-2.0 · experimental · score:0.65 · hot:0.64 · rising:0.65 · durable:0.55 · board:rising · trend:up* `std::execution`, the proposed C++ framework for asynchronous and parallel programming. **Why it matters.** stdexec is an experimental reference implementation of the proposed C++26 std::execution for asynchronous and parallel programming, including NVIDIA-specific extensions for GPU execution. It matters right now because it allows developers to experiment with upcoming C++ standards early, potentially improving performance in high-computational tasks, though its experimental status means it's not ready for production and may change. _themes: async · parallel · c++ · gpu_ #### [google-research/scenic](https://github.com/google-research/scenic) *Python · ★3,794 · Apache-2.0 · experimental · score:0.75 · hot:0.64 · rising:0.64 · durable:0.58 · board:rising · trend:up* Scenic: A Jax Library for Computer Vision Research and Beyond **Why it matters.** Scenic is a JAX library that offers shared utilities and baselines for training attention-based computer vision models, including tools for large-scale experiments on images, video, and multimodal data. It matters right now because JAX's growing adoption in research enables efficient development of vision transformers and other advanced models, addressing common pain points in scaling and prototyping, though its lack of formal releases may limit immediate applicability for production users. _themes: computer-vision · jax · attention · transformers_ #### [huggingface/computer-vision-course](https://github.com/huggingface/computer-vision-course) *Jupyter Notebook · ★817 · MIT · beta · score:0.70 · hot:0.64 · rising:0.65 · durable:0.58 · board:rising · trend:stable* This repo is the homebase of a community driven course on Computer Vision with Neural Networks. Feel free to join us on the Hugging Face discord: hf.co/join/discord **Why it matters.** This repository provides a community-driven, open-source course on computer vision, featuring Jupyter notebooks that cover topics from fundamentals to advanced areas like transformers and generative models. It matters now because computer vision is a key component of modern AI applications, and this resource offers accessible, diverse educational content amid the rapid evolution of AI technologies, though its community-led nature means content quality can vary. _themes: computer-vision · neural-networks · transformers · education_ #### [NVIDIA/apex](https://github.com/NVIDIA/apex) *Python · ★8,948 · BSD-3-Clause · beta · score:0.85 · hot:0.63 · rising:0.67 · durable:0.62 · board:rising · trend:up* A PyTorch Extension: Tools for easy mixed precision and distributed training in Pytorch **Why it matters.** Apex is a NVIDIA-maintained extension for PyTorch that provides tools to simplify mixed precision training for faster GPU utilization and distributed training for scaling across devices. It matters now because efficient training is critical for handling increasingly large AI models, and Apex bridges gaps in PyTorch's ecosystem by offering utilities that may soon be integrated upstream, helping developers optimize workflows amid rapid AI advancements. _themes: pytorch · mixed-precision · distributed-training · optimization_ #### [huggingface/alignment-handbook](https://github.com/huggingface/alignment-handbook) *Python · ★5,571 · Apache-2.0 · beta · score:0.80 · hot:0.63 · rising:0.65 · durable:0.67 · board:durable · trend:stable* Robust recipes to align language models with human and AI preferences **Why it matters.** This repository from Hugging Face provides practical recipes and guidelines for aligning language models using techniques like Reinforcement Learning from Human Feedback (RLHF), covering data collection, training, and evaluation to ensure models better reflect human and AI preferences. It addresses a critical gap in accessible resources amid the rapid proliferation of LLMs, where alignment is essential for improving safety and helpfulness, as demonstrated in recent models like Llama2. However, its lack of a formal release and reliance on a technical report means it may not yet offer fully polished implementations, potentially requiring users to adapt recipes to their needs. _themes: llm · rlhf · fine-tuning · alignment_ #### [google-deepmind/acme](https://github.com/google-deepmind/acme) *Python · ★3,968 · Apache-2.0 · beta · score:0.75 · hot:0.63 · rising:0.67 · durable:0.69 · board:durable · trend:stable* A library of reinforcement learning components and agents **Why it matters.** Acme is a Python library from Google DeepMind that provides modular components and agents for reinforcement learning, serving as reference implementations and baselines for research while supporting scalability. It matters now because reinforcement learning remains crucial for advancing AI in areas like robotics and decision-making, offering researchers reliable tools from a leading organization to build and experiment with algorithms efficiently, though it may lack the polish needed for widespread production adoption. _themes: agents · reinforcement-learning · research_ #### [google-deepmind/dm-haiku](https://github.com/google-deepmind/dm-haiku) *Python · ★3,216 · Apache-2.0 · archived · score:0.30 · hot:0.63 · rising:0.63 · durable:0.53 · board:hot · trend:up* JAX-based neural network library **Why it matters.** Haiku is a JAX-based library for building and training neural networks, offering a simple API inspired by Sonnet for deep learning research. It matters right now primarily for maintaining existing projects, as Google DeepMind has shifted focus to Flax and placed Haiku in maintenance mode with no new features, making it less suitable for new developments. _themes: jax · neural-networks · deep-learning · machine-learning_ #### [microsoft/LLMLingua](https://github.com/microsoft/LLMLingua) *Python · ★6,037 · MIT · beta · score:0.80 · hot:0.63 · rising:0.67 · durable:0.72 · board:durable · trend:stable* [EMNLP'23, ACL'24] To speed up LLMs' inference and enhance LLM's perceive of key information, compress the prompt and KV-Cache, which achieves up to 20x compression with minimal performance loss. **Why it matters.** LLMLingua compresses prompts and KV-Cache for large language models, achieving up to 20x reduction in size with minimal accuracy loss, which speeds up inference and helps manage costs in resource-intensive AI applications. This is particularly relevant now as enterprises and developers deal with escalating computational demands for long-context LLMs, and it builds on established research from EMNLP and ACL conferences. Recent extensions like SCBench and RetrievalAttention further enhance its utility for efficient long-context processing. _themes: inference · compression · llm · prompt-engineering_ #### [facebookresearch/HolisticTraceAnalysis](https://github.com/facebookresearch/HolisticTraceAnalysis) *Python · ★505 · MIT · beta · score:0.70 · hot:0.63 · rising:0.65 · durable:0.62 · board:rising · trend:stable* A library to analyze PyTorch traces. **Why it matters.** Holistic Trace Analysis is a Python library that analyzes PyTorch traces to identify performance bottlenecks in distributed training workloads, offering features like temporal breakdowns, kernel analysis, and trace comparisons. It matters now because optimizing GPU utilization and communication overlaps is essential for efficient large-scale AI training, helping developers reduce costs and improve model training times amid growing demands for high-performance computing. _themes: pytorch · profiling · performance-optimization · distributed-training_ #### [mistralai/mistral-common](https://github.com/mistralai/mistral-common) *Python · ★882 · Apache-2.0 · production · score:0.70 · hot:0.63 · rising:0.67 · durable:0.72 · board:durable · trend:stable* Official inference library for pre-processing of Mistral models **Why it matters.** Mistral-common provides tools for tokenizing and validating inputs for Mistral AI models, including support for text, images, and tool calls, while ensuring backward compatibility. It matters now because as Mistral's models gain adoption in AI applications, this library standardizes pre-processing to reduce integration errors and support custom model development, though its value is limited to users specifically working with Mistral ecosystems. _themes: tokenization · inference · validation · models_ #### [deepseek-ai/FlashMLA](https://github.com/deepseek-ai/FlashMLA) *C++ · ★12,555 · MIT · beta · score:0.85 · hot:0.63 · rising:0.68 · durable:0.71 · board:durable · trend:stable* FlashMLA: Efficient Multi-head Latent Attention Kernels **Why it matters.** FlashMLA is a library providing optimized C++ kernels for multi-head attention in AI models, including sparse and dense variants that enhance efficiency during prefill and decoding stages, as used in DeepSeek's models. It matters now because optimizing attention mechanisms is critical for handling larger models and longer contexts in real-time applications, potentially reducing computational costs amid growing demands for AI inference. With recent updates achieving high TFlops rates, it offers practical improvements for performance-critical workloads. _themes: attention · optimization · inference · efficiency_ #### [microsoft/clarity](https://github.com/microsoft/clarity) *TypeScript · ★2,617 · MIT · production · score:0.80 · hot:0.62 · rising:0.66 · durable:0.61 · board:rising · trend:stable* A behavioral analytics library that uses dom mutations and user interactions to generate aggregated insights. **Why it matters.** Microsoft Clarity is an open-source TypeScript library that tracks user interactions and DOM changes on websites to provide insights like heatmaps and session replays, emphasizing privacy and performance. It matters now because businesses are increasingly prioritizing user privacy amid stricter regulations while seeking actionable analytics to improve website engagement and retention in a competitive digital landscape. _themes: analytics · privacy · session-replay · heatmap_ #### [facebookresearch/dinov2](https://github.com/facebookresearch/dinov2) *Jupyter Notebook · ★12,716 · Apache-2.0 · beta · score:0.90 · hot:0.62 · rising:0.67 · durable:0.67 · board:durable · trend:stable* PyTorch code and models for the DINOv2 self-supervised learning method. **Why it matters.** DINOv2 is a PyTorch-based library for self-supervised learning in computer vision, enabling the training of robust visual features without labeled data, which is particularly useful for applications requiring scalable image understanding. It matters right now due to ongoing updates integrating it with specialized domains like X-ray analysis and biology, building on its established popularity and influence in reducing annotation costs for AI models. _themes: self-supervised · vision · pytorch · feature-learning_ #### [EleutherAI/sparsify](https://github.com/EleutherAI/sparsify) *Python · ★711 · MIT · beta · score:0.70 · hot:0.62 · rising:0.65 · durable:0.67 · board:durable · trend:stable* Sparsify transformers with SAEs and transcoders **Why it matters.** Sparsify is a Python library that trains sparse autoencoders (SAEs) on activations from HuggingFace language models, enforcing sparsity via a TopK approach to potentially enhance model efficiency and interpretability, as outlined in recent research. It matters now because the AI field is grappling with the need for more computationally efficient models, but its on-the-fly activation computation may hinder iterative experimentation compared to alternatives that cache data, potentially limiting its practicality for some users. _themes: sparsity · autoencoders · transformers · efficiency_ #### [microsoft/playwright-python](https://github.com/microsoft/playwright-python) *Python · ★14,533 · Apache-2.0 · production · score:0.80 · hot:0.62 · rising:0.69 · durable:0.78 · board:durable · trend:stable* Python version of the Playwright testing and automation library. **Why it matters.** Playwright-python is a library that automates web browsers like Chromium, Firefox, and WebKit using a unified API, primarily for testing and automation tasks in Python. It matters now because reliable cross-browser automation is essential for modern web development and CI/CD workflows, especially as web applications grow more complex, and this library offers better performance and maintenance than some legacy tools. Its active updates ensure compatibility with the latest browser versions. _themes: web-automation · testing · cross-browser_ #### [chrisliu298/awesome-llm-unlearning](https://github.com/chrisliu298/awesome-llm-unlearning) *? · ★569 · Apache-2.0 · beta · score:0.75 · hot:0.62 · rising:0.62 · durable:0.62 · board:durable · trend:stable* A resource repository for machine unlearning in large language models **Why it matters.** This repository curates a collection of resources, including 475 papers, 17 surveys, and other materials on machine unlearning for large language models, serving as a centralized hub for research in this area. It matters now because unlearning techniques are increasingly critical for addressing AI safety, data privacy regulations, and model alignment issues amid rapid advancements in LLMs, helping researchers stay updated on emerging methods. _themes: unlearning · llm · alignment · machine-learning_ #### [facebookresearch/eb_jepa](https://github.com/facebookresearch/eb_jepa) *Python · ★559 · Apache-2.0 · experimental · score:0.70 · hot:0.61 · rising:0.62 · durable:0.58 · board:rising · trend:stable* An open source library designed to provide community examples of Joint Embedding Predictive Architectures (JEPAs). It contains code and examples for learning representations from images, video, and action-conditioned video, as well as planning using JEPA-based models. **Why it matters.** This repository offers an open-source library and examples for Joint Embedding Predictive Architectures (JEPAs), focusing on self-supervised learning of representations from images, videos, and action-conditioned videos for prediction and planning tasks. It matters now because JEPA advances unsupervised learning techniques, which are increasingly relevant for efficient AI development in areas like world modeling and planning, especially amid growing interest in autonomous agents; however, its experimental nature and lack of a formal release mean it's more of a research tool than a ready-to-use solution. _themes: self-supervised · representation-learning · video · planning_ #### [huggingface/finetrainers](https://github.com/huggingface/finetrainers) *Python · ★1,353 · Apache-2.0 · beta · score:0.65 · hot:0.61 · rising:0.62 · durable:0.66 · board:durable · trend:stable* Scalable and memory-optimized training of diffusion models **Why it matters.** Finetrainers is a library for scalable and memory-optimized training of diffusion models, building on Hugging Face's ecosystem to make training more accessible and efficient. It matters now because diffusion models are widely used in generative AI tasks like image and video creation, but training them is resource-intensive, and this library addresses those challenges amid growing demand for optimized tools in research and development. _themes: diffusion-models · fine-tuning · pytorch · scalable-training_ #### [openai/tiktoken](https://github.com/openai/tiktoken) *Python · ★17,956 · MIT · production · score:0.85 · hot:0.61 · rising:0.69 · durable:0.81 · board:durable · trend:stable* tiktoken is a fast BPE tokeniser for use with OpenAI's models. **Why it matters.** Tiktoken is a fast BPE tokenizer optimized for OpenAI's language models, providing efficient text encoding and decoding for AI workflows. It matters now because tokenization speed is crucial for scaling applications with large language models, and its integration with OpenAI's ecosystem helps developers handle performance bottlenecks in real-time processing. However, its specificity to OpenAI limits broader applicability compared to more general tokenizers. _themes: tokenization · nlp · optimization · inference_ #### [google-research/language](https://github.com/google-research/language) *Python · ★1,770 · Apache-2.0 · experimental · score:0.65 · hot:0.61 · rising:0.62 · durable:0.53 · board:rising · trend:up* Shared repository for open-sourced projects from the Google AI Language team. **Why it matters.** This repository aggregates open-sourced projects from Google's AI Language team, focusing on natural language processing and machine learning research, offering code and implementations for various experimental ideas. It matters for researchers exploring cutting-edge NLP techniques, as it provides insights from a major tech player, but its lack of recent releases and status as not an official product may reduce its practical utility in fast-evolving AI landscapes. _themes: nlp · ml · research · fine-tuning_ #### [google-research/perch](https://github.com/google-research/perch) *Python · ★342 · Apache-2.0 · experimental · score:0.60 · hot:0.61 · rising:0.62 · durable:0.57 · board:rising · trend:stable* **Why it matters.** Perch is a repository for code artifacts related to bioacoustic research, including training models like a bird species classifier using EfficientNet on audio spectrograms, based on datasets of over 10k species. It matters for advancing transfer learning in bioacoustics, supporting applications in wildlife conservation and environmental monitoring as highlighted in associated papers, but its outdated code and reliance on deprecated TensorFlow parts make it less practical for immediate use compared to alternatives. _themes: bioacoustics · audio-classification · transfer-learning · efficientnet_ #### [microsoft/KBLaM](https://github.com/microsoft/KBLaM) *Jupyter Notebook · ★1,445 · MIT · experimental · score:0.70 · hot:0.61 · rising:0.62 · durable:0.57 · board:rising · trend:stable* Official Implementation of "KBLaM: Knowledge Base augmented Language Model" **Why it matters.** KBLaM is a research method for integrating knowledge bases into language models without external retrieval modules, claiming linear scaling with knowledge base size compared to quadratic overhead in in-context learning. It matters now as efficient knowledge augmentation addresses key limitations in LLMs for real-world applications like accurate information retrieval, but its effectiveness is unproven beyond the paper's experiments and requires further validation in diverse scenarios. _themes: llm-augmentation · knowledge-base · efficiency · scaling_ #### [huggingface/nanotron](https://github.com/huggingface/nanotron) *Python · ★2,656 · Apache-2.0 · beta · score:0.70 · hot:0.61 · rising:0.64 · durable:0.66 · board:durable · trend:stable* Minimalistic large language model 3D-parallelism training **Why it matters.** Nanotron is a Hugging Face library that provides a simple API for pretraining transformer-based large language models using 3D-parallelism, focusing on speed and scalability for custom datasets. It matters now because efficient LLM training is a critical bottleneck in AI development due to high computational costs, but its early release version (v0.4) suggests it may still have limitations in stability and widespread adoption. _themes: llm-training · parallelism · scalability · pretraining_ #### [facebookresearch/multimodal](https://github.com/facebookresearch/multimodal) *Python · ★1,714 · BSD-3-Clause · beta · score:0.70 · hot:0.60 · rising:0.62 · durable:0.58 · board:rising · trend:stable* TorchMultimodal is a PyTorch library for training state-of-the-art multimodal multi-task models at scale. **Why it matters.** TorchMultimodal is a PyTorch library that provides modular components, pre-built models like ALBEF and BLIP-2, and examples for training multimodal multi-task models at scale, focusing on content understanding and generative tasks. It matters now because multimodal AI is advancing rapidly in areas like vision-language models, offering researchers a practical toolkit to replicate and build on state-of-the-art research, though it's still in beta and may lack full stability for production use. _themes: multimodal · pytorch · multi-task · generative_ #### [google-deepmind/reverb](https://github.com/google-deepmind/reverb) *C++ · ★773 · Apache-2.0 · experimental · score:0.70 · hot:0.60 · rising:0.60 · durable:0.52 · board:rising · trend:stable* Reverb is an efficient and easy-to-use data storage and transport system designed for machine learning research **Why it matters.** Reverb is a data storage and transport system optimized for machine learning research, particularly as an experience replay buffer for distributed reinforcement learning algorithms, supporting structures like FIFO, LIFO, and priority queues. It matters now because efficient data handling is essential for advancing RL experiments, but its experimental nature and lack of production hardening limit broader adoption amid growing demands for robust ML infrastructure. _themes: rl · data-storage · distributed-ml · experience-replay_ #### [openai/harmony](https://github.com/openai/harmony) *Rust · ★4,317 · Apache-2.0 · beta · score:0.65 · hot:0.59 · rising:0.62 · durable:0.64 · board:durable · trend:stable* Renderer for the harmony response format to be used with gpt-oss **Why it matters.** OpenAI's harmony repo provides a Rust-based renderer for the Harmony response format, which structures outputs from gpt-oss models to handle conversation, reasoning, and tool calls in a standardized way. This is relevant now as the proliferation of open-source AI models requires robust formatting for custom inference setups, but its niche focus on gpt-oss limits broader applicability, and the lack of a latest release raises questions about maintenance and stability. _themes: inference · tool-calling · response-format_ #### [microsoft/node-pty](https://github.com/microsoft/node-pty) *TypeScript · ★1,878 · NOASSERTION · production · score:0.75 · hot:0.59 · rising:0.63 · durable:0.65 · board:durable · trend:stable* Fork pseudoterminals in Node.JS **Why it matters.** Node-pty is a Node.js library that provides bindings for forking pseudoterminals, allowing developers to spawn and interact with shell processes as if they were in a terminal, supporting Linux, macOS, and Windows via the conpty API. It matters right now because terminal emulation is essential for modern web-based tools like IDEs and remote shells, and its Microsoft backing ensures reliable cross-platform support, though it requires recent Windows versions and has dropped older dependencies. However, its last release was in 2021, raising questions about ongoing maintenance for new Node.js versions or security updates. _themes: terminal · pty · emulation · cross-platform_ #### [google-deepmind/chex](https://github.com/google-deepmind/chex) *Python · ★936 · Apache-2.0 · beta · score:0.70 · hot:0.59 · rising:0.61 · durable:0.62 · board:durable · trend:stable* **Why it matters.** Chex is a Python library that provides utilities for debugging, testing, and instrumenting JAX code, helping developers write more reliable machine learning programs by handling tasks like transforming parallel maps and ensuring compatibility with JAX data structures. It matters now because JAX is widely used in AI research for its performance benefits, and as ML projects grow in complexity, robust tooling like Chex is essential to reduce errors and improve productivity in development workflows. _themes: jax · debugging · testing · pytrees_ #### [google-research/FLAN](https://github.com/google-research/FLAN) *Python · ★1,561 · Apache-2.0 · beta · score:0.75 · hot:0.59 · rising:0.62 · durable:0.58 · board:rising · trend:stable* **Why it matters.** This repository provides Python code to generate datasets for instruction tuning language models, specifically the FLAN 2021 and Flan Collection datasets, which are used to fine-tune models like Flan-T5 and Flan-PaLM for zero-shot learning capabilities. It matters now because instruction tuning is a key technique in advancing AI's ability to follow complex instructions, amid growing interest in efficient fine-tuning methods for large language models, though the repo lacks recent updates and formal releases, potentially limiting its immediate applicability. _themes: fine-tuning · datasets · language-models · instruction-tuning_ #### [openai/CLIP](https://github.com/openai/CLIP) *Jupyter Notebook · ★33,241 · MIT · production · score:0.90 · hot:0.58 · rising:0.67 · durable:0.75 · board:durable · trend:stable* CLIP (Contrastive Language-Image Pretraining), Predict the most relevant text snippet given an image **Why it matters.** CLIP is a pre-trained neural network that aligns images and text through contrastive learning, enabling zero-shot image classification by matching images to textual descriptions without task-specific training. It matters because it has influenced multimodal AI research by demonstrating effective vision-language integration, though its performance can vary across datasets and it relies on large-scale pretraining that may not be accessible to all users. _themes: contrastive-learning · zero-shot-learning · multimodal · vision-language_ #### [huggingface/hf_transfer](https://github.com/huggingface/hf_transfer) *Rust · ★554 · Apache-2.0 · experimental · score:0.60 · hot:0.58 · rising:0.59 · durable:0.58 · board:rising · trend:stable* **Why it matters.** hf_transfer is a Rust library designed to accelerate file transfers to and from the Hugging Face Hub, enabling speeds beyond 500MB/s on high-bandwidth networks where Python-based tools fall short. It matters now because the increasing size of AI models and datasets demands efficient data handling for power users, reducing bottlenecks in workflows for advanced machine learning tasks. _themes: file-transfer · performance · huggingface · rust_ #### [huggingface/text-generation-inference](https://github.com/huggingface/text-generation-inference) *Python · ★10,842 · Apache-2.0 · production · score:0.70 · hot:0.57 · rising:0.62 · durable:0.74 · board:durable · trend:stable* Large Language Model Text Generation Inference **Why it matters.** Hugging Face's Text Generation Inference (TGI) is a server for deploying and serving large language models for text generation, built with Rust, Python, and gRPC, and has been used in production for services like Hugging Chat. However, it's now in maintenance mode, meaning it's stable for existing users but not recommended for new projects due to the shift towards more optimized alternatives like vLLM and SGLang. This makes it relevant for legacy inference setups but less innovative in the rapidly evolving LLM inference landscape. _themes: inference · llm · nlp · transformer_ #### [allenai/ir_datasets](https://github.com/allenai/ir_datasets) *Python · ★387 · Apache-2.0 · beta · score:0.70 · hot:0.57 · rising:0.58 · durable:0.61 · board:durable · trend:stable* Provides a common interface to many IR ranking datasets. **Why it matters.** ir_datasets is a Python library that provides a unified interface for accessing and managing various information retrieval (IR) datasets, including downloading and handling documents, queries, and relevance judgments, which simplifies data preparation for IR tasks. This is relevant now as the demand for robust IR systems in AI applications like search engines and recommendation systems grows, but it remains a niche tool that doesn't address broader data ecosystem challenges. However, its standardization efforts help streamline research workflows amid increasing dataset fragmentation. _themes: ir · datasets · retrieval · ranking_ #### [google-deepmind/regress-lm](https://github.com/google-deepmind/regress-lm) *Python · ★335 · Apache-2.0 · experimental · score:0.60 · hot:0.57 · rising:0.58 · durable:0.55 · board:rising · trend:stable* Library for text-to-text regression, applicable to any input string representation and allows pretraining and fine-tuning over multiple regression tasks. **Why it matters.** RegressLM is a Python library from Google DeepMind for performing text-to-text regression, allowing users to handle string inputs and output continuous values through pretraining and fine-tuning on various tasks. It matters now because it addresses a niche in NLP for regression problems, such as in scientific simulations, but its lack of a formal release and modest 335 stars suggest it's still early-stage and not yet widely validated or adopted. _themes: regression · fine-tuning · NLP · pretraining_ #### [google-research/sparf](https://github.com/google-research/sparf) *Python · ★300 · Apache-2.0 · experimental · score:0.70 · hot:0.57 · rising:0.57 · durable:0.56 · board:rising · trend:stable* This is the official code release for SPARF: Neural Radiance Fields from Sparse and Noisy Poses [CVPR 2023-Highlight] **Why it matters.** SPARF is a method that uses Neural Radiance Fields to generate realistic novel views from as few as 2-3 input images with noisy camera poses, incorporating novel constraints like multi-view correspondence and depth-consistency losses for pose refinement. It matters now because it addresses practical challenges in 3D reconstruction with limited data, which is increasingly relevant for applications in augmented reality, robotics, and computer vision, though its effectiveness is limited to experimental settings without broader validation or production-ready tools. _themes: nerf · novel-view-synthesis · pose-refinement · computer-vision_ #### [microsoft/memento](https://github.com/microsoft/memento) *Python · ★339 · MIT · experimental · score:0.70 · hot:0.56 · rising:0.56 · durable:0.59 · board:durable · trend:stable* **Why it matters.** Memento is a technique and library from Microsoft that extends the output length of large language models by breaking chain-of-thought reasoning into blocks and summaries, evicting processed content from the KV cache to fit more reasoning within fixed context limits. This matters now because as LLMs are applied to increasingly complex tasks, efficient context management reduces computational costs and improves performance, addressing a key bottleneck in real-world AI applications like agents and advanced reasoning systems. _themes: inference · kv-cache · reasoning · summarization_ #### [microsoft/msticpy](https://github.com/microsoft/msticpy) *Python · ★1,957 · NOASSERTION · production · score:0.75 · hot:0.56 · rising:0.61 · durable:0.70 · board:durable · trend:stable* Microsoft Threat Intelligence Security Tools **Why it matters.** msticpy is a Python library designed for security investigations and threat hunting in Jupyter Notebooks, enabling users to query logs from various sources, enrich data with threat intelligence, and perform advanced analysis and visualizations. It matters now because cybersecurity threats are escalating, and this tool streamlines incident response for Microsoft Sentinel users while expanding to other platforms, making it practical for enterprise security operations amid growing data volumes. _themes: security · threat-hunting · data-enrichment · visualization_ #### [huggingface/optimum-quanto](https://github.com/huggingface/optimum-quanto) *Python · ★1,037 · Apache-2.0 · archived · score:0.40 · hot:0.56 · rising:0.55 · durable:0.61 · board:durable · trend:stable* A pytorch quantization backend for optimum **Why it matters.** Huggingface/optimum-quanto provides a PyTorch quantization backend for the Optimum library, enabling efficient model compression and faster inference through features like automatic quantization stubs and support for various integer and float precisions. However, it's in maintenance mode with no major updates planned, making it less relevant for new projects compared to actively developed alternatives, as it lacks full feature parity and ongoing improvements. This limits its immediate utility for production environments where reliability and performance are critical. _themes: quantization · pytorch · inference · optimization_ #### [google-research/jax3d](https://github.com/google-research/jax3d) *Python · ★761 · Apache-2.0 · experimental · score:0.65 · hot:0.56 · rising:0.56 · durable:0.52 · board:rising · trend:stable* **Why it matters.** Jax3d is a repository of JAX-based projects for 3D computer vision tasks, including generative models and NeRF implementations, providing tools for researchers to experiment with efficient computations in 3D AI. It matters now due to the increasing demand for 3D content generation and neural rendering in research, but its lack of official releases and experimental nature means it's not yet polished for broader adoption. _themes: jax · 3d · nerf · generative_ #### [google-deepmind/rlax](https://github.com/google-deepmind/rlax) *Python · ★1,418 · Apache-2.0 · beta · score:0.70 · hot:0.55 · rising:0.59 · durable:0.66 · board:durable · trend:stable* **Why it matters.** RLax is a JAX-based library that provides modular building blocks for reinforcement learning, including operations for values, policies, and distributional RL, aimed at helping developers implement agents efficiently. It matters now because JAX's hardware acceleration capabilities make it relevant for modern AI research, especially in RL where experimentation is key, though its early release version suggests it may lack the robustness of more mature alternatives. With DeepMind's backing, it offers reliable implementations but isn't yet a must-use for production. _themes: rl · jax · agents · deep-learning_ #### [microsoft/Bringing-Old-Photos-Back-to-Life](https://github.com/microsoft/Bringing-Old-Photos-Back-to-Life) *Python · ★15,692 · MIT · production · score:0.85 · hot:0.54 · rising:0.64 · durable:0.80 · board:durable · trend:stable* Bringing Old Photo Back to Life (CVPR 2020 oral) **Why it matters.** This repository provides a PyTorch implementation of a GAN-based model for restoring old, degraded photos by removing scratches, restoring colors, and enhancing details, as presented in a CVPR 2020 paper. It matters now because AI-driven image restoration is increasingly important for preserving historical archives and personal memories, with growing applications in cultural heritage and media amid advancements in generative AI. _themes: gans · image-restoration · pytorch · generative-models_ #### [facebookresearch/pytorch3d](https://github.com/facebookresearch/pytorch3d) *Python · ★9,857 · NOASSERTION · production · score:0.85 · hot:0.54 · rising:0.61 · durable:0.70 · board:durable · trend:stable* PyTorch3D is FAIR's library of reusable components for deep learning with 3D data **Why it matters.** PyTorch3D is a library from Facebook's FAIR that offers reusable components for handling 3D data in PyTorch, including mesh operations, rendering, and frameworks for tasks like new-view synthesis, which are essential for 3D computer vision research. It matters now because 3D deep learning is increasingly critical for applications in AR/VR, robotics, and multimodal AI, providing efficient, differentiable tools that integrate seamlessly with PyTorch workflows, though its focus on research may limit immediate production adoption. _themes: 3d-vision · pytorch · mesh-processing · rendering_ #### [huggingface/optimum-nvidia](https://github.com/huggingface/optimum-nvidia) *Python · ★1,032 · Apache-2.0 · beta · score:0.70 · hot:0.54 · rising:0.58 · durable:0.64 · board:durable · trend:stable* **Why it matters.** Optimum-NVIDIA is a library that optimizes AI model inference on NVIDIA hardware, integrating seamlessly with Hugging Face's Transformers to achieve significant speedups, such as running LLaMA 2 at up to 1,200 tokens per second with minimal code changes. It matters now because efficient inference is critical for scaling AI applications in production, especially with the rising demand for real-time processing of large language models on widely available NVIDIA GPUs, potentially reducing costs and improving performance in resource-constrained environments. _themes: inference · optimization · nvidia · llm_ #### [intel/intel-extension-for-pytorch](https://github.com/intel/intel-extension-for-pytorch) *Python · ★2,008 · Apache-2.0 · archived · score:0.20 · hot:0.53 · rising:0.53 · durable:0.52 · board:hot · trend:stable* A Python package for extending the official PyTorch that can easily obtain performance on Intel platform **Why it matters.** This repository provides an extension to the official PyTorch framework, offering optimizations like quantization for better performance on Intel CPUs and GPUs. However, it has been archived with no further development or support, as most features have been upstreamed into PyTorch itself, making it irrelevant for new projects. Users should transition to standard PyTorch for ongoing Intel platform compatibility. _themes: pytorch · quantization · optimization · deep-learning_ #### [microsoft/Detours](https://github.com/microsoft/Detours) *C++ · ★6,221 · MIT · production · score:0.70 · hot:0.53 · rising:0.61 · durable:0.72 · board:durable · trend:stable* Detours is a software package for monitoring and instrumenting API calls on Windows. It is distributed in source code form. **Why it matters.** Detours is a C++ library for intercepting and modifying Windows API calls, allowing developers to monitor and instrument software behavior for purposes like debugging and security analysis. It matters because it's a mature tool from Microsoft that's now open source under the MIT license, making it easier for developers to use in production environments without licensing hurdles, especially amid growing needs for low-level system monitoring in enterprise applications. However, its relevance is limited to Windows-specific development and may not address modern cross-platform needs. _themes: api-hooking · windows · instrumentation · debugging_ #### [facebookresearch/Kats](https://github.com/facebookresearch/Kats) *Python · ★6,293 · MIT · beta · score:0.70 · hot:0.53 · rising:0.60 · durable:0.72 · board:durable · trend:stable* Kats, a kit to analyze time series data, a lightweight, easy-to-use, generalizable, and extendable framework to perform time series analysis, from understanding the key statistics and characteristics, detecting change points and anomalies, to forecasting future trends. **Why it matters.** Kats is a Python library developed by Facebook for time series analysis, providing tools for statistics, anomaly detection, forecasting, and more in a lightweight and extensible manner. It matters because time series data is critical in fields like infrastructure monitoring, finance, and IoT, where efficient analysis can drive real-time decisions, but its version 0.2.0 status indicates it may still have rough edges compared to more mature alternatives. _themes: time-series · forecasting · anomaly-detection · stats_ #### [microsoft/unilm](https://github.com/microsoft/unilm) *Python · ★22,101 · MIT · beta · score:0.85 · hot:0.53 · rising:0.60 · durable:0.73 · board:durable · trend:stable* Large-scale Self-supervised Pre-training Across Tasks, Languages, and Modalities **Why it matters.** Microsoft's UniLM repository provides a collection of large-scale self-supervised pre-trained models and architectures for tasks across languages, modalities, and domains, including innovations like scalable Transformers and multimodal AI. It matters right now because the AI community is focused on efficient foundation models for general-purpose applications, and this repo offers practical research tools that address challenges in training stability, efficiency, and multimodal integration amid rapid advancements in generative AI. However, its Microsoft-centric focus means it may not be as broadly applicable or optimized for independent use as some alternatives. _themes: multimodal · nlp · pre-training · foundation-models_ #### [huggingface/screenenv](https://github.com/huggingface/screenenv) *Python · ★452 · MIT · experimental · score:0.70 · hot:0.53 · rising:0.53 · durable:0.58 · board:durable · trend:stable* A powerful Python library for creating and managing isolated desktop environments using Docker containers. **Why it matters.** ScreenEnv is a Python library that uses Docker to create isolated Ubuntu desktop environments for GUI automation, testing, and development, offering features like mouse/keyboard control and integration with tools like Playwright. It matters now as it addresses the growing demand for reliable, containerized testing in AI workflows, particularly for LLM automation, though its lack of official releases raises questions about stability and maintenance. _themes: automation · testing · docker · gui_ #### [SciPhi-AI/R2R](https://github.com/SciPhi-AI/R2R) *Python · ★7,771 · MIT · production · score:0.80 · hot:0.53 · rising:0.60 · durable:0.78 · board:durable · trend:stable* SoTA production-ready AI retrieval system. Agentic Retrieval-Augmented Generation (RAG) with a RESTful API. **Why it matters.** R2R is a production-ready AI retrieval system that facilitates Retrieval-Augmented Generation (RAG) through a RESTful API, supporting features like multimodal content ingestion, hybrid search, knowledge graphs, and agentic deep research for complex queries. It matters now because RAG is essential for improving the accuracy and reliability of large language model outputs in real-world applications, especially as enterprises seek scalable solutions for handling dynamic data sources amid increasing AI adoption, though its claims of being 'most advanced' require verification against evolving competitors. _themes: rag · agents · retrieval · search_ #### [openai/apps-sdk-ui](https://github.com/openai/apps-sdk-ui) *TypeScript · ★879 · MIT · beta · score:0.60 · hot:0.52 · rising:0.56 · durable:0.64 · board:durable · trend:stable* **Why it matters.** This repo provides a design system and React component library optimized for building apps with OpenAI's Apps SDK, focusing on consistent and accessible UIs for ChatGPT integrations. It matters now because OpenAI is expanding its developer ecosystem, but its early version (v0.2.1) and dependency on specific React and Tailwind versions limit its broader applicability, making it useful only for targeted app development rather than a general solution. _themes: ui · react · design-system · chatgpt_ #### [karpathy/llm.c](https://github.com/karpathy/llm.c) *Cuda · ★29,664 · MIT · experimental · score:0.85 · hot:0.52 · rising:0.60 · durable:0.71 · board:durable · trend:stable* LLM training in simple, raw C/CUDA **Why it matters.** This repository provides a minimal implementation of large language model training using raw C and CUDA, focusing on reproducing models like GPT-2 and GPT-3 without dependencies like PyTorch, emphasizing efficiency and simplicity. It matters now because it highlights potential performance gains in resource-constrained environments and serves as an educational tool for understanding low-level AI training, though its experimental nature means it's not yet ready for widespread production use. _themes: llm-training · cuda · efficiency · pretraining_ #### [google-deepmind/pysc2](https://github.com/google-deepmind/pysc2) *Python · ★8,277 · Apache-2.0 · production · score:0.80 · hot:0.52 · rising:0.62 · durable:0.76 · board:durable · trend:stable* StarCraft II Learning Environment **Why it matters.** PySC2 is a Python library that provides an interface for reinforcement learning agents to interact with StarCraft II, serving as a benchmark environment for AI research in complex decision-making and multi-agent systems. It matters now because it continues to support advanced RL experiments, though its popularity has waned with the rise of more modern benchmarks like those from OpenAI or newer game-based environments. However, it remains a valuable tool for researchers studying strategic AI in partially observable settings. _themes: reinforcement-learning · ai-research · game-ai · benchmark_ #### [huggingface/nfsserve](https://github.com/huggingface/nfsserve) *Rust · ★723 · BSD-3-Clause · experimental · score:0.50 · hot:0.51 · rising:0.52 · durable:0.55 · board:durable · trend:stable* A Rust NFS Server implementation **Why it matters.** This repository implements a Rust-based NFSv3 server for user-mode file systems, enabling cross-platform mounting without relying on FUSE drivers, which simplifies file access in tools like Hugging Face's hf-mount. It matters now as AI/ML developers increasingly need seamless, OS-agnostic file handling for workflows, though its incompleteness limits broader adoption. _themes: rust · nfs · filesystem · cross-platform_ #### [facebookresearch/habitat-sim](https://github.com/facebookresearch/habitat-sim) *C++ · ★3,630 · MIT · beta · score:0.85 · hot:0.51 · rising:0.58 · durable:0.67 · board:durable · trend:stable* A flexible, high-performance 3D simulator for Embodied AI research. **Why it matters.** Habitat-Sim is a high-performance 3D simulator optimized for Embodied AI research, providing fast rendering and physics simulations for environments like indoor scans and robotic interactions. It matters right now because it enables efficient training of AI agents in simulated settings, accelerating advancements in robotics and sim2real transfer amid growing interest in autonomous systems. This tool helps bridge the gap between virtual training and real-world deployment, making it essential for AI research in resource-constrained scenarios. _themes: simulation · robotics · computer-vision · embodied-ai_ #### [openai/chatgpt-retrieval-plugin](https://github.com/openai/chatgpt-retrieval-plugin) *Python · ★21,208 · MIT · beta · score:0.85 · hot:0.51 · rising:0.60 · durable:0.71 · board:durable · trend:stable* The ChatGPT Retrieval Plugin lets you easily find personal or work documents by asking questions in natural language. **Why it matters.** This repository provides a backend for ChatGPT to enable semantic search and retrieval of personal or organizational documents using natural language queries, allowing for customizable control over aspects like document chunking and embedding models. It matters now because it supports the integration of proprietary data into AI workflows, which is increasingly important for productivity in enterprise settings amid the rise of custom GPTs, though its lack of recent releases and deprecated elements may limit long-term reliability. _themes: rag · semantic-search · chatgpt · embeddings_ #### [microsoft/FASTER](https://github.com/microsoft/FASTER) *C# · ★6,617 · MIT · production · score:0.80 · hot:0.50 · rising:0.60 · durable:0.76 · board:durable · trend:stable* Fast persistent recoverable log and key-value store + cache, in C# and C++. **Why it matters.** FASTER provides a high-performance persistent log and key-value store with cache capabilities, designed for concurrent access, heavy updates, and recovery in cloud environments, making it suitable for applications handling large-scale state management. It matters now because modern cloud applications demand low-latency, resilient storage solutions amid growing data volumes and real-time processing needs, where FASTER's performance claims could address bottlenecks in existing systems. However, its advantages should be verified against specific workloads, as real-world results may vary. _themes: concurrency · persistence · key-value · recovery_ #### [karpathy/nanoGPT](https://github.com/karpathy/nanoGPT) *Python · ★57,032 · MIT · archived · score:0.40 · hot:0.50 · rising:0.56 · durable:0.63 · board:durable · trend:stable* The simplest, fastest repository for training/finetuning medium-sized GPTs. **Why it matters.** NanoGPT is a minimal Python repository for training and fine-tuning medium-sized GPT models, offering a simple 300-line training loop and model definition that makes it easy to experiment or adapt for custom needs. However, it is now deprecated in favor of its successor, nanochat, and is considered outdated, so it primarily serves as a historical reference rather than a current tool for active development. _themes: fine-tuning · gpt · training · simplicity_ #### [meta-llama/llama-models](https://github.com/meta-llama/llama-models) *Python · ★7,568 · NOASSERTION · beta · score:0.70 · hot:0.50 · rising:0.57 · durable:0.68 · board:durable · trend:stable* Utilities intended for use with Llama models. **Why it matters.** This repo provides utilities for working with Meta's Llama large language models, including tools for integration, experimentation, and ensuring trust and safety. It matters now because the generative AI field is expanding rapidly, and accessible utilities for popular open LLMs like Llama enable developers and researchers to innovate responsibly amid growing adoption and regulatory scrutiny. _themes: llm · inference · fine-tuning · safety_ #### [microsoft/Swin-Transformer](https://github.com/microsoft/Swin-Transformer) *Python · ★15,856 · MIT · production · score:0.90 · hot:0.50 · rising:0.59 · durable:0.73 · board:durable · trend:stable* This is an official implementation for "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows". **Why it matters.** Swin Transformer is an official implementation of a hierarchical vision transformer that uses shifted windows to improve efficiency and performance in computer vision tasks like image classification and object detection, addressing limitations in traditional transformers by incorporating locality and scalability. It matters because it has become a foundational model in the field, influencing subsequent research and applications due to its state-of-the-art results on benchmarks like ImageNet, and its open-source availability has enabled widespread adoption in academic and industry settings for tasks such as semantic segmentation. _themes: vision-transformers · computer-vision · image-classification · object-detection_ #### [facebookresearch/BenchMARL](https://github.com/facebookresearch/BenchMARL) *Python · ★607 · MIT · beta · score:0.75 · hot:0.50 · rising:0.53 · durable:0.65 · board:durable · trend:stable* BenchMARL is a library for benchmarking Multi-Agent Reinforcement Learning (MARL). BenchMARL allows to quickly compare different MARL algorithms, tasks, and models while being systematically grounded in its two core tenets: reproducibility and standardization. **Why it matters.** BenchMARL is a Python library designed for benchmarking Multi-Agent Reinforcement Learning (MARL) algorithms and environments, emphasizing reproducibility and standardization through integrations like TorchRL and Hydra. It matters now because the increasing complexity of multi-agent systems in research areas such as robotics requires reliable tools for fair comparisons, helping advance the field by ensuring consistent and statistically robust evaluations. _themes: marl · benchmarking · reinforcement-learning · multi-agent_ #### [google-deepmind/tree](https://github.com/google-deepmind/tree) *Python · ★1,021 · Apache-2.0 · beta · score:0.65 · hot:0.50 · rising:0.53 · durable:0.61 · board:durable · trend:stable* tree is a library for working with nested data structures **Why it matters.** Tree is a Python library that provides utilities for handling nested data structures, such as flattening and mapping functions while preserving the structure, extending the built-in map for more complex scenarios. It matters for machine learning workflows where efficient manipulation of hierarchical data is common, but its impact is limited by its niche focus and modest adoption with only 1021 stars, suggesting it's not a transformative tool for most developers right now. _themes: data-structures · functional-programming · ml · performance_ #### [openai/spinningup](https://github.com/openai/spinningup) *Python · ★11,727 · MIT · production · score:0.70 · hot:0.50 · rising:0.60 · durable:0.72 · board:durable · trend:stable* An educational resource to help anyone learn deep reinforcement learning. **Why it matters.** Spinning Up is an OpenAI educational resource providing tutorials, code implementations, and exercises for learning deep reinforcement learning fundamentals. It matters because it offers a structured introduction to RL concepts, which remain foundational for AI research and development, though its content is from 2018 and may not reflect recent advancements in the field. _themes: rl · deep-rl · education · algorithms_ #### [google-deepmind/sonnet](https://github.com/google-deepmind/sonnet) *Python · ★9,916 · Apache-2.0 · production · score:0.60 · hot:0.50 · rising:0.60 · durable:0.74 · board:durable · trend:stable* TensorFlow-based neural network library **Why it matters.** Sonnet is a TensorFlow-based library that offers modular abstractions for building neural networks, primarily through its snt.Module system, allowing users to create and compose custom components for machine learning research. It matters now because it provides a flexible, unopinionated approach for researchers working within the TensorFlow ecosystem, though its adoption may be limited by the growing popularity of alternatives like PyTorch, and it hasn't seen major updates recently despite its established use in DeepMind projects. _themes: deep-learning · neural-networks · tensorflow · modules_ #### [facebookresearch/audiocraft](https://github.com/facebookresearch/audiocraft) *Jupyter Notebook · ★23,200 · MIT · beta · score:0.85 · hot:0.50 · rising:0.58 · durable:0.70 · board:durable · trend:stable* Audiocraft is a library for audio processing and generation with deep learning. It features the state-of-the-art EnCodec audio compressor / tokenizer, along with MusicGen, a simple and controllable music generation LM with textual and melodic conditioning. **Why it matters.** Audiocraft is a PyTorch library for audio generation and processing, offering models like MusicGen for text-to-music creation and AudioGen for text-to-sound, along with EnCodec for audio compression. It matters right now as generative AI for audio is rapidly evolving, with applications in research, music production, and multimedia, and this library provides accessible, state-of-the-art tools from a major research entity to advance these fields. _themes: audio-generation · music-gen · deep-learning · inference_ #### [NVIDIA/DeepLearningExamples](https://github.com/NVIDIA/DeepLearningExamples) *Jupyter Notebook · ★14,784 · no-license · production · score:0.85 · hot:0.50 · rising:0.57 · durable:0.66 · board:durable · trend:stable* State-of-the-Art Deep Learning scripts organized by models - easy to train and deploy with reproducible accuracy and performance on enterprise-grade infrastructure. **Why it matters.** This repository provides a collection of optimized deep learning examples for various models, including computer vision and NLP, designed for easy training and deployment on NVIDIA GPUs with reproducible results. It matters now because the growing demand for efficient AI workflows in enterprise settings requires reliable, high-performance resources, and NVIDIA's examples leverage their hardware ecosystem to address this, though the lack of a clear license may hinder broader adoption. _themes: deep-learning · inference · optimization · gpu_ #### [microsoft/LoRA](https://github.com/microsoft/LoRA) *Python · ★13,444 · MIT · production · score:0.90 · hot:0.49 · rising:0.57 · durable:0.73 · board:durable · trend:stable* Code for loralib, an implementation of "LoRA: Low-Rank Adaptation of Large Language Models" **Why it matters.** LoRA is a library that implements low-rank adaptation for efficiently fine-tuning large language models by training only a small subset of parameters, significantly reducing computational and storage costs. It matters right now because the growing size of LLMs makes traditional fine-tuning impractical for many applications, and LoRA's integration into tools like Hugging Face's PEFT library enhances accessibility for custom model adaptations in research and industry. _themes: fine-tuning · low-rank-adaptation · pytorch · llm_ #### [openai/DALL-E](https://github.com/openai/DALL-E) *Python · ★10,867 · NOASSERTION · beta · score:0.80 · hot:0.49 · rising:0.56 · durable:0.67 · board:durable · trend:stable* PyTorch package for the discrete VAE used for DALL·E. **Why it matters.** This repository provides the official PyTorch implementation of the discrete Variational Autoencoder (VAE) used in OpenAI's DALL·E for encoding images, but it does not include the full text-to-image transformer model. It matters now as it offers insights into the foundational components of early generative AI models, enabling researchers to experiment and build upon DALL·E's architecture, though its utility is limited without the complete system and it's not actively maintained. _themes: vae · generative-ai · pytorch · image-encoding_ #### [facebookresearch/SlowFast](https://github.com/facebookresearch/SlowFast) *Python · ★7,344 · Apache-2.0 · beta · score:0.85 · hot:0.49 · rising:0.55 · durable:0.62 · board:durable · trend:stable* PySlowFast: video understanding codebase from FAIR for reproducing state-of-the-art video models. **Why it matters.** PySlowFast is an open-source PyTorch codebase from Facebook AI Research that implements state-of-the-art video understanding models, such as SlowFast Networks and various transformers, for tasks like video classification and detection, enabling efficient training and evaluation. It matters now because video data analysis is increasingly critical in AI applications like surveillance and content moderation, and this repo provides accessible tools for advancing research in a field where multimodal and efficient models are in high demand. _themes: video-recognition · pytorch · state-of-the-art · self-supervised_ #### [openai/shap-e](https://github.com/openai/shap-e) *Python · ★12,236 · MIT · experimental · score:0.80 · hot:0.49 · rising:0.56 · durable:0.67 · board:durable · trend:stable* Generate 3D objects conditioned on text or images **Why it matters.** Shap-E is an OpenAI repository that provides tools for generating 3D objects from text prompts or images using implicit functions, as detailed in their research paper. It matters now because it advances 3D content creation in AI, potentially impacting applications in design and gaming, but its lack of formal releases and reliance on experimental setups limits immediate practical adoption. _themes: 3d-generation · text-to-3d · image-to-3d · generative-ai_ #### [huggingface/nanoVLM](https://github.com/huggingface/nanoVLM) *Python · ★4,820 · Apache-2.0 · beta · score:0.70 · hot:0.49 · rising:0.57 · durable:0.72 · board:durable · trend:stable* The simplest, fastest repository for training/finetuning small-sized VLMs. **Why it matters.** nanoVLM provides a lightweight PyTorch-based library for training and fine-tuning small vision-language models, emphasizing simplicity and speed to make advanced AI tasks accessible with minimal resources. It matters now due to the increasing demand for efficient AI models in resource-constrained environments, with recent updates supporting features like image splitting and multi-node training, though these come with breaking changes that may disrupt existing workflows. _themes: fine-tuning · vlm · pytorch · efficiency_ #### [google-research/federated](https://github.com/google-research/federated) *Python · ★752 · Apache-2.0 · experimental · score:0.70 · hot:0.49 · rising:0.51 · durable:0.55 · board:durable · trend:stable* A collection of Google research projects related to Federated Learning and Federated Analytics. **Why it matters.** This repository provides code for reproducing Google Research experiments in Federated Learning and Federated Analytics, which enable machine learning on decentralized data without centralizing user information. It matters now because data privacy regulations and edge computing demands are pushing for privacy-preserving AI techniques, though its utility is limited to research contexts and not practical for direct application or production. However, it relies on TensorFlow Federated, making it a niche resource for academic exploration. _themes: federated-learning · privacy · machine-learning · research_ #### [google-research/vision_transformer](https://github.com/google-research/vision_transformer) *Jupyter Notebook · ★12,467 · Apache-2.0 · beta · score:0.90 · hot:0.49 · rising:0.57 · durable:0.69 · board:durable · trend:stable* **Why it matters.** This repository from Google Research provides pre-trained Vision Transformer (ViT) and MLP-Mixer models for image recognition, along with code for fine-tuning them using JAX and Flax, based on influential papers. It matters now because transformers are transforming computer vision by offering scalable alternatives to traditional CNNs, and these models enable efficient experimentation amid growing interest in AI for visual tasks. _themes: vision · transformers · fine-tuning · image-recognition_ #### [allenai/satlas-super-resolution](https://github.com/allenai/satlas-super-resolution) *Python · ★337 · Apache-2.0 · experimental · score:0.70 · hot:0.48 · rising:0.51 · durable:0.57 · board:durable · trend:stable* **Why it matters.** This repository provides code, data, and models for AI-based super-resolution of satellite imagery, specifically enhancing Sentinel-2 data to higher resolutions for global geospatial applications, as detailed in a research paper. It matters now because accurate, frequently updated open geospatial data is crucial for applications like climate monitoring and urban planning, especially with the growing demand for high-quality AI-generated remote sensing data amid advancing Earth observation technologies. _themes: super-resolution · remote-sensing · image-enhancement · geospatial-ai_ #### [artidoro/qlora](https://github.com/artidoro/qlora) *Jupyter Notebook · ★10,886 · MIT · beta · score:0.85 · hot:0.48 · rising:0.55 · durable:0.67 · board:durable · trend:stable* QLoRA: Efficient Finetuning of Quantized LLMs **Why it matters.** QLoRA is a library that facilitates efficient fine-tuning of large language models by quantizing them to 4 bits and using LoRA adapters, enabling the process on a single consumer-grade GPU without significant performance loss. It integrates with existing tools like Hugging Face's PEFT and transformers, making it practical for adapting pre-trained models. This matters now because it lowers the barriers to entry for LLM research and development, especially amid hardware limitations and the rapid growth of AI applications. _themes: fine-tuning · quantization · LLMs · efficient-training_ #### [google-research/football](https://github.com/google-research/football) *Python · ★3,585 · Apache-2.0 · beta · score:0.70 · hot:0.48 · rising:0.56 · durable:0.67 · board:durable · trend:stable* Check out the new game server: **Why it matters.** This repository provides a reinforcement learning environment for simulating football games, allowing researchers to train AI agents in a complex, multi-agent setting based on an open-source game. It matters because it serves as a benchmark for RL algorithms in realistic scenarios and has facilitated competitions and research, though its relevance has waned slightly with newer RL frameworks emerging. _themes: reinforcement-learning · environments · simulation · games_ #### [meta-llama/llama](https://github.com/meta-llama/llama) *Python · ★59,348 · NOASSERTION · archived · score:0.20 · hot:0.48 · rising:0.53 · durable:0.57 · board:durable · trend:stable* Inference code for Llama models **Why it matters.** This repository originally provided inference code for Meta's Llama language models, allowing users to run and experiment with these pre-trained and fine-tuned models. However, it has been deprecated with the Llama 3.1 release, directing users to newer repositories for updated functionality, making it largely irrelevant for current development. As a result, it no longer represents a key resource for AI innovation and should be avoided in favor of the recommended alternatives. _themes: inference · llm · models · deprecated_ #### [google-deepmind/mctx](https://github.com/google-deepmind/mctx) *Python · ★2,611 · Apache-2.0 · experimental · score:0.70 · hot:0.48 · rising:0.54 · durable:0.69 · board:durable · trend:stable* Monte Carlo tree search in JAX **Why it matters.** Mctx is a JAX-native library for implementing Monte Carlo Tree Search algorithms like AlphaZero and MuZero, optimized for parallel execution on accelerators to speed up reinforcement learning tasks. It matters right now because the integration of efficient search with deep learning models is crucial for advancing AI research in games and complex decision-making, especially as hardware capabilities continue to grow and enable scalable computations. This library helps researchers leverage JAX's performance for building and experimenting with state-of-the-art RL systems. _themes: jax · mcts · reinforcement-learning · search_ #### [openai/baselines](https://github.com/openai/baselines) *Python · ★16,693 · MIT · production · score:0.75 · hot:0.48 · rising:0.57 · durable:0.65 · board:durable · trend:stable* OpenAI Baselines: high-quality implementations of reinforcement learning algorithms **Why it matters.** OpenAI Baselines provides high-quality Python implementations of reinforcement learning algorithms like DQN and its variants, serving as reliable baselines for research replication and comparison. It matters now because reinforcement learning remains foundational in AI development, offering a stable reference for experiments, though its maintenance status means it lacks recent advancements and may not address current challenges in the field. _themes: rl · reinforcement-learning · algorithms · baselines_ #### [openai/retro](https://github.com/openai/retro) *C · ★3,581 · MIT · production · score:0.70 · hot:0.48 · rising:0.56 · durable:0.70 · board:durable · trend:stable* Retro Games in Gym **Why it matters.** Gym Retro is a library that integrates classic video games into OpenAI's Gym for reinforcement learning environments, supporting over 1000 games via various emulators, which helps in training agents on diverse, retro-style tasks. It matters for RL research as it provides ready-made environments for experimentation, but its maintenance-only status means it lacks integration with modern RL advancements, potentially limiting its relevance in rapidly evolving AI landscapes. _themes: reinforcement-learning · gym · emulation · retro-games_ #### [openai/guided-diffusion](https://github.com/openai/guided-diffusion) *Python · ★7,354 · MIT · experimental · score:0.80 · hot:0.47 · rising:0.54 · durable:0.64 · board:durable · trend:stable* **Why it matters.** This repository contains code and pre-trained models for guided diffusion models, an advancement in generative AI that uses classifier conditioning to improve image synthesis and outperform GANs, as detailed in a 2021 OpenAI paper. It matters now because diffusion models remain a cornerstone of modern image generation techniques, influencing ongoing research and applications in AI despite the repo's age and lack of recent updates. _themes: diffusion · image-generation · generative-models · conditioning_ #### [deepseek-ai/Engram](https://github.com/deepseek-ai/Engram) *Python · ★4,291 · Apache-2.0 · experimental · score:0.75 · hot:0.47 · rising:0.53 · durable:0.66 · board:durable · trend:stable* Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models **Why it matters.** Engram introduces a new module for large language models that implements conditional memory through scalable lookup, enabling a novel form of sparsity to improve efficiency without sacrificing performance. This matters now as the AI field grapples with the computational demands of scaling LLMs, offering a potential way to optimize models for real-world deployment amid growing concerns over resource constraints and energy consumption. _themes: sparsity · llm · efficiency · conditional-computation_ #### [openai/consistency_models](https://github.com/openai/consistency_models) *Python · ★6,476 · MIT · experimental · score:0.80 · hot:0.47 · rising:0.53 · durable:0.65 · board:durable · trend:stable* Official repo for consistency models. **Why it matters.** This repository provides PyTorch implementations for Consistency Models, a technique that distills diffusion models for faster image generation and editing, as detailed in the associated research paper. It matters now because it addresses efficiency bottlenecks in generative AI, potentially enabling real-time applications in research and development amid growing demands for scalable AI models. _themes: diffusion · image-generation · distillation · efficiency_ #### [google-research/torchsde](https://github.com/google-research/torchsde) *Python · ★1,719 · Apache-2.0 · beta · score:0.70 · hot:0.47 · rising:0.53 · durable:0.67 · board:durable · trend:stable* Differentiable SDE solvers with GPU support and efficient sensitivity analysis. **Why it matters.** This repository provides a PyTorch-based library for solving stochastic differential equations (SDEs) with GPU acceleration and efficient sensitivity analysis, enabling seamless integration into deep learning workflows. It matters now as SDEs are increasingly vital for modeling uncertainty in fields like finance and generative AI, such as diffusion models, and its differentiable features support advanced research in neural differential equations. However, its niche focus means it's most relevant for specialized applications rather than broad adoption. _themes: sdes · pytorch · deep-learning · differentiable-programming_ #### [microsoft/Graphormer](https://github.com/microsoft/Graphormer) *Python · ★2,442 · MIT · beta · score:0.75 · hot:0.47 · rising:0.53 · durable:0.66 · board:durable · trend:stable* Graphormer is a general-purpose deep learning backbone for molecular modeling. **Why it matters.** Graphormer is a deep learning library that serves as a backbone for molecular modeling using graph transformers, allowing users to train models for tasks like drug and material discovery. It matters now because it has demonstrated competitive performance in benchmarks such as the Open Catalyst Challenge and provides pre-trained models, potentially accelerating AI-driven scientific research amid growing interest in AI for science. _themes: graph-neural-networks · transformers · molecular-modeling · deep-learning_ #### [google-deepmind/android_env](https://github.com/google-deepmind/android_env) *Python · ★1,208 · Apache-2.0 · beta · score:0.80 · hot:0.47 · rising:0.54 · durable:0.69 · board:durable · trend:down* RL research on Android devices. **Why it matters.** AndroidEnv is a Python library that turns an Android device into a reinforcement learning environment, allowing agents to interact via touchscreen events and learn custom tasks on real apps. It matters now because it bridges RL research with everyday mobile interactions, potentially advancing AI applications in practical settings like automation or accessibility, though its impact is limited by the need for specialized setup and simulation. _themes: reinforcement-learning · android · agents · simulation_ #### [NVIDIA/FastPhotoStyle](https://github.com/NVIDIA/FastPhotoStyle) *Python · ★11,182 · NOASSERTION · archived · score:0.60 · hot:0.47 · rising:0.51 · durable:0.57 · board:durable · trend:stable* Style transfer, deep learning, feature transform **Why it matters.** This repository implements a fast photorealistic image style transfer algorithm using deep learning, allowing users to apply the style of one photo to another while maintaining realism. It matters as a foundational example from NVIDIA's research, cited in computer vision papers, but its age and lack of updates since 2018 make it less relevant for current applications compared to modern alternatives. _themes: style-transfer · deep-learning · computer-vision · pytorch_ #### [google-deepmind/graph_nets](https://github.com/google-deepmind/graph_nets) *Python · ★5,397 · Apache-2.0 · beta · score:0.70 · hot:0.47 · rising:0.55 · durable:0.66 · board:durable · trend:stable* Build Graph Nets in Tensorflow **Why it matters.** This repository provides DeepMind's library for building graph neural networks using TensorFlow and Sonnet, allowing users to process and update graph-structured data with edge, node, and global attributes. It matters because graph networks are essential for applications involving relational data, such as molecular modeling or social network analysis, but its lack of recent releases suggests it may not be actively maintained, potentially limiting its relevance compared to newer alternatives. _themes: graphs · neural-networks · deep-learning · tensorflow_ #### [facebookresearch/sapiens](https://github.com/facebookresearch/sapiens) *Python · ★5,327 · NOASSERTION · beta · score:0.75 · hot:0.47 · rising:0.53 · durable:0.65 · board:durable · trend:stable* High-resolution models for human tasks. **Why it matters.** Sapiens provides a suite of high-resolution computer vision models specialized in human-centric tasks like 2D pose estimation, segmentation, depth, and normals, trained on 300 million in-the-wild images for strong generalization in real-world scenarios. It matters now because the demand for accurate human analysis is growing in applications such as robotics, AR/VR, and social media, and its focus on high-resolution features addresses limitations in existing models, though the lack of a formal release and unclear licensing may hinder immediate adoption. _themes: computer-vision · human-centric · inference · high-resolution_ #### [google-research/text-to-text-transfer-transformer](https://github.com/google-research/text-to-text-transfer-transformer) *Python · ★6,507 · Apache-2.0 · archived · score:0.40 · hot:0.47 · rising:0.52 · durable:0.60 · board:durable · trend:stable* Code for the paper "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer" **Why it matters.** This repository provides the original TensorFlow implementation for the T5 model, enabling users to reproduce experiments from the paper on transfer learning with text-to-text transformers, including dataset handling and fine-tuning. However, it's no longer actively maintained, with Google recommending the T5X repository for new work, making it more of a historical reference than a current tool for active development. It's still useful for educational purposes or legacy projects involving the original T5 setup. _themes: transfer-learning · nlp · fine-tuning · transformers_ #### [facebookresearch/fairscale](https://github.com/facebookresearch/fairscale) *Python · ★3,406 · NOASSERTION · production · score:0.80 · hot:0.47 · rising:0.54 · durable:0.66 · board:durable · trend:stable* PyTorch extensions for high performance and large scale training. **Why it matters.** FairScale extends PyTorch with tools for high-performance and large-scale model training, offering composable APIs for distributed techniques like FullyShardedDataParallel. It matters now because efficient scaling is essential for handling increasingly large AI models amid resource constraints, and its features have been integrated into PyTorch core, making it relevant for ongoing research and development in deep learning. _themes: distributed-training · pytorch · scaling · large-models_ #### [facebookresearch/flow_matching](https://github.com/facebookresearch/flow_matching) *Python · ★4,347 · NOASSERTION · experimental · score:0.75 · hot:0.47 · rising:0.51 · durable:0.63 · board:durable · trend:stable* A PyTorch library for implementing flow matching algorithms, featuring continuous and discrete flow matching implementations. It includes practical examples for both text and image modalities. **Why it matters.** This repository offers a PyTorch library for implementing flow matching algorithms, including continuous and discrete variants, with practical examples for text and image generation. It matters now as flow matching represents an evolving technique in generative AI research, potentially improving efficiency in model training amid the rapid advancements in diffusion-based systems, as highlighted in its associated ArXiv paper. _themes: generative · diffusion · pytorch · flows_ #### [yingpengma/Awesome-Story-Generation](https://github.com/yingpengma/Awesome-Story-Generation) *Python · ★610 · no-license · beta · score:0.70 · hot:0.47 · rising:0.48 · durable:0.56 · board:durable · trend:down* This repository collects an extensive list of awesome papers about Story Generation / Storytelling, exclusively focusing on the era of Large Language Models (LLMs). **Why it matters.** This repository curates a list of research papers focused on story generation and storytelling using large language models, organized into categories like planning, multimodality, and evaluation, serving as a reference for tracking recent advancements. It matters now because LLMs are driving rapid innovation in narrative AI, but its value is limited to passive information sharing without tools for interaction or practical application, making it a niche resource for academic tracking rather than direct use. _themes: llm · story-generation · nlp · research_ #### [google-research/batch-ppo](https://github.com/google-research/batch-ppo) *Python · ★976 · Apache-2.0 · beta · score:0.70 · hot:0.46 · rising:0.53 · durable:0.68 · board:durable · trend:down* Efficient Batched Reinforcement Learning in TensorFlow **Why it matters.** This repository provides an optimized implementation of Proximal Policy Optimization (PPO) for reinforcement learning in TensorFlow, enabling batched computation across multiple parallel environments for better efficiency. It matters because efficient RL training is crucial for scaling experiments in AI research, though its reliance on older TensorFlow versions (1.3+) may limit its immediate applicability in modern workflows. _themes: reinforcement-learning · tensorflow · batched-computation · optimization_ #### [openai/weak-to-strong](https://github.com/openai/weak-to-strong) *Python · ★2,552 · MIT · experimental · score:0.75 · hot:0.46 · rising:0.52 · durable:0.64 · board:durable · trend:stable* **Why it matters.** This repository implements code for weak-to-strong generalization experiments, allowing fine-tuning of language models and training against labels from stronger models, as outlined in an OpenAI paper, with extensions to vision models. It matters right now for AI research on alignment and generalization, but its untested nature and deviations from the paper mean results are not reliable for production. While timely for addressing AI safety concerns, it requires careful validation due to potential inconsistencies. _themes: fine-tuning · generalization · ai-alignment · vision_ #### [facebookresearch/habitat-lab](https://github.com/facebookresearch/habitat-lab) *Python · ★2,951 · MIT · beta · score:0.70 · hot:0.46 · rising:0.53 · durable:0.65 · board:durable · trend:stable* A modular high-level library to train embodied AI agents across a variety of tasks and environments. **Why it matters.** Habitat-Lab is a Python library for training and evaluating embodied AI agents in simulated indoor environments, supporting tasks like navigation, rearrangement, and human interaction through flexible configurations and integration with Habitat-Sim. It matters now because embodied AI is increasingly relevant for robotics research and real-world applications, such as sim2real transfers, amid growing interest in autonomous agents that can interact with physical spaces, though its focus on research limits broader adoption. _themes: agents · reinforcement-learning · simulation · robotics_ #### [deepseek-ai/DualPipe](https://github.com/deepseek-ai/DualPipe) *Python · ★2,943 · MIT · experimental · score:0.75 · hot:0.46 · rising:0.52 · durable:0.65 · board:durable · trend:stable* A bidirectional pipeline parallelism algorithm for computation-communication overlap in DeepSeek V3/R1 training. **Why it matters.** DualPipe is a bidirectional pipeline parallelism algorithm designed to fully overlap computation and communication phases during the training of large models like DeepSeek V3, thereby reducing pipeline bubbles and improving efficiency. It matters now because distributed training for massive AI models is increasingly resource-constrained, and optimizations like this can significantly cut costs and time, especially as seen in recent large-scale training efforts. _themes: distributed-training · pipeline-parallelism · optimization · efficiency_ #### [huggingface/awesome-papers](https://github.com/huggingface/awesome-papers) *? · ★2,053 · no-license · experimental · score:0.65 · hot:0.46 · rising:0.51 · durable:0.59 · board:durable · trend:stable* Papers & presentation materials from Hugging Face's internal science day **Why it matters.** This repository curates and shares presentation materials and summaries from Hugging Face's internal discussions on key NLP papers, providing community access to expert insights without full interactive elements. It matters for researchers seeking curated research overviews in a fast-evolving AI field, though its value is limited by the lack of recent updates since 2020 and absence of a formal license, potentially reducing its long-term utility. _themes: nlp · papers · research · community_ #### [openai/grok](https://github.com/openai/grok) *Python · ★4,245 · MIT · experimental · score:0.60 · hot:0.46 · rising:0.52 · durable:0.60 · board:durable · trend:stable* **Why it matters.** This repository contains code for experiments on the 'grokking' phenomenon, where neural networks generalize beyond overfitting on small algorithmic datasets, based on an OpenAI research paper. It matters now because insights into generalization can improve AI model training efficiency, especially as the field grapples with data scarcity and scaling challenges in modern machine learning research. _themes: generalization · overfitting · training · experiments_ #### [google-research/disentanglement_lib](https://github.com/google-research/disentanglement_lib) *Python · ★1,424 · Apache-2.0 · beta · score:0.70 · hot:0.46 · rising:0.53 · durable:0.66 · board:durable · trend:stable* disentanglement_lib is an open-source library for research on learning disentangled representations. **Why it matters.** Disentanglement_lib is an open-source Python library that facilitates research on learning disentangled representations by providing models like BetaVAE, metrics for evaluation, and datasets for experimentation, along with pretrained models. It matters for researchers in representation learning due to its foundation in a 2019 ICML best paper, but its utility is limited today by the age of the codebase and the rapid evolution of AI techniques, making it more of a reference than a go-to tool. _themes: disentanglement · vae · representation-learning · metrics_ #### [allenai/OLMo](https://github.com/allenai/OLMo) *Python · ★6,482 · Apache-2.0 · archived · score:0.20 · hot:0.46 · rising:0.50 · durable:0.59 · board:durable · trend:stable* Modeling, training, eval, and inference code for OLMo **Why it matters.** This repository provides code for modeling, training, evaluating, and inferring AI2's OLMo open language models, including a two-stage pretraining procedure on specific datasets. However, it is outdated and no longer active, with users advised to switch to the OLMo-core repository for current developments, making it less relevant except for historical or legacy use. It matters primarily for researchers interested in past implementations of open language models. _themes: language-models · training · inference · open-source_ #### [facebookresearch/dlrm](https://github.com/facebookresearch/dlrm) *Python · ★4,032 · MIT · beta · score:0.70 · hot:0.46 · rising:0.53 · durable:0.63 · board:durable · trend:stable* An implementation of a deep learning recommendation model (DLRM) **Why it matters.** DLRM is an open-source implementation of a deep learning recommendation model that processes dense and sparse features for applications like personalized advertising and content recommendation. It matters for researchers and practitioners in recommendation systems due to its focus on efficient handling of large-scale sparse data, but its relevance is diminished by the lack of recent updates, potentially making it less competitive against more modern alternatives. _themes: recommendation · embeddings · deep-learning · mlp_ #### [huggingface/parler-tts](https://github.com/huggingface/parler-tts) *Python · ★5,565 · Apache-2.0 · beta · score:0.80 · hot:0.46 · rising:0.53 · durable:0.64 · board:durable · trend:stable* Inference and training library for high-quality TTS models. **Why it matters.** Parler-TTS is an open-source library for inference and training of high-quality text-to-speech models that generate natural-sounding speech mimicking specific speaker styles, based on a research paper and trained on extensive audiobook data. It matters now because it provides fully accessible TTS tools with recent optimizations for faster generation, enabling developers to build voice applications amid growing demand for AI-driven speech tech, while promoting community-driven improvements through open datasets and code. _themes: tts · inference · fine-tuning · voice-cloning_ #### [EleutherAI/pythia](https://github.com/EleutherAI/pythia) *Jupyter Notebook · ★2,779 · Apache-2.0 · beta · score:0.80 · hot:0.46 · rising:0.52 · durable:0.65 · board:durable · trend:stable* The hub for EleutherAI's work on interpretability and learning dynamics **Why it matters.** Pythia is a suite of large language models and checkpoints from EleutherAI designed for analyzing interpretability, learning dynamics, and scaling laws in transformers during training. It matters right now because it addresses the critical need for transparency and reproducibility in AI research amid growing concerns over LLM ethics and safety, while providing a unique resource with 154 checkpoints per model that few alternatives match. _themes: interpretability · llm-training · scaling-laws · reproducibility_ #### [google-deepmind/penzai](https://github.com/google-deepmind/penzai) *Python · ★1,880 · Apache-2.0 · beta · score:0.80 · hot:0.46 · rising:0.50 · durable:0.72 · board:durable · trend:down* A JAX research toolkit for building, editing, and visualizing neural networks. **Why it matters.** Penzai is a JAX library that allows researchers to construct, modify, and visualize neural networks as functional pytree structures, with a focus on post-training analysis like inspecting activations and performing model surgery. It matters now because the AI field is emphasizing model interpretability and safety amid growing concerns over black-box systems, providing tools that enable deeper experimentation without proprietary dependencies. This is particularly timely as JAX gains popularity for efficient, research-oriented deep learning workflows. _themes: jax · interpretability · visualization · fine-tuning_ #### [NVIDIA/pix2pixHD](https://github.com/NVIDIA/pix2pixHD) *Python · ★6,943 · NOASSERTION · archived · score:0.60 · hot:0.46 · rising:0.47 · durable:0.51 · board:durable · trend:stable* Synthesizing and manipulating 2048x1024 images with conditional GANs **Why it matters.** This repository offers a PyTorch implementation of pix2pixHD, a conditional GAN designed for high-resolution image-to-image translation, such as converting semantic label maps into photorealistic images up to 2048x1024. It matters because it serves as a foundational example in generative AI for computer vision tasks, but its age (from 2018) means it's largely superseded by more advanced models, making it more relevant for educational or research baseline purposes rather than current production use. _themes: gan · image-generation · computer-vision · pytorch_ #### [google-deepmind/ferminet](https://github.com/google-deepmind/ferminet) *Python · ★828 · Apache-2.0 · experimental · score:0.70 · hot:0.46 · rising:0.50 · durable:0.59 · board:durable · trend:stable* An implementation of the Fermionic Neural Network for ab-initio electronic structure calculations **Why it matters.** FermiNet implements a neural network to approximate ground state wavefunctions for atoms and molecules using variational Monte Carlo, enabling more accurate ab-initio electronic structure calculations without traditional approximations. It matters now because AI-driven quantum simulations are accelerating research in chemistry and materials science, potentially impacting drug discovery and new material design, especially as computational resources and AI frameworks like JAX continue to evolve. However, as a research-level tool under active development, it's not yet ready for widespread production use. _themes: quantum-chemistry · neural-networks · monte-carlo · jax_ #### [facebookresearch/lingua](https://github.com/facebookresearch/lingua) *Python · ★4,759 · BSD-3-Clause · experimental · score:0.60 · hot:0.45 · rising:0.51 · durable:0.63 · board:durable · trend:stable* Meta Lingua: a lean, efficient, and easy-to-hack codebase to research LLMs. **Why it matters.** Meta Lingua is a minimal PyTorch-based library for training, inferring, and evaluating large language models, focused on research by providing easy-to-modify components for experimenting with architectures, losses, and data. It matters now as the LLM field demands rapid prototyping tools to explore new ideas amid growing model complexity, but its under-development status and lack of releases mean it's not yet reliable for broader use. _themes: llm · training · inference · research_ #### [facebookresearch/fvcore](https://github.com/facebookresearch/fvcore) *Python · ★2,235 · Apache-2.0 · production · score:0.70 · hot:0.45 · rising:0.51 · durable:0.60 · board:durable · trend:stable* Collection of common code that's shared among different research projects in FAIR computer vision team. **Why it matters.** fvcore is a lightweight library providing essential utilities for computer vision tasks in PyTorch, including layers, flop counting, and parameter management, primarily shared across Facebook AI Research projects. It matters for researchers working on CV models as it offers tested and benchmarked tools that can streamline development, though its scope is limited to FAIR's ecosystem and may not address broader needs in the field. _themes: pytorch · computer-vision · flop-counting · utilities_ #### [allenai/scispacy](https://github.com/allenai/scispacy) *Python · ★1,944 · Apache-2.0 · beta · score:0.80 · hot:0.45 · rising:0.52 · durable:0.65 · board:durable · trend:stable* A full spaCy pipeline and models for scientific/biomedical documents. **Why it matters.** SciSpacy is a library that extends the spaCy NLP framework with custom pipelines and models specifically trained for scientific and biomedical documents, including tokenizers, POS taggers, parsers, and NER models. It matters now because the rapid growth of scientific literature demands specialized NLP tools for accurate text processing in fields like bioinformatics and healthcare, aiding tasks such as entity recognition and information extraction amid AI-driven research advancements. _themes: nlp · biomedical · inference · fine-tuning_ #### [facebookresearch/encodec](https://github.com/facebookresearch/encodec) *Python · ★3,941 · MIT · experimental · score:0.70 · hot:0.45 · rising:0.50 · durable:0.61 · board:durable · trend:stable* State-of-the-art deep learning based audio codec supporting both mono 24 kHz audio and stereo 48 kHz audio. **Why it matters.** EnCodec is a deep learning-based audio codec that compresses high-fidelity mono 24 kHz and stereo 48 kHz audio at various bitrates, offering better quality than traditional codecs for applications like music streaming and voice communication. It matters now because efficient audio compression is essential for bandwidth-constrained environments, such as mobile apps and online platforms, and this neural approach could improve upon existing methods, though its lack of official releases limits immediate adoption. However, the provided models and tools could accelerate research in audio processing. _themes: audio-compression · neural-codec · deep-learning_ #### [google-research/sam](https://github.com/google-research/sam) *Python · ★627 · Apache-2.0 · experimental · score:0.70 · hot:0.45 · rising:0.47 · durable:0.52 · board:durable · trend:stable* **Why it matters.** This repository implements Sharpness-Aware Minimization (SAM), an optimization technique that improves neural network generalization by simultaneously minimizing loss value and sharpness, leading to better performance on benchmarks like CIFAR and ImageNet. It matters now because overfitting remains a key challenge in training large, overparameterized models, and SAM provides a simple yet effective way to enhance robustness to label noise without additional complexity. Empirical results show it achieves state-of-the-art error rate reductions across various datasets and models. _themes: optimization · generalization · deep-learning_ #### [openai/improved-diffusion](https://github.com/openai/improved-diffusion) *Python · ★3,822 · MIT · experimental · score:0.75 · hot:0.45 · rising:0.50 · durable:0.58 · board:durable · trend:stable* Release for Improved Denoising Diffusion Probabilistic Models **Why it matters.** This repository provides an implementation of improved denoising diffusion probabilistic models for tasks like image generation, based on a specific research paper from OpenAI. It matters because diffusion models are a core technique in current generative AI research, offering better stability and quality in synthetic data creation, though this codebase lacks active maintenance as indicated by no official releases. _themes: diffusion · generative-models · image-generation · pytorch_ #### [huggingface/gsplat.js](https://github.com/huggingface/gsplat.js) *TypeScript · ★1,615 · MIT · beta · score:0.65 · hot:0.45 · rising:0.51 · durable:0.65 · board:durable · trend:down* JavaScript Gaussian Splatting library. **Why it matters.** gsplat.js is a JavaScript library for 3D Gaussian Splatting, offering tools to render and manipulate point-based 3D scenes, similar to three.js but specialized for this technique. It matters now due to the rising interest in efficient real-time 3D rendering for applications like AR/VR and AI-generated content, though its utility is limited by hardware dependencies and niche adoption outside specific graphics communities. _themes: 3d-rendering · gaussian-splat · webgl · typescript_ #### [deepseek-ai/EPLB](https://github.com/deepseek-ai/EPLB) *Python · ★1,360 · MIT · beta · score:0.65 · hot:0.45 · rising:0.51 · durable:0.63 · board:durable · trend:down* Expert Parallelism Load Balancer **Why it matters.** EPLB is a Python library that implements a load balancing algorithm for expert parallelism in AI models, specifically duplicating and strategically placing experts on GPUs to balance loads and reduce inter-node traffic. It matters now because distributed training for large language models is increasingly common, and efficient resource management is critical for performance in mixture-of-experts architectures, as highlighted in the DeepSeek-V3 paper. _themes: parallelism · load-balancing · distributed-training · mixture-of-experts_ #### [mistralai/mistral-finetune](https://github.com/mistralai/mistral-finetune) *Python · ★3,088 · Apache-2.0 · beta · score:0.70 · hot:0.45 · rising:0.51 · durable:0.63 · board:durable · trend:stable* **Why it matters.** Mistral-finetune is a lightweight Python library for memory-efficient finetuning of Mistral AI models using LoRA, which freezes most weights and trains only a small percentage, making it suitable for high-end GPUs like A100 or H100. It matters now because efficient finetuning is increasingly important for customizing large language models in resource-constrained environments, and recent updates like compatibility with Mistral Large v2 address growing demands for handling larger models amid rapid AI advancements. _themes: fine-tuning · LoRA · efficiency · model-adaptation_ #### [NVIDIA/libglvnd](https://github.com/NVIDIA/libglvnd) *C · ★530 · no-license · production · score:0.50 · hot:0.45 · rising:0.51 · durable:0.62 · board:durable · trend:down* The GL Vendor-Neutral Dispatch library **Why it matters.** Libglvnd is a C library that handles OpenGL API calls by dispatching them to the appropriate vendor at runtime, enabling multiple graphics drivers to coexist on the same system without conflicts. This matters for graphics-intensive applications on Linux, where hardware diversity is common, but its relevance is limited as modern graphics ecosystems increasingly standardize or move towards Vulkan. However, it remains a practical solution for legacy OpenGL support in mixed environments. _themes: opengl · graphics · dispatch · cross-vendor_ #### [google-research/self-organising-systems](https://github.com/google-research/self-organising-systems) *Jupyter Notebook · ★406 · Apache-2.0 · experimental · score:0.60 · hot:0.45 · rising:0.50 · durable:0.62 · board:durable · trend:down* **Why it matters.** This repository contains code for self-organising systems, including implementations like growing neural cellular automata, which explore how simple rules can generate complex, emergent behaviors in AI models. It matters now because it aligns with ongoing research in efficient, adaptive AI systems, potentially applicable to generative tasks and simulations, though its impact is limited by the niche focus and lack of widespread adoption beyond academic circles. _themes: cellular-automata · neural-networks · generative-models · research_ #### [facebookresearch/MobileLLM](https://github.com/facebookresearch/MobileLLM) *Python · ★1,424 · NOASSERTION · beta · score:0.75 · hot:0.45 · rising:0.50 · durable:0.62 · board:durable · trend:down* MobileLLM Optimizing Sub-billion Parameter Language Models for On-Device Use Cases. In ICML 2024. **Why it matters.** MobileLLM provides training code and optimized sub-billion parameter language models designed for on-device deployment, incorporating techniques like SwiGLU and grouped-query attention to improve accuracy on commonsense tasks. It matters now because the growing need for efficient, privacy-preserving AI on mobile devices is driving innovation in edge computing, and this work demonstrates scalable improvements that could enable real-world applications in resource-constrained environments. Recent updates show it extends to larger models, making it relevant amid the push for on-device LLMs in mobile apps and IoT. _themes: inference · optimization · language-models · mobile_ #### [google-deepmind/dnc](https://github.com/google-deepmind/dnc) *Python · ★2,538 · Apache-2.0 · archived · score:0.50 · hot:0.45 · rising:0.49 · durable:0.55 · board:durable · trend:stable* A TensorFlow implementation of the Differentiable Neural Computer. **Why it matters.** This repository offers a TensorFlow implementation of the Differentiable Neural Computer, a recurrent neural network with external memory for handling tasks that require dynamic memory access and reasoning. It was influential in early research on memory-augmented networks but is from 2016 and lacks recent updates, making it less relevant today compared to modern architectures like transformers, though it remains a foundational reference for neural memory mechanisms. _themes: rnn · memory-networks · neural-architecture · tensorflow_ #### [facebookresearch/hiera](https://github.com/facebookresearch/hiera) *Python · ★1,060 · Apache-2.0 · beta · score:0.75 · hot:0.44 · rising:0.50 · durable:0.65 · board:durable · trend:down* Hiera: A fast, powerful, and simple hierarchical vision transformer. **Why it matters.** Hiera is a hierarchical vision transformer implementation that simplifies the architecture for computer vision tasks, achieving competitive performance without unnecessary complexities as detailed in its ICML 2023 paper. It matters now because the ongoing evolution of vision models demands efficient alternatives amid growing computational constraints, potentially offering a baseline for further research and deployment in resource-limited scenarios. _themes: vision-transformer · hierarchical · efficiency · computer-vision_ #### [deepseek-ai/ESFT](https://github.com/deepseek-ai/ESFT) *Python · ★733 · MIT · experimental · score:0.70 · hot:0.44 · rising:0.49 · durable:0.60 · board:durable · trend:down* Expert Specialized Fine-Tuning **Why it matters.** ESFT is a fine-tuning technique for Mixture-of-Experts (MoE) based Large Language Models that selectively updates only task-relevant experts, reducing computational resources and improving efficiency. It matters now because the demand for resource-efficient AI customization is growing amid scaling challenges in LLMs, and its recent acceptance to EMNLP 2024 underscores its timely relevance for advancing sparse model research. _themes: fine-tuning · MoE · efficiency_ #### [google-research/fixmatch](https://github.com/google-research/fixmatch) *Python · ★1,209 · Apache-2.0 · beta · score:0.60 · hot:0.44 · rising:0.51 · durable:0.64 · board:durable · trend:down* A simple method to perform semi-supervised learning with limited data. **Why it matters.** FixMatch is a semi-supervised learning algorithm that uses consistency regularization and confident pseudo-labels to train models with limited labeled data, making it efficient for scenarios where labeling is costly. It matters now because semi-supervised techniques are increasingly relevant in AI development to leverage unlabeled data, though this implementation is based on outdated TensorFlow 1, limiting its immediate applicability. Researchers can still draw insights from it for modern adaptations. _themes: semi-supervised · learning · regularization · tensorflow_ #### [google-deepmind/trfl](https://github.com/google-deepmind/trfl) *Python · ★3,134 · Apache-2.0 · archived · score:0.60 · hot:0.44 · rising:0.49 · durable:0.58 · board:durable · trend:stable* TensorFlow Reinforcement Learning **Why it matters.** TRFL is a library built on TensorFlow that provides building blocks like loss functions and utilities for implementing reinforcement learning agents, making it easier to develop RL algorithms. It matters now for legacy TensorFlow-based RL projects, but its relevance is diminished due to lack of recent updates and the availability of more modern alternatives in the evolving RL landscape. _themes: rl · tensorflow · agents · deep-learning_ #### [huggingface/pytorch-openai-transformer-lm](https://github.com/huggingface/pytorch-openai-transformer-lm) *Python · ★1,521 · MIT · archived · score:0.40 · hot:0.44 · rising:0.44 · durable:0.52 · board:durable · trend:stable* 🐥A PyTorch implementation of OpenAI's finetuned transformer language model with a script to import the weights pre-trained by OpenAI **Why it matters.** This repository offers a PyTorch implementation of OpenAI's early transformer language model, including a script to import pre-trained weights from their TensorFlow version, which was useful for researchers experimenting with language models in 2018. It matters today primarily as a historical artifact, showing early efforts in model porting, but its relevance is diminished by more advanced libraries like Hugging Face's Transformers, making it less essential for modern NLP workflows. _themes: transformer · pytorch · language-model · fine-tuning_ #### [huggingface/transfer-learning-conv-ai](https://github.com/huggingface/transfer-learning-conv-ai) *Python · ★1,759 · MIT · beta · score:0.60 · hot:0.44 · rising:0.49 · durable:0.56 · board:durable · trend:stable* 🦄 State-of-the-Art Conversational AI with Transfer Learning **Why it matters.** This repository provides code for fine-tuning GPT and GPT-2 models on conversational AI tasks using transfer learning, enabling users to build and train dialog agents with relatively simple scripts. It matters because it offers an accessible way to reproduce 2018 state-of-the-art results in dialog systems, serving as an educational resource for understanding early transformer-based transfer learning in NLP, though it's outdated compared to current advancements like those in Hugging Face's Transformers library. _themes: dialog · transfer-learning · gpt · nlp_ #### [google-deepmind/dm_pix](https://github.com/google-deepmind/dm_pix) *Python · ★435 · Apache-2.0 · beta · score:0.70 · hot:0.44 · rising:0.50 · durable:0.66 · board:durable · trend:down* PIX is an image processing library in JAX, for JAX. **Why it matters.** PIX is a library that provides image processing functions optimized for JAX, allowing for efficient parallelization and acceleration on GPUs or TPUs, which is essential for computer vision tasks in machine learning. It matters right now because JAX is gaining traction in research for its performance benefits, and PIX fills a gap by integrating image handling directly into this ecosystem, potentially streamlining workflows for developers working on large-scale ML models without needing separate libraries. _themes: jax · image-processing · computer-vision · machine-learning_ #### [huggingface/knockknock](https://github.com/huggingface/knockknock) *Python · ★2,827 · MIT · beta · score:0.70 · hot:0.44 · rising:0.50 · durable:0.61 · board:durable · trend:stable* 🚪✊Knock Knock: Get notified when your training ends with only two additional lines of code **Why it matters.** This repository provides a simple Python library that sends notifications via platforms like email, Slack, or Telegram when machine learning training jobs complete or fail, requiring only a decorator to integrate. It matters now because long-running ML experiments are common, and this tool helps developers and researchers monitor jobs efficiently without constant oversight, potentially saving time in iterative development cycles. _themes: ml-training · notifications · deep-learning · automation_ #### [google-research/electra](https://github.com/google-research/electra) *Python · ★2,372 · Apache-2.0 · production · score:0.80 · hot:0.44 · rising:0.51 · durable:0.62 · board:durable · trend:stable* ELECTRA: Pre-training Text Encoders as Discriminators Rather Than Generators **Why it matters.** ELECTRA is a pre-training method for NLP models that trains transformers as discriminators to identify fake tokens, offering a more efficient alternative to generator-based approaches like BERT. This efficiency allows for strong performance on tasks such as question answering with less computational resources, making it relevant for scenarios where hardware is limited. It remains foundational in NLP research, though newer methods have built upon it. _themes: nlp · pre-training · transformers · fine-tuning_ #### [google-deepmind/synthid-text](https://github.com/google-deepmind/synthid-text) *Python · ★836 · Apache-2.0 · experimental · score:0.75 · hot:0.44 · rising:0.48 · durable:0.58 · board:durable · trend:down* **Why it matters.** This repository provides a reference implementation for watermarking and detecting text generated by AI models like Gemma and GPT-2, based on a research paper published in Nature, but it's not suitable for production. It matters now because watermarking addresses critical issues of content authenticity and AI ethics in an era of increasing misinformation and regulatory scrutiny, though its experimental nature limits immediate applicability. _themes: watermarking · text-generation · detection · ai-ethics_ #### [openai/mujoco-py](https://github.com/openai/mujoco-py) *Cython · ★3,130 · NOASSERTION · archived · score:0.20 · hot:0.43 · rising:0.46 · durable:0.49 · board:durable · trend:stable* MuJoCo is a physics engine for detailed, efficient rigid body simulations with contacts. mujoco-py allows using MuJoCo from Python 3. **Why it matters.** MuJoCo-py is a deprecated Python binding for the MuJoCo physics engine, enabling rigid body simulations for applications like robotics and reinforcement learning. It matters primarily for legacy projects using MuJoCo versions up to 2.1.0, but new users should adopt the official bindings to avoid compatibility issues and ensure ongoing support. _themes: physics · simulation · robotics · rl_ #### [google-deepmind/learning-to-learn](https://github.com/google-deepmind/learning-to-learn) *Python · ★4,070 · Apache-2.0 · experimental · score:0.60 · hot:0.43 · rising:0.50 · durable:0.60 · board:durable · trend:stable* Learning to Learn in TensorFlow **Why it matters.** This repository implements meta-learning techniques for optimizers in TensorFlow, allowing models to learn how to optimize other models more effectively, as demonstrated through problems like MNIST. It matters now because meta-learning addresses the need for more efficient training in complex AI systems, though its age and lack of recent updates mean it may not fully align with current TensorFlow versions or best practices in modern research. _themes: meta-learning · optimizers · deep-learning · tensorflow_ #### [google-research/kubric](https://github.com/google-research/kubric) *Jupyter Notebook · ★2,706 · Apache-2.0 · beta · score:0.80 · hot:0.43 · rising:0.49 · durable:0.60 · board:durable · trend:stable* A data generation pipeline for creating semi-realistic synthetic multi-object videos with rich annotations such as instance segmentation masks, depth maps, and optical flow. **Why it matters.** Kubric is an open-source pipeline that generates semi-realistic synthetic videos with annotations like segmentation masks and depth maps, primarily for training machine learning models in video understanding. It matters now because it addresses the shortage of annotated real-world data for unsupervised learning, enabling researchers to create customizable datasets to accelerate progress in computer vision, especially as demand for robust multi-object datasets grows. _themes: synthetic-data · data-generation · computer-vision · simulation_ #### [allenai/RL4LMs](https://github.com/allenai/RL4LMs) *Python · ★2,387 · Apache-2.0 · beta · score:0.80 · hot:0.43 · rising:0.49 · durable:0.61 · board:durable · trend:down* A modular RL library to fine-tune language models to human preferences **Why it matters.** RL4LMs is a modular library that provides tools for fine-tuning language models using reinforcement learning techniques to align with human preferences, including customizable components like algorithms, reward functions, and metrics for various NLP tasks. It matters now because the growing deployment of large language models in real-world applications demands better alignment with human values for safety and effectiveness, and this library offers a well-tested framework to achieve that amid ongoing advancements in AI ethics and model optimization. _themes: rl · fine-tuning · nlp · language-models_ #### [microsoft/DirectML](https://github.com/microsoft/DirectML) *C++ · ★2,552 · MIT · archived · score:0.40 · hot:0.43 · rising:0.47 · durable:0.53 · board:durable · trend:stable* ⚠️DirectML is in maintenance mode ⚠️ DirectML is a high-performance, hardware-accelerated DirectX 12 library for machine learning. DirectML provides GPU acceleration for common machine learning tasks across a broad range of supported hardware and drivers, including all DirectX 12-capable GPUs from vendors such as AMD, Intel, NVIDIA, and Qualcomm. **Why it matters.** DirectML is a DirectX 12 library that provides GPU acceleration for machine learning tasks, supporting a wide range of hardware from AMD, Intel, NVIDIA, and Qualcomm, making it suitable for high-performance, real-time applications. However, it's now in maintenance mode with no new features planned, so it primarily serves existing projects on older Windows versions rather than new development, where Microsoft recommends switching to Windows ML for better support. This positions it as a stable but diminishing option in the evolving ML acceleration landscape. _themes: gpu-accel · ml-acceleration · directx · inference_ #### [openai/blocksparse](https://github.com/openai/blocksparse) *Cuda · ★1,064 · MIT · beta · score:0.70 · hot:0.43 · rising:0.49 · durable:0.55 · board:durable · trend:stable* Efficient GPU kernels for block-sparse matrix multiplication and convolution **Why it matters.** This repository provides efficient GPU kernels for block-sparse matrix multiplication and convolution, optimized for TensorFlow to accelerate sparse computations in deep learning models. It matters now because the increasing scale of AI workloads demands better hardware efficiency to reduce costs and enable faster training and inference on resource-constrained systems. However, its active development status with potential breaking changes means it's best suited for users who can handle instability. _themes: sparse · gpu-kernels · tensorflow · efficiency_ #### [huggingface/local-gemma](https://github.com/huggingface/local-gemma) *Python · ★384 · Apache-2.0 · beta · score:0.70 · hot:0.43 · rising:0.49 · durable:0.64 · board:durable · trend:down* Gemma 2 optimized for your local machine. **Why it matters.** This repository provides an optimized way to run Gemma-2 models locally on your machine using CLI or Python, focusing on speed and reduced memory usage through integrations with Transformers and bitsandbytes. It matters now because it addresses the growing demand for efficient, privacy-focused local AI inference amid rising computational costs and data privacy concerns, making advanced models more accessible without relying on cloud services. However, its early stage and modest adoption (384 stars) suggest it's still evolving and not yet a mainstream solution. _themes: inference · optimization · local-execution · llm_ #### [allenai/science-parse](https://github.com/allenai/science-parse) *Java · ★699 · Apache-2.0 · production · score:0.60 · hot:0.43 · rising:0.50 · durable:0.60 · board:durable · trend:down* Science Parse parses scientific papers (in PDF form) and returns them in structured form. **Why it matters.** Science Parse is a Java-based tool that extracts structured data like titles, authors, abstracts, sections, and bibliographies from PDF scientific papers, making it easier to process and analyze academic literature programmatically. It matters now because the growing volume of research papers requires efficient automation for tasks like literature reviews and data mining, especially with AI advancements in natural language processing, though its age and the existence of a newer version suggest it may not be the most cutting-edge option. _themes: pdf-parsing · text-extraction · nlp · research-tools_ #### [google-deepmind/treescope](https://github.com/google-deepmind/treescope) *Python · ★467 · Apache-2.0 · beta · score:0.70 · hot:0.43 · rising:0.48 · durable:0.65 · board:durable · trend:down* An interactive HTML pretty-printer for machine learning research in IPython notebooks. **Why it matters.** Treescope is a Python library that serves as an interactive HTML pretty-printer for ML objects in IPython notebooks, enabling features like expandable subtrees, tensor visualizations, and color-coded neural network structures to aid in debugging and exploration. It matters right now because ML research increasingly involves complex, high-dimensional data where intuitive visualization can accelerate insights and reduce errors, and as a drop-in replacement for standard renderers, it integrates seamlessly with popular libraries like JAX and PyTorch amid growing demands for efficient development tools. _themes: visualization · ml-debugging · notebooks · tensor-viz_ #### [google-research/augmix](https://github.com/google-research/augmix) *Python · ★992 · Apache-2.0 · production · score:0.60 · hot:0.43 · rising:0.50 · durable:0.61 · board:durable · trend:down* AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty **Why it matters.** AugMix is a data augmentation technique that mixes augmented images and enforces consistent embeddings to improve model robustness and uncertainty calibration in image classification tasks, without needing extensive tuning. It addresses challenges in datasets with distribution shifts, such as ImageNet-C, and has shown significant improvements in benchmarks, but its 2020 origins mean it may not incorporate recent advancements in the field. While useful for enhancing model reliability, its lack of updates and fixed dependencies could limit applicability in modern workflows. _themes: data-augmentation · robustness · uncertainty-estimation · image-classification_ #### [google-deepmind/disco_rl](https://github.com/google-deepmind/disco_rl) *Python · ★686 · Apache-2.0 · archived · score:0.70 · hot:0.43 · rising:0.45 · durable:0.56 · board:durable · trend:down* Accompanying code for "Discovering State-of-the-art Reinforcement Algorithms" Nature publication **Why it matters.** This repository provides code for experimenting with meta-learned reinforcement learning algorithms, specifically the Disco103 update rule from a Nature publication, using a JAX-based harness for meta-evaluation and meta-training. It allows researchers to reproduce the study's results or extend the work, but its lack of active maintenance limits its long-term utility in a rapidly evolving field like RL. _themes: meta-learning · reinforcement-learning · jax · algorithms_ #### [google-deepmind/dqn_zoo](https://github.com/google-deepmind/dqn_zoo) *Python · ★496 · Apache-2.0 · experimental · score:0.70 · hot:0.43 · rising:0.47 · durable:0.58 · board:durable · trend:down* DQN Zoo is a collection of reference implementations of reinforcement learning agents developed at DeepMind based on the Deep Q-Network (DQN) agent. **Why it matters.** DQN Zoo provides reference implementations of various Deep Q-Network (DQN) based reinforcement learning agents, developed by DeepMind, allowing researchers to replicate and experiment with these algorithms on standard Atari benchmarks. It matters now because foundational RL techniques like DQN remain essential for advancing modern AI research, especially as newer methods build upon them, though its lack of recent updates may limit its appeal for cutting-edge applications. _themes: rl · dqn · reinforcement-learning · jax_ #### [google-deepmind/opro](https://github.com/google-deepmind/opro) *Python · ★733 · Apache-2.0 · experimental · score:0.70 · hot:0.43 · rising:0.47 · durable:0.59 · board:durable · trend:down* official code for "Large Language Models as Optimizers" **Why it matters.** This repository provides code for using large language models like GPT and PaLM as optimizers for tasks such as prompt engineering and solving problems like linear regression or the traveling salesman problem, based on a specific research paper. It matters now because it explores practical applications of LLMs in optimization, potentially improving AI development workflows, though it relies on external APIs and lacks formal releases, making it more of a proof-of-concept than a robust tool. _themes: llm · optimization · prompt-engineering_ #### [allenai/procthor](https://github.com/allenai/procthor) *Python · ★429 · Apache-2.0 · experimental · score:0.70 · hot:0.42 · rising:0.46 · durable:0.59 · board:durable · trend:down* 🏘️ Scaling Embodied AI by Procedurally Generating Interactive 3D Houses **Why it matters.** ProcTHOR is a Python library that procedurally generates interactive 3D houses for use in embodied AI simulations, specifically compatible with AI2-THOR, allowing researchers to create diverse environments for training AI agents in computer vision and robotics tasks. It matters right now because embodied AI is advancing rapidly in areas like robotics and simulation-based learning, and this tool helps address the scalability challenges by automating environment generation, as demonstrated in its award-winning NeurIPS paper. However, its early release version suggests it may still have limitations in stability and features compared to more established simulators. _themes: embodied-ai · procedural-generation · computer-vision_ #### [facebookresearch/Detic](https://github.com/facebookresearch/Detic) *Python · ★1,996 · Apache-2.0 · experimental · score:0.75 · hot:0.42 · rising:0.47 · durable:0.56 · board:durable · trend:down* Code release for "Detecting Twenty-thousand Classes using Image-level Supervision". **Why it matters.** Detic is a Python-based library for training object detectors using only image-level supervision, enabling detection of up to 20,000 classes via integration with CLIP, which reduces the need for costly bounding box annotations. It demonstrates strong generalization across datasets like OpenImages without fine-tuning, making it relevant for scenarios where annotation resources are limited, though its lack of a formal release and reliance on research-grade models may limit immediate practical adoption. This approach aligns with current trends in scalable computer vision but requires users to handle dependencies and potential instability. _themes: object-detection · zero-shot-learning · image-supervision · computer-vision_ #### [facebookresearch/videoseal](https://github.com/facebookresearch/videoseal) *Python · ★630 · MIT · beta · score:0.70 · hot:0.42 · rising:0.47 · durable:0.59 · board:durable · trend:down* Open and efficient video and image watermarking **Why it matters.** VideoSeal is an open-source library for embedding invisible watermarks in images and videos, featuring state-of-the-art models like PixelSeal for high imperceptibility and robustness, and ChunkySeal for increased capacity up to 1024 bits. It matters now due to growing concerns over digital content authenticity amid AI-generated media and deepfakes, with its NeurIPS spotlight highlighting timely advancements in watermarking technology. _themes: watermarking · image · video · ml_ #### [google-research/uda](https://github.com/google-research/uda) *Python · ★2,205 · Apache-2.0 · experimental · score:0.80 · hot:0.42 · rising:0.47 · durable:0.56 · board:durable · trend:stable* Unsupervised Data Augmentation (UDA) **Why it matters.** UDA is a semi-supervised learning method that uses unsupervised data augmentation to enhance model performance on NLP and computer vision tasks with very few labeled examples, outperforming prior state-of-the-art on benchmarks like IMDb and CIFAR-10. It matters now because the growing scarcity and cost of labeled data in AI development make efficient use of unlabeled data essential for advancing models in resource-constrained environments. However, its lack of recent releases may limit immediate applicability in production settings. _themes: semi-supervised · data-augmentation · cv · nlp_ #### [facebookresearch/vggsfm](https://github.com/facebookresearch/vggsfm) *Python · ★1,375 · NOASSERTION · beta · score:0.75 · hot:0.42 · rising:0.46 · durable:0.56 · board:durable · trend:down* VGGSfM: Visual Geometry Grounded Deep Structure From Motion **Why it matters.** VGGSfM is a deep learning-based library for Structure from Motion (SfM) that integrates visual geometry to improve 3D reconstruction from 2D images and videos, including handling dynamic scenes with masks. It matters right now because of the increasing demand for robust 3D modeling in AR/VR and generative AI, with recent updates like video processing and point cloud export making it timely for computer vision research and applications. _themes: computer-vision · 3d-reconstruction · deep-learning · sfm_ #### [facebookresearch/watermark-anything](https://github.com/facebookresearch/watermark-anything) *Jupyter Notebook · ★1,112 · MIT · experimental · score:0.75 · hot:0.42 · rising:0.46 · durable:0.60 · board:durable · trend:down* Official implementation of the paper "Watermark Anything with Localized Messages" **Why it matters.** This repository implements a method for embedding localized watermarks into images, as described in the paper 'Watermark Anything', allowing for potentially multiple watermarks per image to aid in authenticity and copyright protection. It matters now due to increasing concerns over AI-generated content and deepfakes, with the technique's recent acceptance at ICLR 2025 and a new model release highlighting its relevance for research in image security, though its practical robustness remains unproven in real-world scenarios. _themes: image-watermarking · computer-vision · ai-security · deep-learning_ #### [facebookresearch/madgrad](https://github.com/facebookresearch/madgrad) *Python · ★802 · MIT · beta · score:0.60 · hot:0.42 · rising:0.47 · durable:0.60 · board:durable · trend:down* MADGRAD Optimization Method **Why it matters.** MADGRAD is a PyTorch optimizer that combines momentum and adaptive gradient techniques to offer faster convergence than Adam while maintaining SGD-like generalization, making it a potential alternative for training deep learning models. It matters now because optimizing training efficiency is critical in handling larger datasets and models in AI research, though its lack of formal releases and moderate adoption (802 stars) suggests it's still experimental and not widely proven. Users should be cautious, as it requires tuning parameters like learning rate and weight decay. _themes: optimization · pytorch · deep-learning · training_ #### [allenai/papermage](https://github.com/allenai/papermage) *Python · ★794 · Apache-2.0 · experimental · score:0.50 · hot:0.42 · rising:0.45 · durable:0.58 · board:durable · trend:down* library supporting NLP and CV research on scientific papers **Why it matters.** Papermage is a Python library for processing and analyzing scientific papers using NLP and computer vision, allowing users to extract, represent, and manipulate text and visual elements from PDFs. It matters for researchers in multimodal AI due to the growing need for tools in scientific document analysis, but its value is diminished by the lack of active maintenance, as the team has shifted focus to a new project, making it less reliable for ongoing use. _themes: nlp · computer-vision · pdf-processing · multimodal_ #### [google-deepmind/pycolab](https://github.com/google-deepmind/pycolab) *Python · ★663 · Apache-2.0 · beta · score:0.60 · hot:0.42 · rising:0.48 · durable:0.59 · board:durable · trend:down* A highly-customisable gridworld game engine with some batteries included. Make your own gridworld games to test reinforcement learning agents! **Why it matters.** Pycolab is a Python library for creating customizable gridworld environments to test reinforcement learning agents, offering tools for building and running simple games. It matters for educational purposes in RL, as gridworlds are foundational for algorithm development, but its lack of recent updates and support for outdated Python versions make it less relevant in today's rapidly evolving AI landscape, potentially limiting its utility for current projects. _themes: rl · gridworld · environments · simulation_ #### [facebookresearch/MILS](https://github.com/facebookresearch/MILS) *Python · ★459 · NOASSERTION · experimental · score:0.60 · hot:0.42 · rising:0.45 · durable:0.55 · board:durable · trend:down* Code release for "LLMs can see and hear without any training" **Why it matters.** This repository contains code for a paper demonstrating that large language models can handle visual and audio inputs without additional training, likely relying on pre-existing capabilities or simple adapters. It matters now amid growing interest in multimodal AI, but critically, its approach may not be as universally applicable or innovative as implied, given the need for specific datasets and checkpoints, making it a specialized tool for researchers exploring LLM limits. _themes: llms · multimodal · zero-shot · inference_ #### [huggingface/text-clustering](https://github.com/huggingface/text-clustering) *Python · ★602 · Apache-2.0 · experimental · score:0.60 · hot:0.42 · rising:0.46 · durable:0.56 · board:durable · trend:down* Easily embed, cluster and semantically label text datasets **Why it matters.** This repository provides a simple pipeline for embedding, clustering, and semantically labeling text datasets using standard libraries, making it accessible for quick experimentation with large text corpora. It matters now as text clustering is increasingly vital for handling unstructured data in AI workflows, especially with the surge in NLP applications, and this tool offers an easy entry point without requiring extensive custom code. However, being a work in progress, it may lack polish for production use. _themes: nlp · clustering · embedding · visualization_ #### [google-deepmind/aloha_sim](https://github.com/google-deepmind/aloha_sim) *Python · ★307 · NOASSERTION · experimental · score:0.70 · hot:0.42 · rising:0.45 · durable:0.57 · board:durable · trend:down* A collection of tabletop tasks in Mujoco **Why it matters.** Aloha Sim is a Python library that provides simulation environments for the Aloha robot using Mujoco, featuring tabletop tasks for robot learning and evaluation. It matters right now because it supports safe, cost-effective testing of robotic policies, which is increasingly important amid rapid advancements in AI-driven robotics, though its access is restricted to trusted testers for inference features. However, its lack of a formal release and low adoption (307 stars) limits immediate impact. _themes: robotics · simulation · mujoco · evaluation_ #### [google-deepmind/alphastar](https://github.com/google-deepmind/alphastar) *Python · ★559 · Apache-2.0 · experimental · score:0.70 · hot:0.42 · rising:0.46 · durable:0.58 · board:durable · trend:down* **Why it matters.** AlphaStar is a DeepMind package that provides tools for training AI agents in StarCraft II using reinforcement learning architectures and offline training scripts, focusing on offline RL with examples like Behavior Cloning. It matters for researchers exploring complex game environments, as it builds on DeepMind's expertise in AI milestones, but its lack of recent releases and limited OS support make it less practical for ongoing development compared to more maintained alternatives. _themes: reinforcement-learning · agents · offline-rl · game-ai_ #### [deepseek-ai/LPLB](https://github.com/deepseek-ai/LPLB) *Python · ★499 · MIT · experimental · score:0.50 · hot:0.42 · rising:0.45 · durable:0.57 · board:durable · trend:down* An early research stage expert-parallel load balancer for MoE models based on linear programming. **Why it matters.** This repo implements an expert-parallel load balancer for Mixture-of-Experts (MoE) models using linear programming to optimize workload distribution, which helps in dynamically balancing computational loads in distributed training. It matters now because efficient load balancing is critical for scaling large AI models, but its early research stage means it's not yet reliable for production and requires further validation. _themes: moe · load-balancing · distributed-computing · linear-programming_ #### [facebookresearch/sonata](https://github.com/facebookresearch/sonata) *Python · ★719 · Apache-2.0 · beta · score:0.75 · hot:0.41 · rising:0.46 · durable:0.57 · board:durable · trend:down* [CVPR'25 Highlight] Official repository of Sonata: Self-Supervised Learning of Reliable Point Representations **Why it matters.** Sonata provides self-supervised pre-trained models based on Point Transformer V3 for 3D point cloud tasks, focusing on reliable representation learning without labeled data, along with inference code and visualization tools. It matters now as it's a recent CVPR 2025 Highlight, advancing 3D vision techniques amid growing applications in robotics and autonomous systems, though it's limited to providing models rather than full training code. _themes: self-supervised · 3d-vision · point-clouds · inference_ #### [google-deepmind/spriteworld](https://github.com/google-deepmind/spriteworld) *Python · ★373 · Apache-2.0 · production · score:0.60 · hot:0.41 · rising:0.50 · durable:0.67 · board:durable · trend:down* Spriteworld: a flexible, configurable python-based reinforcement learning environment **Why it matters.** Spriteworld is a Python-based reinforcement learning environment that simulates a 2D arena with configurable sprites for tasks like object manipulation and exploration, primarily designed for research in model-based RL and unsupervised learning as seen in the 2019 COBRA paper. It offers flexibility in procedural scene generation and various action spaces, making it useful for testing RL algorithms, but its age and limited features mean it may not address current challenges in more advanced RL setups like those involving complex physics or real-world applications. _themes: rl · environment · procedural · generative_ #### [openai/consistencydecoder](https://github.com/openai/consistencydecoder) *Python · ★2,210 · MIT · experimental · score:0.75 · hot:0.41 · rising:0.45 · durable:0.61 · board:durable · trend:down* Consistency Distilled Diff VAE **Why it matters.** The Consistency Decoder repo from OpenAI provides an improved decoding method for Stable Diffusion VAEs using consistency distillation, which enhances image quality and reduces artifacts in generated outputs. It matters right now because it builds on recent advancements in diffusion models, potentially speeding up image generation workflows for researchers and developers amid growing interest in efficient AI-generated content, as seen in tools like DALL·E 3. _themes: diffusion · vae · image-generation · decoding_ #### [facebookresearch/audioseal](https://github.com/facebookresearch/audioseal) *Python · ★706 · MIT · beta · score:0.70 · hot:0.41 · rising:0.46 · durable:0.56 · board:durable · trend:down* Localized watermarking for AI-generated speech audios, with SOTA on robustness and very fast detector **Why it matters.** AudioSeal is a Python library that implements localized watermarking for AI-generated speech audio, training a generator to embed imperceptible watermarks and a detector to identify them with state-of-the-art robustness and speed. It matters now because the rising threat of deepfake audio in media and communication demands reliable tools for content authentication and combating misinformation, especially as regulations around AI-generated content intensify. _themes: watermarking · audio · ai-safety · detection_ #### [facebookresearch/DepthLM_Official](https://github.com/facebookresearch/DepthLM_Official) *Python · ★327 · NOASSERTION · experimental · score:0.70 · hot:0.41 · rising:0.43 · durable:0.56 · board:durable · trend:down* [ICLR 2026 Oral (top 1.2%)] Official implementation of DepthLM **Why it matters.** DepthLM is a Python implementation that enables vision-language models to perform metric depth estimation with accuracy comparable to specialized vision models, using only standard text-based fine-tuning without architectural changes. This approach simplifies handling multiple 3D tasks like speed estimation and camera pose in a unified model, which could reduce complexity in AI pipelines for applications in robotics and AR/VR. It's relevant now as multimodal AI advancements are accelerating, potentially offering more efficient alternatives to traditional vision-based methods. _themes: vlm · depth-estimation · fine-tuning · 3d-understanding_ #### [openai/sparse_autoencoder](https://github.com/openai/sparse_autoencoder) *Python · ★582 · MIT · experimental · score:0.70 · hot:0.41 · rising:0.46 · durable:0.58 · board:durable · trend:down* **Why it matters.** This repository provides sparse autoencoders trained on GPT2-small activations and a visualizer for inspecting those features, aimed at improving neural network interpretability. It matters now because AI interpretability is a growing concern for model safety and understanding, especially with increasing scrutiny on large language models, though its utility is limited to specific research applications without broader integration or production-ready features. _themes: interpretability · autoencoders · ai-safety_ #### [google-research/dex-lang](https://github.com/google-research/dex-lang) *Haskell · ★1,674 · BSD-3-Clause · experimental · score:0.65 · hot:0.41 · rising:0.45 · durable:0.53 · board:durable · trend:down* Research language for array processing in the Haskell/ML family **Why it matters.** Dex is a research language built on Haskell and ML principles, designed for typed functional array processing with features like automatic differentiation, integration, and compilation to parallel hardware. It matters now as the AI and machine learning fields increasingly demand efficient, type-safe numerical computations, potentially offering insights into optimizing array operations for high-performance applications. However, its experimental nature means it's primarily useful for academic exploration rather than immediate practical deployment. _themes: arrays · functional · differentiation · parallel_ #### [allenai/pdffigures2](https://github.com/allenai/pdffigures2) *Scala · ★740 · Apache-2.0 · beta · score:0.70 · hot:0.41 · rising:0.46 · durable:0.53 · board:durable · trend:stable* Given a scholarly PDF, extract figures, tables, captions, and section titles. **Why it matters.** PDFFigures 2.0 is a Scala library that extracts figures, tables, captions, and section titles from scholarly PDFs, particularly in computer science, by analyzing page elements and saving them as structured data or images. It matters now for automating research workflows and knowledge extraction in AI-driven document processing, but its lack of recent releases and focus on a specific domain may limit broader applicability in fast-evolving tools. _themes: pdf-extraction · document-analysis · figures · tables_ #### [facebookresearch/schedule_free](https://github.com/facebookresearch/schedule_free) *Python · ★2,274 · Apache-2.0 · experimental · score:0.70 · hot:0.41 · rising:0.44 · durable:0.62 · board:durable · trend:down* Schedule-Free Optimization in PyTorch **Why it matters.** This repo provides PyTorch implementations of schedule-free optimizers, such as variants of SGD and AdamW, which eliminate the need for learning rate schedules by using interpolation and averaging techniques. It simplifies training by removing the requirement to specify stopping times or steps in advance, potentially reducing hyperparameter tuning efforts. This matters now as efficient optimization is crucial for scaling large-scale machine learning models in research and production. _themes: optimization · pytorch · deep-learning · efficient-training_ #### [google-deepmind/surface-distance](https://github.com/google-deepmind/surface-distance) *Python · ★594 · Apache-2.0 · experimental · score:0.60 · hot:0.41 · rising:0.45 · durable:0.54 · board:durable · trend:down* Library to compute surface distance based performance metrics for segmentation tasks. **Why it matters.** This repository provides a Python library for computing surface distance-based metrics, such as average surface distance and Hausdorff distance, to evaluate the accuracy of image segmentation tasks by comparing predicted surfaces to ground truth. It matters now because precise segmentation metrics are essential in advancing AI applications in fields like medical imaging and autonomous systems, where traditional metrics like Dice coefficient may not fully capture surface alignment issues; however, its lack of formal releases and basic installation process might limit broader adoption. _themes: segmentation · metrics · computer-vision · evaluation_ #### [google-research/lottery-ticket-hypothesis](https://github.com/google-research/lottery-ticket-hypothesis) *Python · ★727 · Apache-2.0 · experimental · score:0.70 · hot:0.40 · rising:0.45 · durable:0.55 · board:durable · trend:down* A reimplementation of "The Lottery Ticket Hypothesis" (Frankle and Carbin) on MNIST. **Why it matters.** This repository provides a Python reimplementation of the Lottery Ticket Hypothesis, focusing on pruning and retraining subnetworks in neural networks using the MNIST dataset, which helps explore why overparameterized networks train better. It matters now for advancing efficient AI through pruning techniques amid growing demands for lightweight models, but its age and lack of updates mean it may not incorporate recent developments in the field. _themes: pruning · neural-networks · initialization · efficiency_ #### [huggingface/speechbox](https://github.com/huggingface/speechbox) *Python · ★357 · Apache-2.0 · archived · score:0.30 · hot:0.40 · rising:0.42 · durable:0.50 · board:durable · trend:down* **Why it matters.** Hugging Face's Speechbox is a Python library providing basic speech processing tools, primarily focused on tasks like punctuation restoration for transcribed audio. It matters for developers building speech-related applications, but its relevance is limited due to the repository being unmaintained, potentially leading to compatibility issues and lack of updates in a fast-evolving AI landscape. _themes: speech-processing · punctuation-restoration · nlp · machine-learning_ #### [google-research/vmoe](https://github.com/google-research/vmoe) *Jupyter Notebook · ★715 · Apache-2.0 · experimental · score:0.70 · hot:0.40 · rising:0.44 · durable:0.53 · board:durable · trend:down* **Why it matters.** This repository provides code for training and fine-tuning Sparse Mixture of Experts (MoE) models for vision tasks, such as on ImageNet-21k, based on research papers from Google. It matters now because MoE architectures enable efficient scaling of vision models by activating only subsets of parameters, addressing the growing need for cost-effective AI amid resource constraints, though it's still in early stages with limited ready-to-use resources. _themes: vision · moe · scaling · fine-tuning_ #### [facebookresearch/stable_signature](https://github.com/facebookresearch/stable_signature) *Jupyter Notebook · ★509 · NOASSERTION · experimental · score:0.70 · hot:0.40 · rising:0.42 · durable:0.52 · board:durable · trend:down* Official implementation of the paper "The Stable Signature Rooting Watermarks in Latent Diffusion Models" **Why it matters.** This repository implements a watermarking technique for Latent Diffusion Models, enabling the embedding and extraction of watermarks in generated images to verify authenticity and detect origins. It matters now because the proliferation of generative AI raises concerns about misuse, such as deepfakes and intellectual property theft, making robust watermarking essential for research and potential regulatory compliance, though its effectiveness and generalizability remain to be proven in real-world scenarios. _themes: watermarking · diffusion-models · ai-security_ #### [huggingface/open-muse](https://github.com/huggingface/open-muse) *Python · ★358 · Apache-2.0 · experimental · score:0.60 · hot:0.40 · rising:0.43 · durable:0.51 · board:durable · trend:down* Open reproduction of MUSE for fast text2image generation. **Why it matters.** This repository provides an open-source reproduction of the MUSE model, a transformer-based approach for fast text-to-image generation using VQGAN and large datasets. It matters right now because it enables researchers to experiment with and verify advanced generative AI techniques amid growing interest in efficient text-to-image models, potentially accelerating innovation in creative AI applications. _themes: text2image · transformers · diffusion · generative-ai_ #### [google-research/prompt-tuning](https://github.com/google-research/prompt-tuning) *Python · ★699 · Apache-2.0 · experimental · score:0.70 · hot:0.40 · rising:0.43 · durable:0.55 · board:durable · trend:down* Original Implementation of Prompt Tuning from Lester, et al, 2021 **Why it matters.** This repository provides the original implementation for prompt tuning, a parameter-efficient technique to adapt large language models by training only a small set of additional parameters, as introduced in the 2021 paper by Lester et al. It allows researchers to reproduce experiments and apply this method to NLP tasks using frameworks like T5X and JAX, which is relevant today amid growing interest in efficient fine-tuning for resource-constrained environments, though the approach is not the latest in a rapidly evolving field. _themes: prompt-tuning · fine-tuning · nlp · jax_ #### [huggingface/controlnet_aux](https://github.com/huggingface/controlnet_aux) *Python · ★488 · Apache-2.0 · beta · score:0.60 · hot:0.40 · rising:0.44 · durable:0.54 · board:durable · trend:down* **Why it matters.** This repository provides a Python library that repackages auxiliary annotators from lllyasviel's ControlNet for easier installation and use in image processing tasks, such as edge detection and depth estimation, with Hugging Face integration. It matters for developers building AI image generation tools as it simplifies setup and dependency management, but it's essentially a convenience wrapper without original contributions, making it useful yet not transformative in the current AI landscape. _themes: computer-vision · image-processing · stable-diffusion · inference_ #### [huggingface/exporters](https://github.com/huggingface/exporters) *Python · ★695 · Apache-2.0 · experimental · score:0.60 · hot:0.40 · rising:0.42 · durable:0.53 · board:durable · trend:down* Export Hugging Face models to Core ML and TensorFlow Lite **Why it matters.** This repository provides tools to export Hugging Face Transformers models to Core ML for deployment on Apple devices, and recommends Optimum for TensorFlow Lite conversions, but it's still a work in progress with no official releases. It matters for developers building mobile AI apps where on-device inference is key for privacy and performance, though its experimental status means users should expect potential instability and verify compatibility first. _themes: model-conversion · inference · deployment · transformers_ #### [google-deepmind/dm_env](https://github.com/google-deepmind/dm_env) *Python · ★399 · Apache-2.0 · production · score:0.70 · hot:0.39 · rising:0.46 · durable:0.59 · board:durable · trend:down* A Python interface for reinforcement learning environments **Why it matters.** dm_env is a Python library that provides a standardized interface for reinforcement learning environments, including abstract base classes, time step containers, and specification tools to ensure consistency in RL implementations. It matters now because reinforcement learning is a key area in AI research and development, and a reliable interface from DeepMind helps streamline code sharing and testing, especially as RL applications grow in fields like robotics and simulation. _themes: reinforcement-learning · api · environments · rl-interface_ #### [huggingface/diarizers](https://github.com/huggingface/diarizers) *Python · ★326 · no-license · experimental · score:0.60 · hot:0.39 · rising:0.41 · durable:0.52 · board:durable · trend:down* **Why it matters.** This repository provides a library for fine-tuning pyannote speaker diarization models using Hugging Face tools, allowing users to enhance performance on English and multilingual datasets with minimal data and compute resources. It matters because speaker diarization is increasingly important for applications like audio transcription and meeting analysis, but its lack of a license and official releases raises concerns about usability and stability in production environments. _themes: fine-tuning · diarization · audio · huggingface_ #### [facebookresearch/eft](https://github.com/facebookresearch/eft) *Python · ★404 · NOASSERTION · experimental · score:0.60 · hot:0.39 · rising:0.41 · durable:0.48 · board:durable · trend:down* visualization code for 3D human body annotation by EFT (Exemplar Fine-tuning) **Why it matters.** This repository provides code, pseudo-ground truth data, and a pre-trained model for 3D human pose estimation using Exemplar Fine-Tuning, allowing users to train pose estimation algorithms without indoor datasets. It matters for advancing computer vision research in in-the-wild scenarios, as demonstrated in the 2021 paper, but its lack of formal releases and updates since then limits its immediate relevance compared to newer methods. _themes: 3d-pose · fine-tuning · computer-vision · pose-estimation_ #### [google-deepmind/nanodo](https://github.com/google-deepmind/nanodo) *Python · ★308 · Apache-2.0 · experimental · score:0.60 · hot:0.39 · rising:0.44 · durable:0.57 · board:durable · trend:down* **Why it matters.** NanoDO is a minimal implementation of a Transformer decoder-only language model in JAX, aimed at providing a hackable and readable base for researchers to conduct exploratory work without heavy abstractions. It matters now as the AI field demands transparent tools for rapid iteration amid complex model developments, though its experimental nature and lack of a formal release limit immediate broad applicability. _themes: transformer · jax · language-model · minimal-implementation_ #### [QwenLM/ParScale](https://github.com/QwenLM/ParScale) *Python · ★478 · no-license · experimental · score:0.60 · hot:0.39 · rising:0.41 · durable:0.54 · board:durable · trend:down* Parallel Scaling Law for Language Model — Beyond Parameter and Inference Time Scaling **Why it matters.** ParScale introduces a parallel scaling method for language models that uses multiple parallel streams for computation during training and inference, dynamically aggregating outputs to achieve efficiency gains without proportionally increasing parameters or inference time. This matters now because it offers a cost-effective alternative to traditional scaling approaches, potentially reducing resource demands in an era of rapidly growing LLMs, and it demonstrates logarithmic scaling benefits that could enhance performance on reasoning tasks amid ongoing AI efficiency challenges. _themes: scaling-law · parallel-computing · llm · efficiency_ #### [google-deepmind/enn](https://github.com/google-deepmind/enn) *Python · ★313 · Apache-2.0 · experimental · score:0.60 · hot:0.39 · rising:0.43 · durable:0.51 · board:durable · trend:down* **Why it matters.** This repository provides a library for Epistemic Neural Networks (ENNs), which enhance uncertainty estimation in neural networks by incorporating an epistemic index to make joint predictions, distinguishing between genuine ambiguity and model limitations. It matters because reliable uncertainty quantification is increasingly critical in AI applications like safety-critical systems, where traditional methods like Bayesian Neural Networks fall short, and this library offers a flexible interface built on JAX and Haiku for exploring these ideas. _themes: uncertainty · neural-networks · jax · probabilistic-ml_ #### [allenai/s2orc-doc2json](https://github.com/allenai/s2orc-doc2json) *Python · ★464 · Apache-2.0 · beta · score:0.60 · hot:0.39 · rising:0.44 · durable:0.54 · board:durable · trend:down* Parsers for scientific papers (PDF2JSON, TEX2JSON, JATS2JSON) **Why it matters.** This repository provides Python-based parsers to convert scientific papers from PDF, TeX, and JATS formats into a custom JSON schema, primarily for use in projects like S2ORC and CORD-19, enabling structured extraction of paper components for NLP and data analysis. It matters now due to the growing demand for standardized academic data in AI research, but its relevance is tempered by the lack of recent releases and active maintenance, potentially making it less reliable for current applications. _themes: pdf-parsing · json-conversion · scientific-documents · data-processing_ #### [huggingface/hmtl](https://github.com/huggingface/hmtl) *Python · ★1,195 · MIT · archived · score:0.50 · hot:0.39 · rising:0.41 · durable:0.48 · board:durable · trend:down* 🌊HMTL: Hierarchical Multi-Task Learning - A State-of-the-Art neural network model for several NLP tasks based on PyTorch and AllenNLP **Why it matters.** HMTL is a hierarchical multi-task learning model for NLP that combines tasks like named entity recognition and relation extraction to achieve state-of-the-art results as of 2018, using PyTorch and AllenNLP. It demonstrates how shared representations can improve semantic understanding, which remains conceptually relevant for modern NLP research, but the repo's lack of updates since then limits its practical value today compared to more recent frameworks. _themes: multi-task-learning · nlp · pytorch · hierarchical_ #### [allenai/OLMoASR](https://github.com/allenai/OLMoASR) *Python · ★488 · MIT · experimental · score:0.60 · hot:0.39 · rising:0.43 · durable:0.55 · board:durable · trend:down* An open-source implementation of Whisper **Why it matters.** OLMoASR is an open-source repository that outlines the process for training speech recognition models similar to Whisper, covering data processing, training, and evaluation using Python. It matters now because it promotes transparency and accessibility in AI development, enabling researchers to experiment with robust speech models in an era of increasing demand for open alternatives to proprietary systems like OpenAI's Whisper. _themes: speech-recognition · fine-tuning · open-source · data-processing_ #### [google-deepmind/lab2d](https://github.com/google-deepmind/lab2d) *C++ · ★439 · Apache-2.0 · production · score:0.60 · hot:0.38 · rising:0.47 · durable:0.66 · board:durable · trend:down* A customisable 2D platform for agent-based AI research **Why it matters.** DeepMind Lab2D is a C++ and Lua-based library for creating customizable 2D grid worlds, allowing AI researchers to define environments with text maps and scripts for agent interactions via Python or C APIs. It supports performance-focused machine learning experiments, particularly in reinforcement learning, but its modest 439 stars and v1.0.0 release indicate it's not as widely adopted or actively updated as some alternatives. While useful for specialized agent-based research, it may lack the ecosystem support of more popular frameworks in the current AI landscape. _themes: agents · reinforcement-learning · simulation_ #### [microsoft/triton-shared](https://github.com/microsoft/triton-shared) *MLIR · ★330 · MIT · archived · score:0.30 · hot:0.38 · rising:0.40 · durable:0.44 · board:durable · trend:down* Shared Middle-Layer for Triton Compilation **Why it matters.** This repository provides a shared middle layer for the Triton compiler, translating Triton IR into MLIR dialects to enable hardware-agnostic compilation and sharing across backends. However, as it is no longer maintained, its relevance is mostly as a reference for researchers exploring MLIR-based compiler designs. It reflects broader efforts in improving portability for accelerator programming, but lacks ongoing development, making it less impactful in the current ecosystem. _themes: mlir · compiler · triton · acceleration_ #### [google-research/morph-net](https://github.com/google-research/morph-net) *Python · ★1,035 · Apache-2.0 · beta · score:0.70 · hot:0.38 · rising:0.45 · durable:0.56 · board:durable · trend:down* Fast & Simple Resource-Constrained Learning of Deep Network Structure **Why it matters.** MorphNet is a library for learning and optimizing deep network structures under resource constraints like FLOPs and model size during training, using regularizers to prune unnecessary parts. It matters now because efficient model design is critical for deploying AI on edge devices and mobile platforms, where computational resources are limited, and its FiGS extension improves upon traditional pruning methods for better architecture search. However, with no recent releases since its 2018 introduction, its practical applicability may be waning compared to more actively maintained alternatives. _themes: automl · neural-architecture-search · pruning · deep-learning_ #### [huggingface/dataspeech](https://github.com/huggingface/dataspeech) *Python · ★394 · MIT · experimental · score:0.60 · hot:0.38 · rising:0.42 · durable:0.51 · board:durable · trend:down* **Why it matters.** Data-Speech is a collection of Python scripts for annotating speech datasets with speaker characteristics and audio transformations, primarily to support text-to-speech model development as described in a specific research paper. It matters for researchers fine-tuning TTS models like Parler-TTS, given the ongoing demand for high-quality annotated datasets in AI speech synthesis, but its niche focus and lack of official releases limit broader applicability. _themes: speech · annotation · fine-tuning · TTS_ #### [google-research/leaf-audio](https://github.com/google-research/leaf-audio) *Python · ★524 · Apache-2.0 · experimental · score:0.70 · hot:0.38 · rising:0.42 · durable:0.52 · board:durable · trend:down* LEAF is a learnable alternative to audio features such as mel-filterbanks, that can be initialized as an approximation of mel-filterbanks, and then be trained for the task at hand, while using a very small number of parameters. **Why it matters.** LEAF is a learnable audio frontend that serves as an alternative to fixed features like mel-filterbanks, allowing it to be initialized as an approximation and then fine-tuned for specific tasks with minimal parameters, making it efficient for audio processing in machine learning pipelines. It matters now because the growing demand for optimized audio models in applications like speech recognition and sound classification benefits from adaptable features that can improve performance without excessive computational overhead. _themes: audio · feature-extraction · learnable · tensorflow_ #### [allenai/openie-standalone](https://github.com/allenai/openie-standalone) *Scala · ★333 · NOASSERTION · archived · score:0.40 · hot:0.37 · rising:0.39 · durable:0.45 · board:durable · trend:down* Quality information extraction at web scale. Edit **Why it matters.** This repo provides an Open Information Extraction system that extracts relations from text sentences into structured triples or n-ary formats, useful for NLP tasks like relation identification. However, its relevance is limited today due to lack of recent updates and maintenance, making it less competitive against modern libraries in a rapidly evolving AI landscape. It remains a foundational academic tool for understanding basic information extraction techniques. _themes: nlp · information-extraction · openie · text-processing_ #### [allenai/bi-att-flow](https://github.com/allenai/bi-att-flow) *Python · ★1,542 · Apache-2.0 · archived · score:0.30 · hot:0.37 · rising:0.41 · durable:0.44 · board:durable · trend:stable* Bi-directional Attention Flow (BiDAF) network is a multi-stage hierarchical process that represents context at different levels of granularity and uses a bi-directional attention flow mechanism to achieve a query-aware context representation without early summarization. **Why it matters.** This repository implements the BiDAF model for question-answering tasks, using bi-directional attention to process context and queries in NLP, particularly on datasets like SQuAD. It was influential in early QA research but is now outdated, relying on TensorFlow 0.11 from 2016, making it less relevant for modern applications where advanced models like transformers dominate. Its value lies primarily in educational or historical contexts for understanding foundational attention mechanisms. _themes: nlp · question-answering · attention · tensorflow_ #### [openai/vdvae](https://github.com/openai/vdvae) *Python · ★452 · MIT · experimental · score:0.65 · hot:0.37 · rising:0.43 · durable:0.54 · board:durable · trend:down* Repository for the paper "Very Deep VAEs Generalize Autoregressive Models and Can Outperform Them on Images" **Why it matters.** This repository implements Very Deep VAEs, a generative model that extends traditional VAEs to handle complex image distributions and can outperform autoregressive models in certain image generation tasks, as detailed in the associated research paper. It matters for AI research because it challenges the dominance of autoregressive methods and provides a scalable alternative for image synthesis, though its practical impact is limited by its age and specific hardware requirements. The code enables experimentation with datasets like CIFAR-10 and ImageNet, but lacks modern updates or broader applicability. _themes: vae · generative-models · image-generation · deep-learning_ #### [openai/ebm_code_release](https://github.com/openai/ebm_code_release) *Python · ★365 · no-license · experimental · score:0.40 · hot:0.37 · rising:0.39 · durable:0.49 · board:durable · trend:down* Code for Implicit Generation and Generalization with Energy Based Models **Why it matters.** This repository provides code for implementing Energy Based Models (EBMs) for implicit generation and generalization, based on a 2019 OpenAI paper, allowing users to train models on datasets like CIFAR-10 and ImageNet. It matters for researchers exploring alternative generative techniques, but its age and lack of updates make it less relevant amid modern advancements in diffusion and GAN models. The absence of a license and official releases limits its practical adoption. _themes: generative-models · energy-based · ml-research_ #### [allenai/dont-stop-pretraining](https://github.com/allenai/dont-stop-pretraining) *Python · ★541 · no-license · experimental · score:0.60 · hot:0.36 · rising:0.39 · durable:0.49 · board:durable · trend:down* Code associated with the Don't Stop Pretraining ACL 2020 paper **Why it matters.** This repository offers code for domain-adaptive pretraining (DAPT) and task-adaptive pretraining (TAPT) of language models, based on a 2020 ACL paper, enabling researchers to fine-tune models for specific domains and tasks to improve performance. It matters for understanding foundational techniques in NLP adaptation, but its age and lack of updates or a proper license make it less practical for current applications compared to more recent tools. The reliance on a pinned, outdated version of AllenNLP could hinder reproducibility and integration with modern frameworks. _themes: nlp · pretraining · fine-tuning · domain-adaptation_ #### [google-research/tf-slim](https://github.com/google-research/tf-slim) *Python · ★372 · Apache-2.0 · archived · score:0.40 · hot:0.35 · rising:0.38 · durable:0.46 · board:durable · trend:down* **Why it matters.** TF-Slim is a lightweight library that simplifies defining, training, and evaluating neural networks in TensorFlow by providing high-level abstractions and pre-built models like VGG and AlexNet, reducing boilerplate code. However, it is based on older TensorFlow versions (up to 2.2) and has not been updated recently, making it less relevant today with Keras integrated into TensorFlow as the standard high-level API. It may still be useful for legacy projects but offers little new value for current development. _themes: tensorflow · deep-learning · model-training · computer-vision_ #### [allenai/XNOR-Net](https://github.com/allenai/XNOR-Net) *Lua · ★866 · NOASSERTION · archived · score:0.50 · hot:0.34 · rising:0.37 · durable:0.43 · board:durable · trend:down* ImageNet classification using binary Convolutional Neural Networks **Why it matters.** This repository provides a Torch 7.0 implementation of XNOR-Net, a binary convolutional neural network for ImageNet classification that uses binary weights to reduce computational demands. It matters because efficient neural networks are still relevant for edge devices and resource-constrained environments, but this 2016 code is outdated, relies on deprecated tools like Lua and Torch, and lacks modern maintenance, making it more of a historical artifact than a practical solution today. _themes: binary-nets · cnn · efficiency · imagenet_ #### [allenai/kb](https://github.com/allenai/kb) *Python · ★375 · Apache-2.0 · experimental · score:0.50 · hot:0.32 · rising:0.36 · durable:0.47 · board:durable · trend:down* KnowBert -- Knowledge Enhanced Contextual Word Representations **Why it matters.** KnowBert enhances BERT by integrating knowledge from sources like Wikipedia and WordNet to improve contextual word representations, which can boost performance in NLP tasks requiring external knowledge. However, as this 2019 research lacks recent updates and active maintenance, its practical value is limited compared to modern alternatives that build on similar ideas with better scalability. It's primarily useful for academic exploration rather than production use today. _themes: nlp · bert · knowledge-enhanced · pretraining_ ### model (114) #### [NVIDIA/Isaac-GR00T](https://github.com/NVIDIA/Isaac-GR00T) *Python · ★6,735 · Apache-2.0 · beta · score:0.80 · hot:0.80 · rising:0.81 · durable:0.72 · board:rising · trend:up* NVIDIA Isaac GR00T N1.7 - A Foundation Model for Generalist Robots. **Why it matters.** NVIDIA Isaac GR00T is a foundation model for generalist robots, enabling them to perform tasks through vision-language models and fine-tuning on custom data. It matters right now as it advances AI-driven robotics amid growing interest in physical AI applications, though it's in early access with limited stability. This release from NVIDIA provides pre-trained weights and tools for experimentation, potentially accelerating research in embodied AI. _themes: robotics · foundation-model · fine-tuning · inference_ #### [openai/whisper](https://github.com/openai/whisper) *Python · ★98,030 · MIT · production · score:0.90 · hot:0.77 · rising:0.85 · durable:0.88 · board:durable · trend:up* Robust Speech Recognition via Large-Scale Weak Supervision **Why it matters.** Whisper is an open-source speech recognition model from OpenAI that performs multilingual transcription, translation, and language identification using a transformer-based architecture trained on diverse audio data. It matters now because it provides accessible, high-accuracy tools for developers and researchers in an era of increasing demand for voice AI in applications like transcription services and multilingual interfaces, though its performance can vary with audio quality and accents. _themes: speech-recognition · multilingual · inference · multitask_ #### [google-deepmind/alphagenome_research](https://github.com/google-deepmind/alphagenome_research) *Python · ★735 · Apache-2.0 · beta · score:0.80 · hot:0.76 · rising:0.76 · durable:0.74 · board:rising · trend:up* Research code accompanying AlphaGenome **Why it matters.** AlphaGenome is a DNA sequence model that predicts regulatory variant effects and genome functions by analyzing sequences up to 1 million base pairs, offering insights into gene expression and chromatin features. It matters now because AI-driven genomics is advancing rapidly, potentially improving disease research and personalized medicine, though its research-focused code requires specialized hardware, limiting immediate accessibility. _themes: genomics · ai-biology · sequence-modeling · jax_ #### [microsoft/VibeVoice](https://github.com/microsoft/VibeVoice) *Python · ★40,291 · MIT · experimental · score:0.75 · hot:0.73 · rising:0.77 · durable:0.73 · board:rising · trend:up* Open-Source Frontier Voice AI **Why it matters.** VibeVoice is an open-source repository from Microsoft offering advanced voice AI models, including a speech-to-text model for long-form audio with structured transcriptions and a real-time text-to-speech model supporting multiple languages; it provides tools for finetuning and inference, making it useful for handling complex audio tasks. It matters now because voice AI is increasingly critical for applications like transcription and virtual assistants, and this release democratizes access to high-quality, multilingual models that were previously proprietary, though its experimental status means it may require further refinement for production use. _themes: asr · tts · multilingual · fine-tuning_ #### [google-deepmind/alphafold3](https://github.com/google-deepmind/alphafold3) *Python · ★7,858 · NOASSERTION · production · score:0.90 · hot:0.73 · rising:0.78 · durable:0.80 · board:durable · trend:up* AlphaFold 3 inference pipeline. **Why it matters.** AlphaFold 3 provides an inference pipeline for predicting 3D structures of proteins and their interactions with other biomolecules, extending the capabilities of previous versions for more accurate biomolecular simulations. It matters now because it advances drug discovery and biological research amid growing AI applications in science, especially following the recent publication of its research paper, though access to model parameters is restricted and requires direct approval from Google. This repo enables researchers to run inferences if they obtain the parameters, but it's limited to non-commercial use and subject to specific terms. _themes: inference · protein-folding · structure-prediction · ai-for-science_ #### [google-research/timesfm](https://github.com/google-research/timesfm) *Python · ★18,147 · Apache-2.0 · beta · score:0.85 · hot:0.73 · rising:0.77 · durable:0.79 · board:durable · trend:up* TimesFM (Time Series Foundation Model) is a pretrained time-series foundation model developed by Google Research for time-series forecasting. **Why it matters.** TimesFM is a pretrained foundation model for time-series forecasting, enabling efficient predictions on various datasets with minimal fine-tuning. It matters now due to its integration into Google tools like BigQuery and Sheets, addressing the growing demand for scalable forecasting in industries such as finance and IoT, though it's not officially supported and has evolving versions. _themes: forecasting · time-series · foundation-model · fine-tuning_ #### [fishaudio/fish-speech](https://github.com/fishaudio/fish-speech) *Python · ★29,813 · NOASSERTION · beta · score:0.85 · hot:0.70 · rising:0.72 · durable:0.79 · board:durable · trend:stable* SOTA Open Source TTS **Why it matters.** Fish Audio S2 is an open-source text-to-speech system built with advanced models like transformers and VITS, enabling high-quality multilingual voice generation, though its custom license may restrict commercial use. It matters now amid growing interest in AI-driven speech synthesis for applications like virtual assistants, but users must navigate potential legal risks due to its disclaimers on illegal applications. _themes: tts · inference · multilingual · voice-generation_ #### [facebookresearch/sam-3d-objects](https://github.com/facebookresearch/sam-3d-objects) *Python · ★6,468 · NOASSERTION · experimental · score:0.75 · hot:0.69 · rising:0.69 · durable:0.63 · board:rising · trend:up* SAM 3D Objects **Why it matters.** This repository provides models for 3D object reconstruction as part of Meta's SAM 3D project, focusing on generating meshes from inputs like images or point clouds. It matters right now because 3D reconstruction is advancing rapidly in fields like AR/VR and robotics, and this work from Meta could enhance accuracy and efficiency, though its lack of a formal release and unspecified license may limit immediate adoption. _themes: 3d-reconstruction · computer-vision · mesh-generation · deep-learning_ #### [google-deepmind/graphcast](https://github.com/google-deepmind/graphcast) *Python · ★6,599 · Apache-2.0 · beta · score:0.80 · hot:0.69 · rising:0.71 · durable:0.76 · board:durable · trend:stable* **Why it matters.** GraphCast is a deep learning model from Google DeepMind that uses graph neural networks for accurate medium-range weather forecasting, outperforming traditional methods in research benchmarks. This repository provides Python code, pretrained weights, and utilities to run and train the model using external datasets, making it accessible for replication and experimentation. It matters now because advancing AI-driven weather prediction can enhance climate modeling and disaster response amid growing environmental challenges, though its reliance on specific datasets and computational resources limits broader adoption. _themes: weather-forecasting · graph-nn · inference · fine-tuning_ #### [google-deepmind/alphafold](https://github.com/google-deepmind/alphafold) *Python · ★14,497 · Apache-2.0 · production · score:0.95 · hot:0.68 · rising:0.74 · durable:0.79 · board:durable · trend:stable* Open source code for AlphaFold 2. **Why it matters.** This repository provides the open-source implementation of AlphaFold 2, a deep learning model for predicting protein structures from amino acid sequences with high accuracy. It matters now because it enables widespread access to advanced structural biology tools, accelerating research in drug development and disease understanding, though users should note that features like AlphaFold-Multimer are still in progress and not fully stable. _themes: protein-folding · deep-learning · inference · structure-prediction_ #### [facebookresearch/sam3](https://github.com/facebookresearch/sam3) *Python · ★9,092 · NOASSERTION · experimental · score:0.75 · hot:0.67 · rising:0.67 · durable:0.60 · board:hot · trend:up* The repository provides code for running inference and finetuning with the Meta Segment Anything Model 3 (SAM 3), links for downloading the trained model checkpoints, and example notebooks that show how to use the model. **Why it matters.** SAM 3 is an advanced computer vision model for segmenting objects in images, building on Meta's previous Segment Anything series by providing code for inference and fine-tuning, along with pre-trained checkpoints and example notebooks. It matters now because image segmentation is crucial for applications like autonomous systems and medical imaging, where precise object detection can drive innovation, though its research-focused nature means it may require adaptation for production use. _themes: segmentation · inference · fine-tuning · computer-vision_ #### [QwenLM/Qwen3.6](https://github.com/QwenLM/Qwen3.6) *? · ★2,873 · Apache-2.0 · beta · score:0.60 · hot:0.65 · rising:0.66 · durable:0.64 · board:rising · trend:stable* Qwen3.6 is the large language model series developed by Qwen team, Alibaba Group. **Why it matters.** Qwen3.6 is a large language model developed by Alibaba's Qwen team, focusing on improvements in coding assistance, such as agentic workflows and context retention across conversations, building on the Qwen3.5 foundation. It aims to enhance developer productivity by prioritizing stability and real-world utility based on community feedback, but its lack of a formal release and incomplete documentation raises questions about its readiness. Overall, it represents incremental progress in the LLM space amid ongoing competition from established models. _themes: llm · agents · coding · efficiency_ #### [openai/gpt-oss](https://github.com/openai/gpt-oss) *Python · ★20,019 · Apache-2.0 · beta · score:0.80 · hot:0.64 · rising:0.70 · durable:0.81 · board:durable · trend:stable* gpt-oss-120b and gpt-oss-20b are two open-weight language models by OpenAI **Why it matters.** This repository provides OpenAI's open-weight language models, gpt-oss-120b and gpt-oss-20b, designed for reasoning and agentic tasks, with specific hardware requirements and a mandatory harmony response format for functionality. It matters now as it signals OpenAI's cautious move towards open-source AI, potentially fostering broader access and innovation, but its early release version and restrictions may limit immediate adoption compared to more mature alternatives. _themes: agents · inference · open-source · large-language-models_ #### [facebookresearch/sam2](https://github.com/facebookresearch/sam2) *Jupyter Notebook · ★18,979 · Apache-2.0 · beta · score:0.90 · hot:0.64 · rising:0.67 · durable:0.70 · board:durable · trend:stable* The repository provides code for running inference with the Meta Segment Anything Model 2 (SAM 2), links for downloading the trained model checkpoints, and example notebooks that show how to use the model. **Why it matters.** SAM 2 is an open-source implementation for running inference on Meta's Segment Anything Model 2, which enables promptable visual segmentation in both images and videos, building on the original SAM for more dynamic applications. It matters now because video segmentation is increasingly critical for fields like video editing and autonomous systems, and this release from Meta provides accessible tools and checkpoints that could accelerate research and development, though it lacks a formal release version which might indicate ongoing refinements. _themes: segmentation · vision · inference · video_ #### [facebookresearch/seamless_communication](https://github.com/facebookresearch/seamless_communication) *Jupyter Notebook · ★11,776 · NOASSERTION · beta · score:0.85 · hot:0.64 · rising:0.66 · durable:0.66 · board:durable · trend:stable* Foundational Models for State-of-the-Art Speech and Text Translation **Why it matters.** This repository from Facebook Research introduces the Seamless family of AI models, including SeamlessM4T for multilingual multimodal translation, SeamlessExpressive for preserving prosody and voice style, and SeamlessStreaming for real-time translation across about 100 languages, enabling more natural cross-language communication. It matters now because advancements in real-time, expressive translation address growing needs in global collaboration, media, and AI-driven applications, though its lack of formal releases and unclear licensing may limit immediate adoption. The models build on recent multimodal AI progress but require further evaluation for practical reliability. _themes: multilingual · speech-translation · multimodal · real-time_ #### [facebookresearch/perception_models](https://github.com/facebookresearch/perception_models) *Jupyter Notebook · ★2,250 · Apache-2.0 · beta · score:0.75 · hot:0.63 · rising:0.64 · durable:0.63 · board:rising · trend:stable* State-of-the-art Image & Video CLIP, Multimodal Large Language Models, and More! **Why it matters.** This repository from Meta Research provides state-of-the-art models like the Perception Encoder for encoding images, videos, and audio, and the Perception Language Model for decoding, enabling advanced multimodal perception tasks. It matters right now because it advances audiovisual understanding in AI, building on CLIP and similar technologies, which are increasingly vital for applications in content analysis, AR/VR, and social media amid growing demand for integrated multimodal systems. _themes: multimodal · vision · audio · language-model_ #### [facebookresearch/boxer](https://github.com/facebookresearch/boxer) *Python · ★444 · NOASSERTION · experimental · score:0.60 · hot:0.63 · rising:0.60 · durable:0.56 · board:hot · trend:stable* Code for the Boxer research paper **Why it matters.** Boxer is a tool that converts 2D object detections from images and point clouds into 3D oriented bounding boxes, specifically for indoor environments, using pre-trained models for inference only. It matters now because accurate 3D perception is increasingly important for applications in AR/VR, robotics, and spatial computing, though its research-focused nature limits immediate widespread adoption. _themes: 3d-detection · computer-vision · inference · point-clouds_ #### [facebookresearch/EUPE](https://github.com/facebookresearch/EUPE) *Python · ★502 · NOASSERTION · experimental · score:0.70 · hot:0.61 · rising:0.59 · durable:0.57 · board:hot · trend:stable* Efficient Universal Perception Encoder: a single on-device vision encoder with versatile representations that match or exceed specialized experts across multiple task domains. **Why it matters.** EUPE is a vision encoder from Meta AI that uses distillation to create a single, efficient model capable of handling diverse tasks like image understanding and vision-language modeling, potentially outperforming specialized models. It matters now because the push for on-device AI efficiency amid resource constraints makes versatile encoders valuable for reducing deployment complexity, though its real-world performance remains to be validated beyond benchmarks. _themes: vision · encoder · distillation · efficiency_ #### [facebookresearch/all-atom-diffusion-transformer](https://github.com/facebookresearch/all-atom-diffusion-transformer) *Python · ★302 · NOASSERTION · experimental · score:0.70 · hot:0.61 · rising:0.58 · durable:0.56 · board:hot · trend:stable* Official implementation of All Atom Diffusion Transformers (ICML 2025) **Why it matters.** This repository implements All Atom Diffusion Transformers, a unified latent diffusion framework for generating both periodic materials and non-periodic molecules, as detailed in an upcoming ICML 2025 paper. It matters because it could enhance generative modeling in chemistry and materials science by providing a single approach for diverse systems, potentially aiding drug discovery and material design, though its real-world impact remains unproven given the lack of releases or extensive validation. _themes: diffusion · generative · transformers · materials-science_ #### [facebookresearch/actionmesh](https://github.com/facebookresearch/actionmesh) *Python · ★335 · NOASSERTION · experimental · score:0.70 · hot:0.60 · rising:0.58 · durable:0.58 · board:hot · trend:stable* 🎬ActionMesh: A fast video to animated mesh model with unprecedented quality. Generate animated mesh seamlessly importable into any 3D software in less than a minute. **Why it matters.** ActionMesh is a Python-based model that generates animated 3D meshes from videos in under a minute, with options to integrate existing meshes while preserving topology and texture, making it useful for 3D content creation. It matters now because it advances accessible 4D reconstruction techniques amid growing demand in AR/VR and gaming, with features like low-RAM mode and a benchmark dataset enhancing its practicality for rapid prototyping and evaluation. _themes: 4d-generation · mesh-generation · animation · video-processing_ #### [facebookresearch/MHR](https://github.com/facebookresearch/MHR) *Python · ★656 · Apache-2.0 · beta · score:0.70 · hot:0.60 · rising:0.62 · durable:0.68 · board:durable · trend:stable* Momentum Human Rig is an anatomically-inspired parametric full-body digital human model developed at Meta. It includes: A parametric body skeletal model; A realistic 3D mesh skinned to the skeleton with levels of detail;A body blendshape and pose corrective model; A facial blendshape model.Its design is friendly for both CG and CV communities. **Why it matters.** MHR is a parametric 3D human body model from Meta that enables detailed control over body identity, pose, and facial expressions for applications in computer graphics and computer vision. It matters now due to the growing demand for realistic digital humans in AI-driven fields like AR/VR and animation, though its value is limited by its specificity and early-stage adoption, potentially overlapping with existing models without clear innovation. _themes: 3d-modeling · human-pose · animation · pytorch_ #### [facebookresearch/lagernvs](https://github.com/facebookresearch/lagernvs) *Python · ★309 · NOASSERTION · experimental · score:0.60 · hot:0.59 · rising:0.55 · durable:0.54 · board:hot · trend:stable* Official code for "LagerNVS Latent Geometry for Fully Neural Real-time Novel View Synthesis" (CVPR 2026) **Why it matters.** This repository provides the official code for LagerNVS, a feed-forward neural model that synthesizes novel views from input images in real-time without explicit 3D representations, making it suitable for in-the-wild scenarios but with limitations like unsatisfactory results when camera intrinsics differ significantly. It matters for computer vision research as it advances fully neural approaches to novel view synthesis, though its experimental nature and known issues, such as the camera intrinsics problem, mean it's not yet reliable for practical applications. _themes: neural-rendering · novel-view-synthesis · real-time-rendering · computer-vision_ #### [deepseek-ai/DeepSeek-V3](https://github.com/deepseek-ai/DeepSeek-V3) *Python · ★102,727 · MIT · production · score:0.85 · hot:0.59 · rising:0.70 · durable:0.89 · board:durable · trend:stable* **Why it matters.** DeepSeek-V3 is an open-source large language model from DeepSeek AI, focused on advanced natural language processing and generation capabilities, building on prior versions for improved efficiency and performance. It matters right now due to the ongoing AI model arms race, with its high star count indicating significant community adoption and potential for real-world applications in research and development; however, its true impact depends on benchmark results and accessibility compared to established alternatives. _themes: llm · inference · fine-tuning · open-source_ #### [facebookresearch/dinov3](https://github.com/facebookresearch/dinov3) *Jupyter Notebook · ★10,163 · NOASSERTION · beta · score:0.85 · hot:0.58 · rising:0.60 · durable:0.67 · board:durable · trend:stable* Reference PyTorch implementation and models for DINOv3 **Why it matters.** DINOv3 is a self-supervised vision transformer model providing PyTorch implementations and pre-trained backbones for tasks like semantic segmentation and monocular depth estimation, with recent updates including distillation code and integrations with Hugging Face. It matters now because these enhancements make it more accessible for advanced computer vision applications, building on prior work while addressing real-world needs like improved accuracy in height mapping, though the future-dated README entries raise questions about the timeline of developments. _themes: self-supervised · vision-transformers · segmentation · depth-estimation_ #### [deepseek-ai/DeepSeek-R1](https://github.com/deepseek-ai/DeepSeek-R1) *? · ★91,960 · MIT · beta · score:0.80 · hot:0.58 · rising:0.68 · durable:0.85 · board:durable · trend:stable* **Why it matters.** DeepSeek-R1 is an AI reasoning model, including DeepSeek-R1-Zero, which is trained using large-scale reinforcement learning without supervised fine-tuning, aiming to enhance autonomous learning in AI systems. This approach matters now because it challenges traditional training methods and could lead to more efficient model development, especially amid growing interest in scalable RL techniques for complex tasks, though its real-world effectiveness remains to be proven given it's a first-generation release. _themes: reinforcement-learning · reasoning · ai-models · large-language-models_ #### [huggingface/smollm](https://github.com/huggingface/smollm) *Python · ★3,720 · Apache-2.0 · beta · score:0.80 · hot:0.57 · rising:0.60 · durable:0.66 · board:durable · trend:stable* Everything about the SmolLM and SmolVLM family of models **Why it matters.** This repository from Hugging Face features the SmolLM and SmolVLM families, which are open-source, lightweight language and vision-language models optimized for on-device efficiency, including a 3B parameter model that competes with larger alternatives while being fully transparent about training. It matters now amid the push for resource-efficient AI, as it provides accessible models and detailed training recipes for developers facing hardware constraints, though its lack of a formal release raises questions about long-term support and reliability. _themes: efficient-models · open-source · multimodal · on-device_ #### [QwenLM/Qwen3-Coder](https://github.com/QwenLM/Qwen3-Coder) *Python · ★16,398 · no-license · beta · score:0.80 · hot:0.57 · rising:0.61 · durable:0.69 · board:durable · trend:stable* Qwen3-Coder is the code version of Qwen3, the large language model series developed by Qwen team. **Why it matters.** Qwen3-Coder is a large language model specialized for code-related tasks, such as generating, completing, and debugging code, built on the Qwen3 series. It matters now because the growing demand for AI-assisted development tools is driving productivity in software engineering, especially as open-source LLMs continue to compete with proprietary options in a rapidly evolving AI landscape. _themes: llm · code-generation · inference · fine-tuning_ #### [QwenLM/Qwen3-VL-Embedding](https://github.com/QwenLM/Qwen3-VL-Embedding) *Python · ★1,189 · Apache-2.0 · beta · score:0.70 · hot:0.57 · rising:0.57 · durable:0.59 · board:durable · trend:stable* **Why it matters.** This repo provides state-of-the-art multimodal embedding and reranking models based on Qwen3-VL, enabling processing of text, images, videos, and mixed inputs for improved information retrieval and cross-modal tasks. It matters now due to the growing demand for efficient multimodal AI in applications like search and content recommendation, building on recent advancements in foundation models like Qwen3-VL. However, without a formal release, its immediate practical impact is limited pending further development and validation. _themes: multimodal · embedding · retrieval · reranking_ #### [facebookresearch/tribev2](https://github.com/facebookresearch/tribev2) *Jupyter Notebook · ★1,909 · NOASSERTION · experimental · score:0.70 · hot:0.56 · rising:0.56 · durable:0.60 · board:durable · trend:stable* This repository contains the code to train and evaluate TRIBE v2, a multimodal model for brain response prediction **Why it matters.** TRIBE v2 is a multimodal model that predicts fMRI brain responses to video, audio, and text stimuli using a unified Transformer architecture, primarily for neuroscience research. It matters now because it advances in-silico brain modeling, potentially aiding studies in cognitive science amid growing interest in AI for healthcare, though its specialized application limits broader adoption. _themes: multimodal · neuroscience · fMRI · transformer_ #### [google-deepmind/tapnet](https://github.com/google-deepmind/tapnet) *Jupyter Notebook · ★1,850 · Apache-2.0 · experimental · score:0.75 · hot:0.55 · rising:0.56 · durable:0.61 · board:durable · trend:stable* Tracking Any Point (TAP) **Why it matters.** TAPNet provides datasets like TAP-Vid and TAPVid-3D for benchmarking point tracking in videos, along with the TAPIR model that accurately tracks arbitrary points across frames using a two-stage approach, and RoboTAP for applying this to robotics tasks. It matters now because precise point tracking advances computer vision applications in robotics and video analysis, where real-time accuracy is critical for emerging technologies like autonomous systems, and it outperforms existing methods on standard benchmarks. _themes: computer-vision · point-tracking · deep-learning · robotics_ #### [xai-org/grok-1](https://github.com/xai-org/grok-1) *Python · ★51,519 · Apache-2.0 · experimental · score:0.80 · hot:0.55 · rising:0.62 · durable:0.74 · board:durable · trend:stable* Grok open release **Why it matters.** This repo provides JAX code for loading and running the Grok-1 model, a 314B parameter Mixture of Experts (MoE) AI model with open weights, allowing users to experiment with a large-scale language model on compatible hardware. It matters right now because it represents xAI's entry into open-source AI, potentially accelerating research amid the competitive landscape of large language models, though its inefficient implementation limits practical use. The release highlights ongoing efforts to democratize access to advanced AI capabilities, but requires significant resources, making it more of a proof-of-concept than a ready-to-use tool. _themes: llm · jax · inference · moe_ #### [QwenLM/Qwen3-TTS](https://github.com/QwenLM/Qwen3-TTS) *Python · ★10,790 · Apache-2.0 · beta · score:0.80 · hot:0.54 · rising:0.59 · durable:0.71 · board:durable · trend:stable* Qwen3-TTS is an open-source series of TTS models developed by the Qwen team at Alibaba Cloud, supporting stable, expressive, and streaming speech generation, free-form voice design, and vivid voice cloning. **Why it matters.** Qwen3-TTS is an open-source text-to-speech model series from Alibaba's Qwen team, offering features like expressive speech generation, voice cloning, and streaming capabilities for applications needing high-quality audio output. It matters now because it advances TTS technology with user-friendly tools for voice design and cloning, potentially democratizing access to sophisticated speech synthesis amid growing demand in AI-driven apps, though its lack of a formal release raises questions about stability. _themes: tts · voice-cloning · speech-synthesis · inference_ #### [facebookresearch/segment-anything](https://github.com/facebookresearch/segment-anything) *Jupyter Notebook · ★53,975 · Apache-2.0 · production · score:0.95 · hot:0.54 · rising:0.63 · durable:0.78 · board:durable · trend:stable* The repository provides code for running inference with the SegmentAnything Model (SAM), links for downloading the trained model checkpoints, and example notebooks that show how to use the model. **Why it matters.** This repository provides the Segment Anything Model (SAM) for zero-shot image segmentation and its extension SAM 2 for video segmentation, enabling users to perform promptable visual segmentation with pre-trained models and example code. It matters now because SAM has become a foundational tool in computer vision, influencing research and applications, while SAM 2's real-time video capabilities address emerging needs in video analysis, AI content creation, and multimodal tasks, especially with the growing demand for efficient foundation models. _themes: segmentation · computer-vision · foundation-model · inference_ #### [deepseek-ai/Janus](https://github.com/deepseek-ai/Janus) *Python · ★17,690 · MIT · beta · score:0.80 · hot:0.53 · rising:0.58 · durable:0.73 · board:durable · trend:stable* Janus-Series: Unified Multimodal Understanding and Generation Models **Why it matters.** Janus is a series of unified models for multimodal understanding and generation, handling tasks like text, images, and their combinations through a single framework, as seen in its latest iterations like Janus-Pro and JanusFlow. It matters right now because the demand for efficient, open-source multimodal AI is surging amid advancements in foundation models, enabling researchers and developers to experiment with 'any-to-any' capabilities without relying on proprietary systems, though it still faces issues like bugs in tokenization that affect performance. _themes: multimodal · llm · foundation-models · vision-language_ #### [google-deepmind/videoprism](https://github.com/google-deepmind/videoprism) *Python · ★365 · Apache-2.0 · beta · score:0.80 · hot:0.53 · rising:0.54 · durable:0.63 · board:durable · trend:stable* Official repository for "VideoPrism: A Foundational Visual Encoder for Video Understanding" (ICML 2024) **Why it matters.** VideoPrism is a pre-trained video encoder from Google DeepMind that handles tasks like classification, retrieval, and question answering by leveraging a massive dataset of image-text pairs and video clips, achieving state-of-the-art results on most benchmarks with a single frozen model. It matters now because video understanding is a growing priority in AI, especially for applications in multimodal learning and content analysis, and this model provides an efficient foundation for researchers to build upon amid increasing demands for scalable video AI solutions. _themes: self-supervised-learning · video-understanding · vision-transformer · multimodal_ #### [zai-org/ChatGLM-6B](https://github.com/zai-org/ChatGLM-6B) *Python · ★41,158 · Apache-2.0 · production · score:0.80 · hot:0.53 · rising:0.61 · durable:0.74 · board:durable · trend:stable* ChatGLM-6B: An Open Bilingual Dialogue Language Model | 开源双语对话语言模型 **Why it matters.** ChatGLM-6B is an open-source, bilingual (Chinese-English) dialogue language model with 6.2 billion parameters, based on the GLM architecture, optimized for low-resource deployment on consumer hardware via quantization. It matters now because it provides accessible AI for multilingual applications in a landscape dominated by English-centric models, and its popularity reflects growing demand for open-source alternatives amid rapid LLM advancements, though the project seems less active with no recent releases. _themes: bilingual · dialogue · inference · quantization_ #### [QwenLM/Qwen3](https://github.com/QwenLM/Qwen3) *Python · ★27,147 · no-license · beta · score:0.80 · hot:0.53 · rising:0.58 · durable:0.72 · board:durable · trend:stable* Qwen3 is the large language model series developed by Qwen team, Alibaba Cloud. **Why it matters.** Qwen3 is a series of large language models developed by Alibaba's Qwen team, focused on advanced text generation and understanding capabilities, with models available on platforms like Hugging Face. It matters now because it builds on prior Qwen iterations with potential improvements in efficiency and performance, as detailed in a recent paper, but its lack of a specified license limits accessibility and adoption for commercial or derivative uses. _themes: inference · language-model · fine-tuning · deployment_ #### [facebookresearch/vjepa2](https://github.com/facebookresearch/vjepa2) *Python · ★3,655 · MIT · experimental · score:0.60 · hot:0.53 · rising:0.55 · durable:0.61 · board:durable · trend:stable* PyTorch code and models for VJEPA2 self-supervised learning from video. **Why it matters.** V-JEPA 2 is a PyTorch implementation for self-supervised video learning models that focus on motion understanding and action anticipation using internet-scale data, with extensions for robot manipulation tasks. It matters now because it demonstrates efficient video representation learning without extensive labeling, potentially advancing applications in robotics and AI planning, though its real-world impact remains unproven due to the lack of official releases and future-dated announcements. However, as self-supervised methods gain traction for scalable AI, this could influence research in temporal data handling. _themes: self-supervised · video-learning · robotics · pytorch_ #### [deepseek-ai/DeepSeek-Coder](https://github.com/deepseek-ai/DeepSeek-Coder) *Python · ★23,057 · MIT · beta · score:0.85 · hot:0.52 · rising:0.60 · durable:0.74 · board:durable · trend:stable* DeepSeek Coder: Let the Code Write Itself **Why it matters.** DeepSeek Coder is an open-source suite of code language models trained on 2 trillion tokens, primarily code, to enable advanced code completion and infilling across multiple languages, with sizes from 1B to 33B parameters. It matters now because it offers a competitive, accessible alternative to proprietary tools like GitHub Copilot, potentially improving developer productivity amid growing demand for AI-assisted coding, though its real-world performance beyond benchmarks remains to be fully validated. _themes: code-generation · llm · inference · fine-tuning_ #### [facebookresearch/vggt](https://github.com/facebookresearch/vggt) *Python · ★12,893 · NOASSERTION · beta · score:0.85 · hot:0.51 · rising:0.57 · durable:0.68 · board:durable · trend:stable* [CVPR 2025 Best Paper Award] VGGT: Visual Geometry Grounded Transformer **Why it matters.** VGGT is a transformer-based model that incorporates visual geometry to enhance computer vision tasks, as highlighted by its CVPR 2025 Best Paper Award. It matters now because the recent update allows commercial use for non-military applications, making advanced AI research more accessible, while its high adoption (over 12,000 stars) indicates strong interest in improving vision model accuracy and efficiency. _themes: vision · transformer · geometry · fine-tuning_ #### [google-deepmind/recurrentgemma](https://github.com/google-deepmind/recurrentgemma) *Python · ★670 · Apache-2.0 · beta · score:0.75 · hot:0.51 · rising:0.55 · durable:0.71 · board:durable · trend:down* Open weights language model from Google DeepMind, based on Griffin. **Why it matters.** RecurrentGemma is an open-weights language model from Google DeepMind that uses the Griffin architecture to improve inference efficiency for long sequences by replacing global attention with a mix of local attention and linear recurrences. It matters now as the AI field increasingly prioritizes computational efficiency for real-world applications like text generation, offering a potential alternative to resource-intensive transformer models amid growing concerns over cost and scalability. _themes: inference · fine-tuning · language-models · recurrent_ #### [deepseek-ai/DeepSeek-OCR](https://github.com/deepseek-ai/DeepSeek-OCR) *Python · ★22,845 · MIT · experimental · score:0.80 · hot:0.51 · rising:0.57 · durable:0.70 · board:durable · trend:stable* Contexts Optical Compression **Why it matters.** DeepSeek-OCR is an open-source OCR model that focuses on vision encoders integrated with large language models, aiming to enhance text recognition from images by adopting an LLM-centric approach. It matters right now because it bridges vision and language processing, potentially improving multimodal AI applications, and its integration with tools like vLLM makes it practical for developers amid growing demand for efficient OCR in real-world scenarios. _themes: ocr · vision · inference · llm_ #### [NVIDIA/personaplex](https://github.com/NVIDIA/personaplex) *Python · ★9,459 · MIT · experimental · score:0.70 · hot:0.50 · rising:0.56 · durable:0.68 · board:durable · trend:stable* PersonaPlex code. **Why it matters.** PersonaPlex is a real-time speech-to-speech model that supports full-duplex conversations with customizable personas via text prompts and voice conditioning, based on the Moshi architecture for low-latency interactions. It matters now because it addresses the growing demand for natural, personalized voice AI in applications like virtual assistants and interactive media, though its reliance on NVIDIA hardware and lack of a formal release may limit accessibility. The open-source MIT license and integration with Hugging Face make it a practical starting point for developers exploring advanced speech tech. _themes: speech · conversational-ai · inference · real-time_ #### [NVIDIA/DreamDojo](https://github.com/NVIDIA/DreamDojo) *Python · ★744 · Apache-2.0 · beta · score:0.75 · hot:0.50 · rising:0.52 · durable:0.63 · board:durable · trend:stable* Official Codebase for "DreamDojo: A Generalist Robot World Model from Large-Scale Human Videos" **Why it matters.** NVIDIA's DreamDojo repo provides code for training a generalist robot world model using 44k hours of human egocentric videos, enabling robots to generalize across diverse objects and environments through pretraining and distillation for real-time interactions. It matters now as it advances foundation models for robotics, potentially accelerating development in autonomous systems amid growing interest in AI-driven physical interactions, though its real-world applicability remains to be proven beyond demonstrations. _themes: world-model · robotics · video-pretraining · distillation_ #### [QwenLM/Qwen](https://github.com/QwenLM/Qwen) *Python · ★21,027 · Apache-2.0 · archived · score:0.40 · hot:0.50 · rising:0.53 · durable:0.63 · board:durable · trend:stable* The official repo of Qwen (通义千问) chat & pretrained large language model proposed by Alibaba Cloud. **Why it matters.** Qwen is a large language model developed by Alibaba Cloud, focused on chat and pretrained capabilities with a strong emphasis on Chinese language processing, including features like flash-attention for efficiency. It matters as it demonstrates advancements in open-source LLMs from major tech firms, but its relevance is limited now since the repository is no longer actively maintained and users are directed to the successor, Qwen2. Despite over 20,000 stars, it represents historical value rather than current innovation due to the codebase being deprecated. _themes: llm · chinese-nlp · pretrained-models · flash-attention_ #### [microsoft/TRELLIS](https://github.com/microsoft/TRELLIS) *Python · ★12,211 · MIT · experimental · score:0.85 · hot:0.50 · rising:0.55 · durable:0.67 · board:durable · trend:stable* Official repo for paper "Structured 3D Latents for Scalable and Versatile 3D Generation" (CVPR'25 Spotlight). **Why it matters.** TRELLIS is a 3D asset generation model that uses text or image prompts to produce high-quality 3D outputs like radiance fields and meshes, leveraging structured latents and large-scale pre-trained transformers. It matters right now because it advances scalable 3D generation amid increasing demand for AI-driven content creation in fields like gaming and design, while offering flexible formats and editing capabilities that outperform many existing methods, though its lack of a formal release may limit immediate adoption. _themes: 3d-generation · text-to-3d · image-to-3d · transformers_ #### [QwenLM/Qwen3-VL](https://github.com/QwenLM/Qwen3-VL) *Jupyter Notebook · ★18,996 · Apache-2.0 · beta · score:0.85 · hot:0.50 · rising:0.56 · durable:0.70 · board:durable · trend:stable* Qwen3-VL is the multimodal large language model series developed by Qwen team, Alibaba Cloud. **Why it matters.** Qwen3-VL is a multimodal large language model developed by Alibaba's Qwen team, focusing on enhanced vision-language integration for tasks like text understanding, visual reasoning, and agent interactions. It matters now because the AI industry is prioritizing multimodal capabilities for real-world applications, and this open-source model offers scalable architectures amid increasing competition from tech giants, potentially lowering barriers for developers working on vision-based AI. _themes: multimodal · vision-language · llm · inference_ #### [deepseek-ai/DeepSeek-LLM](https://github.com/deepseek-ai/DeepSeek-LLM) *Makefile · ★6,820 · MIT · beta · score:0.70 · hot:0.50 · rising:0.56 · durable:0.69 · board:durable · trend:stable* DeepSeek LLM: Let there be answers **Why it matters.** DeepSeek-LLM provides open-source large language models with up to 67 billion parameters, trained on 2 trillion tokens in English and Chinese, focusing on strengths in reasoning, coding, math, and Chinese comprehension. It matters now as it offers a multilingual alternative to models like Llama2, potentially advancing research in underrepresented languages and specialized tasks, but its claims of superiority are based on internal evaluations and lack independent verification, with no formal releases raising questions about readiness. _themes: llm · multilingual · coding · math_ #### [01-ai/Yi](https://github.com/01-ai/Yi) *Jupyter Notebook · ★7,831 · Apache-2.0 · beta · score:0.75 · hot:0.50 · rising:0.56 · durable:0.70 · board:durable · trend:stable* A series of large language models trained from scratch by developers @01-ai **Why it matters.** Yi is a series of large language models trained from scratch by 01-ai, offering base and chat variants for tasks like text generation and conversation. It matters now as an open-source option in the competitive LLM landscape, providing accessible tools for fine-tuning and deployment amid growing demand for customizable AI models, though its lack of a formal release may limit immediate adoption. _themes: llm · fine-tuning · inference · quantization_ #### [facebookresearch/omnilingual-asr](https://github.com/facebookresearch/omnilingual-asr) *Python · ★2,775 · NOASSERTION · beta · score:0.70 · hot:0.49 · rising:0.54 · durable:0.70 · board:durable · trend:stable* Omnilingual ASR Open-Source Multilingual SpeechRecognition for 1600+ Languages **Why it matters.** Omnilingual ASR is an open-source multilingual speech recognition model that supports over 1600 languages, using zero-shot learning to add new ones with minimal data, making it accessible for underrepresented languages. It matters now because it addresses the growing need for inclusive AI in global applications, such as research and community tools, but its accuracy for low-resource languages remains unproven and the unclear licensing could hinder adoption. The recent updates in December 2025 introduce improved models, signaling ongoing development amid increasing demand for multilingual tech. _themes: asr · multilingual · zero-shot · fine-tuning_ #### [deepseek-ai/DeepSeek-Coder-V2](https://github.com/deepseek-ai/DeepSeek-Coder-V2) *? · ★6,629 · MIT · beta · score:0.75 · hot:0.49 · rising:0.56 · durable:0.68 · board:durable · trend:stable* DeepSeek-Coder-V2: Breaking the Barrier of Closed-Source Models in Code Intelligence **Why it matters.** DeepSeek-Coder-V2 is an open-source Mixture-of-Experts language model specialized in code generation, mathematical reasoning, and general tasks, achieving performance comparable to closed-source models like GPT4-Turbo while supporting 338 programming languages and a 128K context length. It matters now because it provides a high-quality, accessible alternative to proprietary AI tools, potentially accelerating innovation in code intelligence amid growing demand for efficient, open-source solutions in software development. _themes: code-generation · llm · moe · inference_ #### [allenai/molmo2](https://github.com/allenai/molmo2) *Python · ★502 · Apache-2.0 · beta · score:0.70 · hot:0.49 · rising:0.52 · durable:0.63 · board:durable · trend:stable* Code for the Molmo2 Vision-Language Model **Why it matters.** This repository provides code for training and using the Molmo2 vision-language model, which focuses on advanced point-driven grounding tasks in images, videos, and multi-image scenarios, building on open-source AI innovations. It matters now because vision-language models are key to multimodal AI advancements, offering accessible tools for researchers to experiment with spatial understanding, though its lack of a formal release and unverified claims of being state-of-the-art warrant cautious adoption amid rapid competition in the field. _themes: vision-language · grounding · inference · fine-tuning_ #### [deepseek-ai/DeepSeek-VL2](https://github.com/deepseek-ai/DeepSeek-VL2) *Python · ★5,263 · MIT · beta · score:0.80 · hot:0.49 · rising:0.55 · durable:0.67 · board:durable · trend:stable* DeepSeek-VL2: Mixture-of-Experts Vision-Language Models for Advanced Multimodal Understanding **Why it matters.** DeepSeek-VL2 is a series of Mixture-of-Experts vision-language models designed for advanced multimodal tasks like visual question answering, optical character recognition, and document understanding, building on its predecessor with improved efficiency and performance. It matters now because it achieves state-of-the-art results with fewer activated parameters, making it a timely advancement in efficient AI for multimodal applications amid growing demand for open-source vision-language models. _themes: moe · vision-language · multimodal · inference_ #### [facebookresearch/ImageBind](https://github.com/facebookresearch/ImageBind) *Python · ★9,017 · NOASSERTION · experimental · score:0.80 · hot:0.49 · rising:0.53 · durable:0.64 · board:durable · trend:stable* ImageBind One Embedding Space to Bind Them All **Why it matters.** ImageBind is a PyTorch model that learns a joint embedding space for six modalities—images, text, audio, depth, thermal, and IMU data—enabling cross-modal tasks like retrieval and generation without additional training. It matters now because multimodal AI is increasingly important for applications in robotics, AR/VR, and content analysis, building on recent advances in foundation models to handle real-world data integration more effectively. _themes: multimodal · embedding · zero-shot · computer-vision_ #### [openai/glide-text2im](https://github.com/openai/glide-text2im) *Python · ★3,692 · MIT · beta · score:0.80 · hot:0.49 · rising:0.55 · durable:0.68 · board:durable · trend:stable* GLIDE: a diffusion-based text-conditional image synthesis model **Why it matters.** GLIDE is an OpenAI-developed diffusion-based model for generating photorealistic images from text prompts, including features like inpainting and classifier-free guidance. It matters now because diffusion models are at the forefront of generative AI advancements, influencing applications in creative fields and research, though this specific repo represents an earlier implementation that may not include the latest improvements. The open-source code enables experimentation but lacks active releases, potentially limiting its immediate practical use. _themes: diffusion · text-to-image · image-generation · generative-ai_ #### [openai/point-e](https://github.com/openai/point-e) *Python · ★6,875 · MIT · experimental · score:0.70 · hot:0.49 · rising:0.54 · durable:0.65 · board:durable · trend:stable* Point cloud diffusion for 3D model synthesis **Why it matters.** The Point-E repository provides code and models for generating 3D point clouds from text or image prompts using diffusion techniques, along with tools for converting point clouds to meshes and evaluating results, as per the associated research paper. It matters for advancing 3D synthesis in AI research, but its limitations—such as the low-quality text-to-3D model and lack of formal releases—mean it's not yet ready for widespread production use, making it more relevant for experimental exploration in a field that's evolving rapidly. _themes: diffusion · 3d-generation · point-cloud · synthesis_ #### [microsoft/TRELLIS.2](https://github.com/microsoft/TRELLIS.2) *Python · ★5,471 · MIT · experimental · score:0.80 · hot:0.48 · rising:0.52 · durable:0.64 · board:durable · trend:stable* Native and Compact Structured Latents for 3D Generation **Why it matters.** TRELLIS.2 is a 4-billion-parameter model for image-to-3D generation that uses a sparse voxel structure called O-Voxel to create high-fidelity 3D assets with complex topologies and full PBR materials, while being efficient in processing. It matters now because 3D generation is a growing area in AI for applications like gaming and AR/VR, and this model addresses key limitations in existing methods by providing compact latents and better handling of arbitrary structures, potentially accelerating adoption in research and development. _themes: 3d-generation · generative-ai · voxels · image-to-3d_ #### [deepseek-ai/DeepSeek-V2](https://github.com/deepseek-ai/DeepSeek-V2) *? · ★5,006 · MIT · beta · score:0.80 · hot:0.48 · rising:0.53 · durable:0.67 · board:durable · trend:stable* DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model **Why it matters.** DeepSeek-V2 is a Mixture-of-Experts language model with 236B total parameters, where only 21B are activated per token, designed to deliver strong performance while reducing training costs by 42.5% and improving inference efficiency. It matters now because the AI community is increasingly focused on cost-effective and scalable models amid rising computational demands, potentially making advanced AI more accessible for research and deployment. _themes: moe · language-model · efficiency · inference_ #### [google-deepmind/alphageometry](https://github.com/google-deepmind/alphageometry) *Python · ★4,826 · Apache-2.0 · experimental · score:0.90 · hot:0.48 · rising:0.52 · durable:0.63 · board:durable · trend:stable* **Why it matters.** This repository provides code to reproduce AlphaGeometry and DDAR, AI systems that solve geometry problems at the level of math Olympiads without relying on human demonstrations, as detailed in a 2024 Nature paper. It matters right now because it advances AI reasoning in mathematics, potentially influencing fields like education and automated proof systems, especially with the recent update to AlphaGeometry2 in 2026 highlighting ongoing progress in this area. _themes: ai-reasoning · theorem-proving · deep-learning · geometry_ #### [huggingface/distil-whisper](https://github.com/huggingface/distil-whisper) *Python · ★4,069 · MIT · beta · score:0.85 · hot:0.48 · rising:0.53 · durable:0.66 · board:durable · trend:stable* Distilled variant of Whisper for speech recognition. 6x faster, 50% smaller, within 1% word error rate. **Why it matters.** Distil-Whisper is a distilled version of the OpenAI Whisper model for English speech recognition, reducing model size by about 50% and increasing speed by up to 6x while keeping word error rates within 1% of the original. This matters now because efficient AI models are essential for real-time applications on devices with limited resources, such as mobiles or edge computing, amid growing demands for accessible speech tech in everyday use. _themes: speech-recognition · model-distillation · inference · audio_ #### [facebookresearch/large_concept_model](https://github.com/facebookresearch/large_concept_model) *Python · ★2,351 · MIT · experimental · score:0.70 · hot:0.48 · rising:0.51 · durable:0.64 · board:durable · trend:stable* Large Concept Models: Language modeling in a sentence representation space **Why it matters.** This repository implements Large Concept Models (LCM), which are sequence-to-sequence models that perform language modeling in a sentence representation space using concepts from the SONAR embedding, supporting up to 200 languages. It explores approaches like MSE regression and diffusion-based generation for auto-regressive sentence prediction, trained on massive datasets like 1.3T tokens with 1.6B parameter models. While it advances multilingual and multimodal AI by focusing on higher-level semantics, its experimental status with no official release limits immediate applicability for production use. _themes: language-models · nlp · sequence-to-sequence · multilingual_ #### [allenai/molmoact](https://github.com/allenai/molmoact) *Python · ★327 · Apache-2.0 · beta · score:0.70 · hot:0.48 · rising:0.50 · durable:0.61 · board:durable · trend:down* Official Repository for MolmoAct **Why it matters.** This repository provides the official code for MolmoAct, a vision-language model from AllenAI, including tools for training, fine-tuning, evaluation, and data processing, enabling users to replicate and experiment with the model's pipeline. It matters now because it offers fully open-source resources for advancing multimodal AI research, especially with recent releases of datasets and models, though the future-dated updates in the README suggest potential inconsistencies or ongoing development that could affect reliability. _themes: vision-language · fine-tuning · inference · evaluation_ #### [facebookresearch/sam-audio](https://github.com/facebookresearch/sam-audio) *Python · ★3,461 · NOASSERTION · experimental · score:0.80 · hot:0.48 · rising:0.50 · durable:0.62 · board:durable · trend:stable* The repository provides code for running inference with the Meta Segment Anything Audio Model (SAM-Audio), links for downloading the trained model checkpoints, and example notebooks that show how to use the model. **Why it matters.** SAM-Audio is a foundation model from Meta that segments and isolates specific sounds in audio using text, visual, or temporal prompts, building on multimodal techniques for audio processing. It matters now because it advances AI applications in areas like content editing and accessibility, amid growing interest in multimodal models, but its experimental nature and dependency on restricted checkpoints may hinder widespread use. _themes: audio · segmentation · multimodal · inference_ #### [QwenLM/Qwen3-Omni](https://github.com/QwenLM/Qwen3-Omni) *Jupyter Notebook · ★3,688 · Apache-2.0 · beta · score:0.70 · hot:0.47 · rising:0.53 · durable:0.68 · board:durable · trend:stable* Qwen3-omni is a natively end-to-end, omni-modal LLM developed by the Qwen team at Alibaba Cloud, capable of understanding text, audio, images, and video, as well as generating speech in real time. **Why it matters.** Qwen3-Omni is an end-to-end multilingual omni-modal large language model that processes text, images, audio, and video inputs while generating real-time text and speech outputs, making it suitable for integrated AI applications. It matters now because it advances multimodal AI capabilities from Alibaba Cloud, potentially competing with models like GPT-4o in a rapidly evolving field where seamless handling of diverse data types is becoming essential for real-world use cases. _themes: multimodal · llm · inference · speech_ #### [deepseek-ai/DeepSeek-Math](https://github.com/deepseek-ai/DeepSeek-Math) *Python · ★3,237 · MIT · experimental · score:0.70 · hot:0.47 · rising:0.52 · durable:0.63 · board:durable · trend:stable* DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models **Why it matters.** DeepSeek-Math is an open-source language model fine-tuned for mathematical reasoning, based on DeepSeek-Coder and trained on math-related data, achieving 51.7% on the MATH benchmark without external tools. It matters because it represents a step forward in accessible AI for specialized tasks like math problem-solving, potentially challenging proprietary models, though its lack of a formal release limits immediate applicability. This could support research in AI reasoning amid growing demand for reliable math capabilities in education and automation. _themes: fine-tuning · mathematical-reasoning · llm · inference_ #### [facebookresearch/jepa](https://github.com/facebookresearch/jepa) *Python · ★3,725 · NOASSERTION · experimental · score:0.75 · hot:0.47 · rising:0.50 · durable:0.60 · board:durable · trend:stable* PyTorch code and models for V-JEPA self-supervised learning from video. **Why it matters.** V-JEPA is a PyTorch implementation for self-supervised learning from video, using a joint embedding predictive architecture to learn visual representations without labels or pretraining. It matters now because it advances efficient video understanding in AI, potentially reducing reliance on annotated data for applications like robotics and multimedia, and builds on ongoing research from Meta AI led by experts like Yann LeCun. _themes: self-supervised · video-learning · pytorch · feature-prediction_ #### [deepseek-ai/DeepSeek-VL](https://github.com/deepseek-ai/DeepSeek-VL) *Python · ★4,088 · MIT · beta · score:0.75 · hot:0.47 · rising:0.53 · durable:0.67 · board:durable · trend:stable* DeepSeek-VL: Towards Real-World Vision-Language Understanding **Why it matters.** DeepSeek-VL is an open-source vision-language model that processes diverse real-world inputs like diagrams, web pages, and scientific literature to enable multimodal understanding, with variants in 7B and 1.3B parameters for base and chat applications. It matters now because vision-language models are increasingly critical for AI advancements in areas like embodied AI and scientific analysis, offering accessible alternatives to proprietary models amid growing demand for open-source tools, though its lack of a formal release may limit immediate adoption. _themes: vision-language · multimodal · pretraining · inference_ #### [deepseek-ai/DreamCraft3D](https://github.com/deepseek-ai/DreamCraft3D) *Python · ★3,006 · MIT · experimental · score:0.70 · hot:0.47 · rising:0.52 · durable:0.63 · board:durable · trend:stable* [ICLR 2024] Official implementation of DreamCraft3D: Hierarchical 3D Generation with Bootstrapped Diffusion Prior **Why it matters.** DreamCraft3D is a Python-based implementation for generating hierarchical 3D objects from 2D images, using techniques like score distillation sampling and bootstrapped diffusion priors to enhance geometric and textural consistency. It addresses key challenges in 3D generation, such as view consistency, which is relevant amid growing interest in AI-driven content creation for applications like gaming and AR, though its reliance on experimental methods may limit immediate practical adoption. _themes: diffusion-models · 3d-generation · image-to-3d · generative-ai_ #### [facebookresearch/sam-3d-body](https://github.com/facebookresearch/sam-3d-body) *Python · ★2,831 · NOASSERTION · experimental · score:0.80 · hot:0.47 · rising:0.49 · durable:0.60 · board:durable · trend:stable* The repository provides code for running inference with the SAM 3D Body Model (3DB), links for downloading the trained model checkpoints and datasets, and example notebooks that show how to use the model. **Why it matters.** This repository provides code and models for SAM 3D Body, a promptable AI model that performs single-image full-body 3D human mesh recovery, including pose estimation for body, feet, and hands using the Momentum Human Rig. It matters now because accurate 3D human reconstruction is increasingly important for applications in AR/VR, gaming, and animation, and this model claims state-of-the-art performance with strong generalization, though its lack of a formal release and unspecified license may limit immediate adoption. _themes: computer-vision · 3d-reconstruction · human-pose · inference_ #### [deepseek-ai/DeepSeek-Math-V2](https://github.com/deepseek-ai/DeepSeek-Math-V2) *Python · ★1,575 · Apache-2.0 · experimental · score:0.75 · hot:0.46 · rising:0.50 · durable:0.64 · board:durable · trend:down* **Why it matters.** DeepSeek-Math-V2 is a language model specialized in mathematical reasoning, using reinforcement learning to improve accuracy on problems from competitions like AIME and HMMT, though it faces potential limitations in scaling. It matters now because advancements in AI for math could enhance scientific research and education, but the lack of a formal release and incomplete documentation raises questions about its immediate applicability amid rapid LLM developments. _themes: math-reasoning · llm · reinforcement-learning · fine-tuning_ #### [QwenLM/Qwen-Image](https://github.com/QwenLM/Qwen-Image) *Python · ★7,791 · Apache-2.0 · beta · score:0.70 · hot:0.46 · rising:0.51 · durable:0.63 · board:durable · trend:stable* Qwen-Image is a powerful image generation foundation model capable of complex text rendering and precise image editing. **Why it matters.** Qwen-Image is a 20B parameter foundation model for text-to-image generation and precise image editing, with a focus on complex text rendering, especially for Chinese. It matters right now because it advances multilingual capabilities in a competitive AI landscape, potentially benefiting applications in content creation and design, though its lack of formal releases may limit immediate adoption. _themes: image-generation · image-editing · text-rendering · multilingual_ #### [facebookresearch/co-tracker](https://github.com/facebookresearch/co-tracker) *Jupyter Notebook · ★4,908 · NOASSERTION · experimental · score:0.75 · hot:0.46 · rising:0.49 · durable:0.59 · board:durable · trend:stable* CoTracker is a model for tracking any point (pixel) on a video. **Why it matters.** CoTracker is a transformer-based model that tracks any pixel in a video, supporting manual or grid-based point selection for applications like video analysis and editing. It matters right now because accurate point tracking advances computer vision research, particularly in handling real-world video data, though its lack of official releases and reliance on experimental demos limits immediate practical adoption. _themes: tracking · optical-flow · transformers · video_ #### [openai/following-instructions-human-feedback](https://github.com/openai/following-instructions-human-feedback) *? · ★1,257 · no-license · experimental · score:0.80 · hot:0.46 · rising:0.50 · durable:0.62 · board:durable · trend:down* **Why it matters.** This repository presents the InstructGPT method, which fine-tunes language models like GPT-3 using supervised learning and reinforcement learning from human feedback to better align with user intentions, reducing issues like untruthfulness and toxicity. It matters now because AI alignment is a critical challenge in deploying large language models safely in applications, and this approach has influenced real-world systems like ChatGPT, highlighting the value of human feedback in model training despite the method's experimental nature and potential limitations in generalizing beyond specific prompts. _themes: fine-tuning · rlhf · alignment · language-models_ #### [QwenLM/Qwen2.5-Omni](https://github.com/QwenLM/Qwen2.5-Omni) *Jupyter Notebook · ★3,979 · Apache-2.0 · beta · score:0.80 · hot:0.46 · rising:0.51 · durable:0.63 · board:durable · trend:stable* Qwen2.5-Omni is an end-to-end multimodal model by Qwen team at Alibaba Cloud, capable of understanding text, audio, vision, video, and performing real-time speech generation. **Why it matters.** Qwen2.5-Omni is an open-source multimodal AI model that processes text, images, audio, and video inputs while supporting real-time speech generation, making it suitable for advanced applications in perception and interaction. It matters now because multimodal models are key to next-generation AI systems, and this model's strong benchmark performance, like topping spoken language understanding tests, highlights its competitiveness in a rapidly evolving field dominated by proprietary alternatives. _themes: multimodal · inference · speech-generation · benchmarks_ #### [deepseek-ai/DeepSeek-Prover-V1.5](https://github.com/deepseek-ai/DeepSeek-Prover-V1.5) *Python · ★562 · MIT · experimental · score:0.75 · hot:0.46 · rising:0.49 · durable:0.60 · board:durable · trend:down* **Why it matters.** DeepSeek-Prover-V1.5 is an open-source language model specialized for theorem proving in Lean 4, building on its predecessor by integrating reinforcement learning from proof assistant feedback and a Monte-Carlo tree search variant for better proof path exploration. It achieves state-of-the-art results on benchmarks like miniF2F and ProofNet, which could advance AI in formal mathematics, but its lack of a formal release and limited evaluation details raise questions about immediate practical utility. However, as AI models increasingly tackle complex reasoning tasks, this represents a timely step in specialized machine learning for verification. _themes: theorem-proving · reinforcement-learning · mcts · fine-tuning_ #### [QwenLM/Qwen-Image-Layered](https://github.com/QwenLM/Qwen-Image-Layered) *Python · ★1,790 · Apache-2.0 · beta · score:0.70 · hot:0.46 · rising:0.50 · durable:0.64 · board:durable · trend:down* Qwen-Image-Layered: Layered Decomposition for Inherent Editablity **Why it matters.** This repository introduces Qwen-Image-Layered, a model that decomposes images into multiple RGBA layers to enable independent editing of elements like resizing, repositioning, and recoloring without affecting other parts. It matters because it addresses the growing need for precise, non-destructive image manipulation in AI applications, potentially improving workflows in content creation and editing amid advancements in generative AI. However, its real-world utility depends on further validation beyond the initial release. _themes: image-editing · ai-vision · decomposition · inference_ #### [QwenLM/Qwen-VL](https://github.com/QwenLM/Qwen-VL) *Python · ★6,624 · NOASSERTION · beta · score:0.70 · hot:0.45 · rising:0.49 · durable:0.60 · board:durable · trend:stable* The official repo of Qwen-VL (通义千问-VL) chat & pretrained large vision language model proposed by Alibaba Cloud. **Why it matters.** Qwen-VL is a large vision-language model from Alibaba Cloud that combines visual and textual processing for tasks like chat and image understanding, building on pretrained foundations. It matters now due to the growing demand for multimodal AI in applications such as visual question answering and content generation, with Alibaba's backing providing a competitive open-source alternative in a rapidly evolving field dominated by Western models. However, its lack of a formal release and unclear licensing may limit immediate adoption. _themes: multimodal · vision-language · inference · large-language-models_ #### [deepseek-ai/DeepSeek-MoE](https://github.com/deepseek-ai/DeepSeek-MoE) *Python · ★1,909 · MIT · experimental · score:0.75 · hot:0.45 · rising:0.50 · durable:0.63 · board:durable · trend:down* DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Models **Why it matters.** DeepSeekMoE is a Mixture-of-Experts language model with 16.4B parameters that optimizes computation through expert specialization, achieving performance comparable to denser models like DeepSeek 7B and LLaMA2 7B using only about 40% of the computations. It matters now because it addresses the critical need for efficient AI models amid rising hardware costs and environmental concerns, potentially accelerating research in scalable language processing. However, its experimental nature and lack of official releases mean it's more of a proof-of-concept than a ready-to-use solution. _themes: moe · llm · efficiency · inference_ #### [deepseek-ai/DeepSeek-OCR-2](https://github.com/deepseek-ai/DeepSeek-OCR-2) *Python · ★2,713 · Apache-2.0 · experimental · score:0.65 · hot:0.45 · rising:0.48 · durable:0.60 · board:durable · trend:down* Visual Causal Flow **Why it matters.** DeepSeek-OCR-2 is an open-source OCR model focused on text extraction from images and PDFs, utilizing advanced vision-language techniques as detailed in its associated paper. It matters right now because the growing demand for accurate document processing in AI applications, such as automated workflows and data extraction, makes efficient OCR tools essential, and this repo provides inference setups with popular frameworks like vLLM. _themes: ocr · inference · vision-ai · document-processing_ #### [facebookresearch/ShapeR](https://github.com/facebookresearch/ShapeR) *Python · ★753 · NOASSERTION · experimental · score:0.70 · hot:0.45 · rising:0.47 · durable:0.60 · board:durable · trend:down* Code for the ShapeR research paper **Why it matters.** ShapeR is a research tool that generates 3D meshes from casual image sequences using multimodal inputs like SLAM points, images, poses, and captions via a transformer-based model, requiring significant preprocessing for accurate metric reconstruction. It matters now because it advances 3D scene understanding in computer vision applications, such as AR/VR and robotics, but its reliance on off-the-shelf methods and lack of a stable release limits immediate practical adoption. _themes: 3d-reconstruction · multimodal-ai · shape-generation · computer-vision_ #### [microsoft/MoGe](https://github.com/microsoft/MoGe) *Python · ★2,417 · NOASSERTION · beta · score:0.75 · hot:0.45 · rising:0.48 · durable:0.59 · board:durable · trend:stable* [CVPR'25 Oral] MoGe: Unlocking Accurate Monocular Geometry Estimation for Open-Domain Images with Optimal Training Supervision **Why it matters.** MoGe is a model that estimates 3D geometry, including metric depth maps, point maps, and normal maps, from single monocular images using optimized training supervision, all in one forward pass. It matters now because it addresses challenges in open-domain computer vision applications like AR/VR and autonomous driving, where accurate monocular depth estimation is critical, and the recent MoGe-2 update improves scale accuracy and detail sharpness. However, its lack of a formal release and unclear licensing may limit immediate adoption. _themes: 3d-vision · depth-estimation · monocular · inference_ #### [QwenLM/Qwen3-ASR](https://github.com/QwenLM/Qwen3-ASR) *Python · ★2,452 · Apache-2.0 · beta · score:0.70 · hot:0.45 · rising:0.49 · durable:0.63 · board:durable · trend:down* Qwen3-ASR is an open-source series of ASR models developed by the Qwen team at Alibaba Cloud, supporting stable multilingual speech/music/song recognition, language detection and timestamp prediction. **Why it matters.** Qwen3-ASR provides open-source ASR models for multilingual speech recognition across 52 languages, including features like language detection, timestamp prediction, and forced alignment for text-speech pairs in 11 languages, making it useful for applications in transcription and multimedia processing. It matters now because the demand for accurate, multilingual voice AI is growing with global digital content, offering accessible tools from Alibaba Cloud that could enhance real-time applications, though its novelty is tempered by existing competitors and the lack of a formal release. _themes: asr · multilingual · speech-recognition · alignment_ #### [allenai/longformer](https://github.com/allenai/longformer) *Python · ★2,193 · Apache-2.0 · beta · score:0.80 · hot:0.45 · rising:0.49 · durable:0.60 · board:durable · trend:stable* Longformer: The Long-Document Transformer **Why it matters.** Longformer is a pretrained transformer model that extends sequence lengths to handle long documents, up to 16K tokens with specific hardware, addressing limitations in standard transformers for tasks like summarization and question answering. It matters now because processing long texts is essential for real-world NLP applications amid growing data volumes, and its integration with Hugging Face makes it practical for adaptation. However, it hasn't seen a formal release since 2020, potentially indicating limited ongoing maintenance. _themes: nlp · transformer · long-context · summarization_ #### [facebookresearch/nwm](https://github.com/facebookresearch/nwm) *Python · ★594 · NOASSERTION · experimental · score:0.70 · hot:0.45 · rising:0.46 · durable:0.57 · board:durable · trend:down* Official code for the CVPR 2025 paper "Navigation World Models". **Why it matters.** This repository provides the official PyTorch implementation for training Navigation World Models, a Conditional Diffusion Transformer (CDiT) for simulating and navigating environments based on visual inputs, as presented in a CVPR 2025 paper. It matters now because world models are increasingly important for advancing AI in robotics and autonomous systems, offering potential improvements in efficiency and accuracy, though its experimental nature and lack of a formal release mean it's still unproven for practical applications. _themes: navigation · world-models · diffusion · transformers_ #### [NVIDIA/audio-flamingo](https://github.com/NVIDIA/audio-flamingo) *? · ★1,090 · no-license · experimental · score:0.60 · hot:0.45 · rising:0.45 · durable:0.55 · board:durable · trend:down* PyTorch implementation of Audio Flamingo: Series of Advanced Audio Understanding Language Models **Why it matters.** This repository contains PyTorch implementations of the Audio Flamingo series, which are advanced language models for tasks like audio captioning, question answering, and reasoning, including extensions for music understanding. It matters because it advances multimodal AI in audio processing, potentially influencing research in areas like long-form audio analysis, but its lack of a license, code releases, or clear implementation details makes it less accessible and more of a conceptual showcase based on academic papers. _themes: audio-language-models · multimodal-ai · audio-reasoning · few-shot-learning_ #### [google-research/population-dynamics](https://github.com/google-research/population-dynamics) *Jupyter Notebook · ★400 · Apache-2.0 · experimental · score:0.70 · hot:0.45 · rising:0.48 · durable:0.61 · board:durable · trend:down* PDFM Embeddings: location-based vectors for geo-spatial analysis. **Why it matters.** This repository offers location-based embeddings generated via a Graph Neural Network using aggregated data like search trends and environmental factors to analyze population dynamics while preserving privacy. It matters now because geospatial analysis is increasingly critical for applications in public health and socioeconomic modeling amid growing data privacy concerns and urban challenges. However, its experimental nature and lack of a formal release mean it's more of a proof-of-concept than a ready-to-use tool. _themes: gnn · embeddings · geospatial · privacy_ #### [facebookresearch/cwm](https://github.com/facebookresearch/cwm) *Python · ★862 · NOASSERTION · experimental · score:0.70 · hot:0.45 · rising:0.46 · durable:0.60 · board:durable · trend:down* Research code artifacts for Code World Model (CWM) including inference tools, reproducibility, and documentation. **Why it matters.** Code World Model (CWM) is a 32-billion-parameter LLM specialized in code generation and reasoning, trained on Python execution traces and agent interactions to better model code's effects on system states. It matters now as it addresses gaps in verifiable coding and multi-turn tasks, potentially enhancing AI for software engineering amid increasing focus on reliable code models, though its research-oriented nature means real-world adoption depends on further validation. The repo provides inference tools and weights for experimentation, but lacks a formal release or clear licensing, which may limit immediate usability. _themes: code-generation · llm · inference · reasoning_ #### [google-deepmind/searchless_chess](https://github.com/google-deepmind/searchless_chess) *Python · ★621 · Apache-2.0 · experimental · score:0.70 · hot:0.45 · rising:0.47 · durable:0.61 · board:durable · trend:down* Grandmaster-Level Chess Without Search **Why it matters.** This repository implements a transformer-based model for playing chess at grandmaster level without using search algorithms, trained via supervised learning on a large dataset of annotated chess games. It matters now because it highlights the potential of large-scale transformers for complex planning tasks, offering insights into AI generalization that could influence ongoing research in decision-making systems, especially amid the rapid evolution of transformer architectures in AI. However, its reliance on a specific domain like chess limits broader applicability without further validation. _themes: transformers · planning · chess-ai · supervised-learning_ #### [google-deepmind/ithaca](https://github.com/google-deepmind/ithaca) *Jupyter Notebook · ★577 · Apache-2.0 · experimental · score:0.70 · hot:0.45 · rising:0.48 · durable:0.60 · board:durable · trend:down* Restoring and attributing ancient texts using deep neural networks **Why it matters.** Ithaca is a deep neural network designed to restore damaged ancient Greek inscriptions, attribute them to their original locations, and estimate their dates, thereby assisting historians in epigraphy. It matters now because it showcases AI's potential to enhance historical research accuracy and efficiency, addressing challenges in cultural heritage preservation amid growing interest in AI-human collaborations. This tool could influence interdisciplinary fields by providing a model for applying machine learning to humanities. _themes: deep-learning · nlp · text-restoration · historical-analysis_ #### [deepseek-ai/DeepSeek-V3.2-Exp](https://github.com/deepseek-ai/DeepSeek-V3.2-Exp) *Python · ★1,561 · MIT · experimental · score:0.65 · hot:0.44 · rising:0.47 · durable:0.59 · board:durable · trend:down* **Why it matters.** DeepSeek-V3.2-Exp is an experimental AI model that introduces a sparse attention mechanism to improve efficiency in handling long text sequences during training and inference, while maintaining performance comparable to its predecessor. It matters now because optimizing computational resources for extended contexts is increasingly critical for real-world AI applications dealing with large datasets, and this release provides a research-validated step toward more efficient transformer architectures. However, as an experimental version, it primarily serves as a proof-of-concept rather than a ready-to-use solution. _themes: sparse-attention · efficiency · inference · transformers_ #### [deepseek-ai/DeepSeek-Prover-V2](https://github.com/deepseek-ai/DeepSeek-Prover-V2) *? · ★1,257 · NOASSERTION · experimental · score:0.60 · hot:0.44 · rising:0.46 · durable:0.57 · board:durable · trend:down* **Why it matters.** DeepSeek-Prover-V2 appears to be an AI model or extension focused on automated theorem proving, building on DeepSeek's prior language models, as indicated by references to ProverBench and model summaries in the README. It matters now due to growing interest in AI for formal verification and mathematical reasoning, but its unclear licensing, lack of releases, and minimal documentation raise concerns about accessibility and reliability for practical use. _themes: theorem-proving · ai-reasoning · formal-verification · language-model_ #### [google-research/inksight](https://github.com/google-research/inksight) *Jupyter Notebook · ★979 · Apache-2.0 · experimental · score:0.70 · hot:0.44 · rising:0.47 · durable:0.61 · board:durable · trend:down* **Why it matters.** InkSight is a system that converts images of handwritten text into digital ink using a Vision Transformer and mT5 architecture, supporting both word-level and full-page processing without specialized hardware. It matters now because accurate handwriting digitization addresses growing needs in document archiving and accessibility, though its experimental nature limits immediate practical adoption compared to established OCR tools. _themes: vision · handwriting · ocr · inference_ #### [allenai/scibert](https://github.com/allenai/scibert) *Python · ★1,690 · Apache-2.0 · production · score:0.75 · hot:0.44 · rising:0.49 · durable:0.60 · board:durable · trend:stable* A BERT model for scientific text. **Why it matters.** SciBERT is a BERT model fine-tuned on a large corpus of scientific papers, providing improved performance for NLP tasks in scientific domains compared to general-purpose models. It matters now because specialized language models are increasingly essential for handling the growing volume of academic literature, though its relevance is somewhat diminished by newer advancements in transformer-based architectures. However, it remains a solid baseline for researchers working on scientific text analysis. _themes: bert · nlp · scientific-text · fine-tuning_ #### [QwenLM/Qwen3-Embedding](https://github.com/QwenLM/Qwen3-Embedding) *Python · ★1,904 · no-license · beta · score:0.80 · hot:0.44 · rising:0.47 · durable:0.59 · board:durable · trend:down* **Why it matters.** Qwen3-Embedding is a series of proprietary text embedding and reranking models in sizes like 0.6B, 4B, and 8B, designed for tasks such as text retrieval, classification, clustering, and multilingual processing, building on the Qwen3 family. It matters now because it achieves state-of-the-art performance, topping benchmarks like MTEB with a score of 70.58 as of June 2025, which could advance applications in AI-driven search and ranking amid growing demand for efficient multilingual embeddings. _themes: embeddings · multilingual · retrieval · ranking_ #### [allenai/OLMoE](https://github.com/allenai/OLMoE) *Jupyter Notebook · ★1,009 · Apache-2.0 · beta · score:0.70 · hot:0.44 · rising:0.48 · durable:0.61 · board:durable · trend:down* OLMoE: Open Mixture-of-Experts Language Models **Why it matters.** This repository provides resources for OLMoE, an open-source mixture-of-experts language model, including code, checkpoints, and data for pretraining, fine-tuning, and evaluation, as detailed in a recent paper. It matters now because it offers accessible tools for researchers to explore efficient scaling techniques in LLMs, potentially addressing computational efficiency challenges in AI development, though its lack of formal releases limits immediate practical application. _themes: moe · llm · pretraining · fine-tuning_ #### [QwenLM/Qwen-Audio](https://github.com/QwenLM/Qwen-Audio) *Python · ★1,891 · NOASSERTION · experimental · score:0.70 · hot:0.44 · rising:0.45 · durable:0.55 · board:durable · trend:down* The official repo of Qwen-Audio (通义千问-Audio) chat & pretrained large audio language model proposed by Alibaba Cloud. **Why it matters.** Qwen-Audio is a large audio language model that processes diverse audio inputs like speech and music along with text to generate text outputs, supporting tasks such as audio understanding and multi-turn dialogues. It matters now because the growing interest in multimodal AI for real-world applications like voice interfaces and content analysis highlights the need for versatile models, and this repo from Alibaba Cloud provides an open-source option for researchers to build upon, though its lack of formal releases may limit immediate adoption. _themes: audio · multimodal · language-model · fine-tuning_ #### [QwenLM/Qwen2-Audio](https://github.com/QwenLM/Qwen2-Audio) *Python · ★2,063 · no-license · experimental · score:0.70 · hot:0.44 · rising:0.46 · durable:0.55 · board:durable · trend:down* The official repo of Qwen2-Audio chat & pretrained large audio language model proposed by Alibaba Cloud. **Why it matters.** Qwen2-Audio is a large audio language model from Alibaba Cloud that processes audio inputs for voice chat and analysis, allowing users to interact via speech or provide audio for instructions. It matters now as multimodal AI gains traction for applications like voice assistants, but its lack of a license and formal releases raises barriers to adoption and verification. Overall, it's a solid research effort but not yet polished for practical use. _themes: audio · llm · multimodal · inference_ #### [google-deepmind/kinetics-i3d](https://github.com/google-deepmind/kinetics-i3d) *Python · ★1,833 · Apache-2.0 · archived · score:0.60 · hot:0.43 · rising:0.45 · durable:0.50 · board:durable · trend:stable* Convolutional neural network model for video classification trained on the Kinetics dataset. **Why it matters.** This repository provides pre-trained I3D convolutional neural network models for video classification, specifically trained on the Kinetics dataset, allowing users to fine-tune them for tasks like action recognition on datasets such as UCF101 or HMDB51. It matters now as a foundational resource for computer vision research, offering a benchmark for video models, though its 2017 origins mean it's less competitive against modern architectures like transformers, making it more relevant for educational or baseline comparisons rather than cutting-edge applications. _themes: video-classification · cnn · kinetics · fine-tuning_ #### [QwenLM/Qwen2.5-Math](https://github.com/QwenLM/Qwen2.5-Math) *Python · ★1,075 · no-license · beta · score:0.70 · hot:0.43 · rising:0.46 · durable:0.56 · board:durable · trend:down* A series of math-specific large language models of our Qwen2 series. **Why it matters.** Qwen2.5-Math is a series of large language models specialized for mathematical reasoning, supporting English and Chinese with techniques like Chain-of-Thought and Tool-integrated Reasoning, building on the Qwen2 family. It matters now because it addresses the growing need for accurate AI in math-intensive fields like education and research, where general LLMs often fall short, though its lack of a specified license could limit broader adoption. _themes: llm · mathematics · fine-tuning · reasoning_ #### [EleutherAI/math-lm](https://github.com/EleutherAI/math-lm) *Python · ★1,098 · MIT · experimental · score:0.70 · hot:0.43 · rising:0.46 · durable:0.59 · board:durable · trend:down* **Why it matters.** This repository provides open-source code, data, and pre-trained models for Llemma, a language model specialized in mathematical reasoning and theorem proving, including datasets like Proof-Pile-2 and training scripts. It matters for advancing AI in mathematics, as it enables researchers to experiment with math-focused LLMs, but its utility is limited by the lack of formal releases and reliance on a single paper, making it more of a niche resource than a broadly applicable tool. _themes: mathematics · language-models · fine-tuning · evaluation_ #### [facebookresearch/EdgeTAM](https://github.com/facebookresearch/EdgeTAM) *Jupyter Notebook · ★912 · Apache-2.0 · experimental · score:0.70 · hot:0.43 · rising:0.45 · durable:0.57 · board:durable · trend:down* [CVPR 2025] Official PyTorch implementation of "EdgeTAM: On-Device Track Anything Model" **Why it matters.** EdgeTAM is a PyTorch-based implementation of an on-device variant of the SAM 2 model, designed for efficient promptable segmentation and tracking in videos, achieving 16 FPS on iPhone 15 Pro Max without quantization. It matters now due to the growing demand for lightweight AI models that enable real-time processing on mobile devices for applications like AR/VR, potentially improving privacy and reducing latency, though its experimental nature means it may lack full optimization or robustness for production use. _themes: on-device-ai · segmentation · tracking · inference_ #### [allenai/molmo](https://github.com/allenai/molmo) *Python · ★899 · Apache-2.0 · beta · score:0.70 · hot:0.43 · rising:0.46 · durable:0.58 · board:durable · trend:down* Code for the Molmo Vision-Language Model **Why it matters.** Molmo is an open-source repository providing code, datasets, and tools for training and evaluating vision-language models, extending the OLMo framework with vision encoding capabilities. It matters now because multimodal AI is rapidly evolving, and this release from AllenAI offers accessible resources like the PixMo datasets for researchers to advance VLM development, especially amid growing interest in integrating vision and language for real-world applications. _themes: multimodal · vision-language · fine-tuning · datasets_ #### [google-research/lasertagger](https://github.com/google-research/lasertagger) *Python · ★604 · Apache-2.0 · experimental · score:0.60 · hot:0.43 · rising:0.45 · durable:0.53 · board:durable · trend:down* **Why it matters.** LaserTagger is a BERT-based text-editing model that predicts token-level operations like keep, delete, add phrases, or swap sentences to transform source text into target text, making it useful for precise NLP tasks such as rephrasing or correction. It matters now for applications needing data-efficient and hallucination-resistant editing, especially in research or legacy systems, though its 2019 origins mean it may not incorporate recent advancements in LLMs. _themes: text-editing · bert · fine-tuning · nlp_ #### [google-research/pix2struct](https://github.com/google-research/pix2struct) *Python · ★684 · Apache-2.0 · experimental · score:0.70 · hot:0.42 · rising:0.44 · durable:0.54 · board:durable · trend:down* **Why it matters.** Pix2Struct is a Google Research repository that provides pretrained models and code for screenshot parsing as a pretraining technique to enhance visual language understanding, allowing finetuning on specific tasks like document analysis. It matters now because multimodal AI models are increasingly important for real-world applications such as web and document processing, but its reliance on Google Cloud tools and lack of a formal release may limit immediate adoption due to setup complexity and experimental nature. _themes: vision-language · pretraining · finetuning · multimodal_ #### [stanford-crfm/BioMedLM](https://github.com/stanford-crfm/BioMedLM) *Python · ★640 · no-license · experimental · score:0.70 · hot:0.42 · rising:0.43 · durable:0.52 · board:durable · trend:down* **Why it matters.** BioMedLM provides code for pre-training and fine-tuning a GPT-2 based language model specialized in biomedical literature, allowing researchers to generate and analyze medical text. It matters right now due to the growing demand for domain-specific AI in healthcare, such as accelerating drug discovery and clinical research, though its lack of a license and formal releases may hinder broader adoption. _themes: biomedical · llm · fine-tuning · pre-training_ #### [microsoft/VidTok](https://github.com/microsoft/VidTok) *Python · ★444 · MIT · beta · score:0.65 · hot:0.42 · rising:0.44 · durable:0.61 · board:durable · trend:down* a family of versatile and state-of-the-art video tokenizers. **Why it matters.** VidTok is a family of video tokenizers designed for efficient continuous and discrete tokenization, featuring advancements like reduced computational complexity and improved quantization to handle large-scale video data. It matters right now because efficient video processing is crucial for growing applications in AI-driven video analysis and generation, where it outperforms existing models on key metrics, though its lack of a formal release and moderate adoption (444 stars) suggest it's still emerging. _themes: video-tokenization · efficiency · quantization · computer-vision_ #### [facebookresearch/OrienterNet](https://github.com/facebookresearch/OrienterNet) *Python · ★555 · NOASSERTION · experimental · score:0.70 · hot:0.42 · rising:0.42 · durable:0.51 · board:durable · trend:down* Source Code for Paper "OrienterNet Visual Localization in 2D Public Maps with Neural Matching" **Why it matters.** OrienterNet is a deep learning model for visual localization that matches images to 2D public maps like OpenStreetMap to estimate position and orientation, avoiding the need for complex 3D point clouds. It matters now as accurate, lightweight localization is increasingly vital for AR/VR applications and mobile navigation, leveraging freely available data for broader accessibility and efficiency. _themes: computer-vision · visual-localization · neural-networks · mapping_ #### [01-ai/Yi-Coder](https://github.com/01-ai/Yi-Coder) *HTML · ★445 · no-license · experimental · score:0.60 · hot:0.41 · rising:0.43 · durable:0.54 · board:durable · trend:down* 🌟 Yi-Coder is a series of open-source code language models that delivers state-of-the-art coding performance with fewer than 10 billion parameters. **Why it matters.** Yi-Coder is an open-source series of language models optimized for code generation and understanding, supporting 52 programming languages with up to 128K token context length and under 10 billion parameters, aiming for state-of-the-art performance in a compact size. It matters for developers seeking efficient AI tools for coding tasks amid growing demand for lightweight models, but its lack of a license and no official releases hinder immediate adoption and raise questions about its readiness and legal usability. _themes: code-generation · llm · efficiency · long-context_ #### [google-research/byt5](https://github.com/google-research/byt5) *Python · ★543 · Apache-2.0 · experimental · score:0.65 · hot:0.41 · rising:0.43 · durable:0.56 · board:durable · trend:down* **Why it matters.** ByT5 is a tokenizer-free language model that processes text directly at the byte level using UTF-8, extending the mT5 architecture to eliminate preprocessing steps and handle noisy or spelling-sensitive data more effectively. It matters now because it simplifies NLP pipelines in an era of increasing data complexity, potentially reducing computational overhead, though its advantages are task-specific and may not outperform subword-based models universally. _themes: nlp · byte-level · transformer · fine-tuning_ #### [QwenLM/Qwen3Guard](https://github.com/QwenLM/Qwen3Guard) *Python · ★448 · no-license · beta · score:0.70 · hot:0.41 · rising:0.42 · durable:0.56 · board:durable · trend:down* Qwen3Guard is a multilingual guardrail model series developed by the Qwen team at Alibaba Cloud. **Why it matters.** Qwen3Guard is a series of multilingual AI models for safety moderation, built on Qwen3 and trained on 1.19 million labeled prompts and responses, offering variants for full prompt/response classification and real-time token-level monitoring. It matters now because AI safety is a pressing concern amid increasing regulatory scrutiny and the need to prevent harmful outputs in real-world applications, though its lack of a specified license and release may limit immediate adoption. The models' sizes and features make them practical for developers integrating safety checks into AI systems. _themes: safety · moderation · llm · real-time_ #### [allenai/specter](https://github.com/allenai/specter) *Python · ★576 · Apache-2.0 · beta · score:0.70 · hot:0.41 · rising:0.44 · durable:0.55 · board:durable · trend:down* SPECTER: Document-level Representation Learning using Citation-informed Transformers **Why it matters.** SPECTER is a transformer-based model that learns document-level representations by incorporating citation information, primarily for scientific texts, enabling better tasks like document retrieval and clustering. It matters now because the growing volume of academic literature demands efficient representation learning, and its integration with HuggingFace makes it more accessible for NLP applications in research and development. With over 500 stars and ongoing use, it addresses a niche but important need in AI for knowledge management. _themes: transformers · embeddings · citation-analysis · document-representation_ #### [QwenLM/QwQ](https://github.com/QwenLM/QwQ) *Python · ★516 · Apache-2.0 · experimental · score:0.70 · hot:0.40 · rising:0.42 · durable:0.55 · board:durable · trend:down* QwQ is the reasoning model series developed by Qwen team, Alibaba Cloud. **Why it matters.** QwQ is a series of AI models specialized in reasoning and critical thinking, developed by Alibaba's Qwen team, aimed at improving performance on complex problem-solving tasks compared to standard instruction-tuned models. It competes with models like DeepSeek-R1 and o1-mini, making it relevant for applications needing advanced reasoning capabilities. However, with no official release and potential issues like endless repetitions, its practical value is still unproven and may require further refinement. _themes: reasoning · llm · inference · evaluation_ #### [01-ai/Yi-1.5](https://github.com/01-ai/Yi-1.5) *? · ★559 · Apache-2.0 · beta · score:0.60 · hot:0.39 · rising:0.43 · durable:0.58 · board:durable · trend:down* Yi-1.5 is an upgraded version of Yi, delivering stronger performance in coding, math, reasoning, and instruction-following capability. **Why it matters.** Yi-1.5 is an open-source language model upgrade that improves performance in coding, math, reasoning, and instruction-following through additional pre-training and fine-tuning on large datasets. It matters now as the AI community continues to prioritize specialized models for practical tasks, offering a free alternative amid ongoing advancements in open-source LLMs, though its impact is limited by modest adoption with only 559 stars. _themes: llm · fine-tuning · inference · reasoning_ #### [EleutherAI/polyglot](https://github.com/EleutherAI/polyglot) *? · ★485 · Apache-2.0 · experimental · score:0.70 · hot:0.39 · rising:0.42 · durable:0.57 · board:durable · trend:down* Polyglot: Large Language Models of Well-balanced Competence in Multi-languages **Why it matters.** Polyglot focuses on developing large language models that enhance performance in non-English languages, starting with a Korean model trained on 1.2TB of data, to address the dissatisfaction with existing multilingual models. This matters now because the global push for inclusive AI requires better multilingual capabilities, potentially filling gaps left by models like mBERT and BLOOM, though its impact is limited without formal releases or broader evaluations. It serves as a research tool for comparing monolingual and multilingual approaches in underrepresented languages. _themes: multilingual · llm · non-english · training_ ### sdk (58) #### [BerriAI/litellm](https://github.com/BerriAI/litellm) *Python · ★44,312 · NOASSERTION · production · score:0.90 · hot:0.89 · rising:0.89 · durable:0.79 · board:rising · trend:up* Python SDK, Proxy Server (AI Gateway) to call 100+ LLM APIs in OpenAI (or native) format, with cost tracking, guardrails, loadbalancing and logging. [Bedrock, Azure, OpenAI, VertexAI, Cohere, Anthropic, Sagemaker, HuggingFace, VLLM, NVIDIA NIM] **Why it matters.** LiteLLM is a Python SDK and proxy server that provides a unified interface for calling over 100 LLM APIs from providers like OpenAI, Anthropic, and Azure, simplifying integration with features like cost tracking, load balancing, and guardrails. It matters right now because the rapid proliferation of LLM providers creates complexity in development and operations, and LiteLLM reduces this friction, enabling faster, more scalable AI application building amid growing enterprise adoption of multi-provider strategies. _themes: llm · inference · gateway · load-balancing_ #### [vercel/ai](https://github.com/vercel/ai) *TypeScript · ★23,709 · NOASSERTION · production · score:0.80 · hot:0.87 · rising:0.87 · durable:0.75 · board:hot · trend:up* The AI Toolkit for TypeScript. From the creators of Next.js, the AI SDK is a free open-source library for building AI-powered applications and agents **Why it matters.** The Vercel AI SDK is a TypeScript library that offers a unified API for integrating various AI models from providers like OpenAI and Anthropic, simplifying the development of AI-powered applications across frameworks like Next.js and React. It matters now because the proliferation of AI services requires tools that abstract provider differences, enabling faster prototyping and easier model switching in a competitive landscape, though it still depends on Vercel's ecosystem for full benefits. However, its reliance on external gateways like Vercel AI Gateway could introduce potential latency or dependency issues for users not on Vercel's platform. _themes: generative-ai · llm · agents · inference_ #### [openai/openai-python](https://github.com/openai/openai-python) *Python · ★30,559 · Apache-2.0 · production · score:0.95 · hot:0.87 · rising:0.92 · durable:0.82 · board:rising · trend:up* The official Python library for the OpenAI API **Why it matters.** This repository hosts the official Python library for the OpenAI API, providing easy-to-use synchronous and asynchronous clients to access OpenAI's AI models and services from Python applications. It matters right now because OpenAI's APIs are central to the ongoing AI boom, with developers increasingly building production-grade applications for tasks like text generation and chatbots, and this library ensures reliable, type-safe integration amid rapid advancements in generative AI. _themes: inference · api · llms · chatbots_ #### [microsoft/semantic-kernel](https://github.com/microsoft/semantic-kernel) *C# · ★27,759 · MIT · production · score:0.85 · hot:0.86 · rising:0.90 · durable:0.81 · board:rising · trend:up* Integrate cutting-edge LLM technology quickly and easily into your apps **Why it matters.** Semantic Kernel is an SDK for integrating LLMs into applications, enabling the creation of AI agents and multi-agent systems with features like model flexibility and orchestration. It matters now because enterprises are rapidly adopting AI for workflows, but its multi-language support and reliance on external services could introduce integration challenges and potential vendor lock-in. While it offers enterprise-grade tools, developers should evaluate its performance in real-world scenarios against simpler alternatives. _themes: agents · llm · orchestration · plugins_ #### [CopilotKit/CopilotKit](https://github.com/CopilotKit/CopilotKit) *TypeScript · ★30,326 · MIT · production · score:0.85 · hot:0.84 · rising:0.87 · durable:0.80 · board:rising · trend:up* The Frontend Stack for Agents & Generative UI. React + Angular. Makers of the AG-UI Protocol **Why it matters.** CopilotKit is an SDK for building frontend applications with AI agents, offering features like generative UI, shared state, and human-in-the-loop interactions using React and Angular. It matters now because the growing adoption of AI agents in production environments demands seamless UI integration, and its backing by the AG-UI Protocol, used by companies like Google and LangChain, positions it as a practical tool for enhancing interactive AI experiences. _themes: agents · generative-ui · ai · react_ #### [openai/openai-node](https://github.com/openai/openai-node) *TypeScript · ★10,834 · Apache-2.0 · production · score:0.90 · hot:0.84 · rising:0.88 · durable:0.77 · board:rising · trend:up* Official JavaScript / TypeScript library for the OpenAI API **Why it matters.** This repository provides the official JavaScript and TypeScript library for accessing the OpenAI API, enabling developers to interact with OpenAI's models and services through a generated, convenient interface. It matters right now because the growing demand for AI integration in applications requires reliable tools, and this library ensures seamless compatibility with OpenAI's evolving ecosystem, especially as developers build on recent advancements in LLMs. _themes: api · inference · llm · sdk_ #### [cactus-compute/cactus](https://github.com/cactus-compute/cactus) *C · ★4,668 · NOASSERTION · beta · score:0.75 · hot:0.82 · rising:0.79 · durable:0.72 · board:hot · trend:up* Low-latency AI engine for mobile devices & wearables **Why it matters.** Cactus is an AI inference engine optimized for mobile devices and wearables, focusing on low-latency execution of language, speech, and vision models using ARM CPUs with features like zero-copy memory and NPU acceleration. It matters now because the growing demand for on-device AI in edge computing prioritizes privacy, energy efficiency, and reduced cloud dependency, though its claims of being the 'fastest' need independent verification amid increasing competition in mobile AI frameworks. _themes: edge-ai · mobile-inference · llm-inference · quantization_ #### [anthropics/anthropic-sdk-python](https://github.com/anthropics/anthropic-sdk-python) *Python · ★3,292 · MIT · production · score:0.75 · hot:0.81 · rising:0.84 · durable:0.70 · board:rising · trend:up* **Why it matters.** This repository is the official Python SDK for Anthropic's Claude API, allowing developers to integrate Claude's AI models into their applications with simple API calls. It matters now because Claude is a competitive alternative in the AI landscape, especially for tasks requiring advanced language understanding, amid growing demand for reliable LLMs; however, its value depends on specific use cases and comparisons with established options like OpenAI's offerings. _themes: api · sdk · llm · inference_ #### [traceloop/openllmetry](https://github.com/traceloop/openllmetry) *Python · ★7,030 · Apache-2.0 · beta · score:0.75 · hot:0.80 · rising:0.81 · durable:0.69 · board:rising · trend:up* Open-source observability for your GenAI or LLM application, based on OpenTelemetry **Why it matters.** OpenLLMetry is an open-source extension of OpenTelemetry that provides observability for GenAI and LLM applications, enabling monitoring of metrics, traces, and logs with integrations to tools like Datadog. It matters now because the rapid adoption of LLMs requires robust monitoring to ensure reliability and performance in production environments, and it standardizes observability practices specifically for AI workloads. This helps developers and operators address common issues like model hallucinations or latency without reinventing tools. _themes: llm · observability · monitoring · genai_ #### [anthropics/anthropic-sdk-typescript](https://github.com/anthropics/anthropic-sdk-typescript) *TypeScript · ★1,880 · MIT · beta · score:0.70 · hot:0.79 · rising:0.82 · durable:0.66 · board:rising · trend:up* Access to Anthropic's safety-first language model APIs in TypeScript **Why it matters.** This repository provides an official TypeScript SDK for accessing Anthropic's Claude API, allowing developers to integrate safety-focused language models into their server-side applications with ease. It matters now as AI safety and ethical considerations gain prominence in development, offering a straightforward alternative to other LLM APIs amid growing regulatory pressures and the need for reliable, secure AI integrations. _themes: api · inference · llm · safety_ #### [openai/openai-agents-js](https://github.com/openai/openai-agents-js) *TypeScript · ★2,745 · MIT · beta · score:0.70 · hot:0.79 · rising:0.80 · durable:0.72 · board:rising · trend:up* A lightweight, powerful framework for multi-agent workflows and voice agents **Why it matters.** This SDK offers a framework for building multi-agent workflows and voice agents using JavaScript/TypeScript, featuring tools for agent configuration, handoffs, guardrails, and real-time interactions with OpenAI APIs. It matters now because the growing interest in AI agents for applications like chatbots and autonomous systems requires accessible tools for development, but its early release version suggests potential limitations in stability and full feature maturity. _themes: agents · workflows · realtime · tools_ #### [mistralai/client-python](https://github.com/mistralai/client-python) *Python · ★722 · Apache-2.0 · production · score:0.70 · hot:0.78 · rising:0.81 · durable:0.67 · board:rising · trend:up* Python client library for Mistral AI platform **Why it matters.** This repository offers a Python client library for the Mistral AI platform, facilitating access to their APIs for tasks like chat completions and embeddings, which simplifies integration for developers. It matters now as Mistral AI is an emerging alternative in the competitive AI landscape, providing open-source-friendly options amid growing demand for diverse LLM providers, though its adoption is still niche compared to established players. _themes: api-client · inference · embeddings · llms_ #### [microsoft/presidio](https://github.com/microsoft/presidio) *Python · ★7,657 · MIT · production · score:0.80 · hot:0.78 · rising:0.82 · durable:0.75 · board:rising · trend:up* An open-source framework for detecting, redacting, masking, and anonymizing sensitive data (PII) across text, images, and structured data. Supports NLP, pattern matching, and customizable pipelines. **Why it matters.** Presidio is an open-source SDK that detects, redacts, and anonymizes personally identifiable information (PII) in text, images, and structured data using NLP and pattern matching, with customizable pipelines for specific needs. It matters now due to escalating data privacy regulations like GDPR and CCPA, as well as frequent data breaches, making PII protection essential for compliance and risk management in organizations. _themes: nlp · privacy · anonymization · data-protection_ #### [anthropics/claude-agent-sdk-python](https://github.com/anthropics/claude-agent-sdk-python) *Python · ★6,419 · MIT · beta · score:0.70 · hot:0.77 · rising:0.80 · durable:0.69 · board:rising · trend:up* **Why it matters.** The Claude Agent SDK for Python provides a straightforward interface for developers to interact with Anthropic's Claude AI models, enabling asynchronous queries and agent-building capabilities through a simple Python library. It matters now because it simplifies integration of advanced AI agents into applications amid growing demand for conversational AI tools, though its early version (v0.1.63) suggests potential limitations in stability and features. However, as an official SDK, it offers direct access to Anthropic's ecosystem, which could be valuable for those already invested in Claude. _themes: agents · sdk · inference · llm_ #### [microsoft/Foundry-Local](https://github.com/microsoft/Foundry-Local) *TypeScript · ★2,194 · NOASSERTION · production · score:0.75 · hot:0.77 · rising:0.78 · durable:0.68 · board:rising · trend:up* **Why it matters.** Foundry Local provides an end-to-end solution for running AI models locally on devices, including SDKs for multiple languages, optimized models for chat and speech-to-text, and automatic hardware acceleration via ONNX Runtime, all in a compact package. It matters now due to growing demands for privacy, offline capabilities, and cost savings in AI applications, especially as edge computing rises, though its v1.0.0 status and unclear licensing may pose adoption risks for enterprises. _themes: inference · on-device · hardware-acceleration · local-ai_ #### [openai/openai-dotnet](https://github.com/openai/openai-dotnet) *C# · ★2,570 · MIT · production · score:0.80 · hot:0.76 · rising:0.79 · durable:0.73 · board:rising · trend:up* The official .NET library for the OpenAI API **Why it matters.** This repository offers the official .NET library for interacting with OpenAI's API, allowing developers to easily integrate features like chat completions, embeddings, and image generation into C# applications. It matters right now because the growing adoption of AI in enterprise software demands reliable, platform-specific tools, and this library ensures seamless access for the .NET ecosystem amid OpenAI's rapid updates and expanding capabilities. _themes: api · ai-integration · chat · embeddings_ #### [openai/openai-go](https://github.com/openai/openai-go) *Go · ★3,160 · Apache-2.0 · production · score:0.80 · hot:0.76 · rising:0.79 · durable:0.70 · board:rising · trend:up* The official Go library for the OpenAI API **Why it matters.** This repository provides the official Go library for interacting with the OpenAI API, enabling developers to integrate AI capabilities like text generation into Go applications with ease. It matters right now because the growing adoption of AI in production systems demands reliable, maintained SDKs, and this library ensures compatibility with OpenAI's evolving API, though users must watch for occasional breaking changes as noted in the changelog. _themes: api · ai · inference · go-sdk_ #### [huggingface/huggingface.js](https://github.com/huggingface/huggingface.js) *TypeScript · ★2,389 · MIT · beta · score:0.70 · hot:0.76 · rising:0.77 · durable:0.67 · board:rising · trend:up* Use Hugging Face with JavaScript **Why it matters.** Hugging Face JS provides a set of TypeScript libraries for interacting with the Hugging Face Hub, managing repositories, and performing inferences on machine learning models using various providers. It matters now because it simplifies integrating Hugging Face's extensive model ecosystem into JavaScript applications, enabling developers to build AI features without dealing with low-level API complexities amid the growing demand for web-based AI tools. _themes: inference · api-client · machine-learning · hub_ #### [microsoft/OpenAPI.NET](https://github.com/microsoft/OpenAPI.NET) *C# · ★1,581 · MIT · production · score:0.80 · hot:0.76 · rising:0.79 · durable:0.69 · board:rising · trend:up* The OpenAPI.NET SDK contains a useful object model for OpenAPI documents in .NET along with common serializers to extract raw OpenAPI JSON and YAML documents from the model. **Why it matters.** OpenAPI.NET provides a .NET object model for representing and serializing OpenAPI documents, including readers and writers for JSON and YAML formats across OpenAPI versions 2 and 3. It matters now because OpenAPI is essential for standardizing API definitions in modern development, and the recent v3.5.2 release adds support for OpenAPI 3.2, helping developers integrate evolving API standards in .NET projects. _themes: openapi · api · dotnet · serialization_ #### [microsoft/mcp](https://github.com/microsoft/mcp) *C# · ★2,994 · MIT · production · score:0.75 · hot:0.76 · rising:0.79 · durable:0.70 · board:rising · trend:up* Catalog of official Microsoft MCP (Model Context Protocol) server implementations for AI-powered data access and tool integration **Why it matters.** Microsoft MCP is a protocol that standardizes how AI applications provide context to large language models for accessing data sources and tools, enabling consistent integration across environments. It matters now because the rapid adoption of LLMs in enterprise settings demands interoperable standards to reduce fragmentation and enhance efficiency in AI development. This repo serves as a central hub for building and maintaining MCP server implementations, helping developers avoid duplication. _themes: llm · integration · protocol · ai-tools_ #### [NVIDIA/TensorRT](https://github.com/NVIDIA/TensorRT) *C++ · ★12,914 · Apache-2.0 · production · score:0.90 · hot:0.75 · rising:0.80 · durable:0.74 · board:rising · trend:up* NVIDIA® TensorRT™ is an SDK for high-performance deep learning inference on NVIDIA GPUs. This repository contains the open source components of TensorRT. **Why it matters.** NVIDIA TensorRT is an SDK that optimizes and accelerates deep learning inference on NVIDIA GPUs, providing tools for model deployment with high performance. It matters right now because it's a standard for production-scale AI inference in industries like autonomous driving and healthcare, but users must prepare for upcoming API changes in version 11.0, which could disrupt workflows if not addressed early. _themes: inference · gpu-acceleration · deep-learning · optimization_ #### [microsoft/windows-rs](https://github.com/microsoft/windows-rs) *Rust · ★12,115 · Apache-2.0 · production · score:0.85 · hot:0.75 · rising:0.81 · durable:0.78 · board:rising · trend:up* Rust for Windows **Why it matters.** This repo provides Rust crates that generate bindings for Windows APIs, allowing developers to call any Windows API directly from Rust as if it were native code, based on metadata from Microsoft. It matters now because Rust's focus on safety and performance is increasingly important for systems programming, and this integration enables seamless development of Windows applications in Rust, especially with Microsoft's endorsement amid growing cross-platform demands. As Windows continues to evolve, this tool ensures Rust developers can access the latest APIs without manual effort. _themes: rust · windows · bindings · ffi_ #### [anthropics/anthropic-cli](https://github.com/anthropics/anthropic-cli) *Go · ★311 · MIT · production · score:0.65 · hot:0.75 · rising:0.77 · durable:0.69 · board:rising · trend:up* The CLI for the Claude API **Why it matters.** This repository provides an official command-line interface for Anthropic's Claude API, allowing users to send messages and manage resources directly from the terminal, which simplifies API interactions for developers. It matters now as AI API tools become essential for efficient prototyping and automation in a competitive landscape, but its limited adoption with only 310 stars suggests it may not yet be a go-to solution compared to more established alternatives. _themes: cli · api · inference · ai-models_ #### [NVIDIA/cuda-python](https://github.com/NVIDIA/cuda-python) *Cython · ★3,221 · NOASSERTION · beta · score:0.75 · hot:0.75 · rising:0.76 · durable:0.63 · board:rising · trend:up* CUDA Python: Performance meets Productivity **Why it matters.** CUDA Python provides Python bindings and tools for accessing NVIDIA's CUDA platform, enabling GPU-accelerated computing in Python for tasks like machine learning and scientific simulations. It matters now because the growing demand for high-performance computing in AI and data processing requires accessible GPU tools, and this library bridges the gap between Python's productivity and CUDA's performance without needing deep C++ expertise. _themes: gpu · cuda · bindings · acceleration_ #### [openai/openai-java](https://github.com/openai/openai-java) *Kotlin · ★1,427 · Apache-2.0 · production · score:0.80 · hot:0.75 · rising:0.78 · durable:0.69 · board:rising · trend:up* The official Java library for the OpenAI API **Why it matters.** This repository provides the official Java library for interacting with OpenAI's REST API, enabling developers to integrate AI capabilities like text generation and chat models into Java applications. It matters right now as AI adoption accelerates in enterprise software, and Java remains a dominant language in production environments, making this SDK a practical bridge for developers building scalable AI features amid growing demand for tools like GPT models. _themes: inference · api · sdk · chat_ #### [anthropics/anthropic-sdk-go](https://github.com/anthropics/anthropic-sdk-go) *Go · ★989 · MIT · production · score:0.70 · hot:0.74 · rising:0.76 · durable:0.66 · board:rising · trend:up* Access to Anthropic's safety-first language model APIs via Go **Why it matters.** This repository offers a Go SDK for accessing Anthropic's Claude API, allowing developers to integrate safety-focused AI language models into their applications with ease. It matters now as AI adoption accelerates and developers seek reliable, language-specific tools for production use, especially in Go for scalable backends, amid growing emphasis on AI safety and compliance. _themes: inference · llm · api · go_ #### [anthropics/anthropic-sdk-ruby](https://github.com/anthropics/anthropic-sdk-ruby) *Ruby · ★323 · MIT · production · score:0.70 · hot:0.74 · rising:0.76 · durable:0.67 · board:rising · trend:up* **Why it matters.** This repository provides an official Ruby SDK for interacting with Anthropic's Claude API, enabling developers to integrate AI capabilities like text generation into their Ruby applications. It matters now because the growing adoption of AI models in software development demands reliable SDKs, and this one simplifies access to Claude's advanced features amid increasing competition in the AI space. _themes: api · sdk · llm · inference_ #### [anthropics/claude-agent-sdk-typescript](https://github.com/anthropics/claude-agent-sdk-typescript) *Shell · ★1,305 · no-license · beta · score:0.70 · hot:0.73 · rising:0.73 · durable:0.62 · board:hot · trend:up* **Why it matters.** The Claude Agent SDK is a TypeScript library from Anthropic that allows developers to build autonomous AI agents capable of understanding codebases, editing files, running commands, and executing workflows using Claude's AI capabilities. It matters now because AI agents are increasingly essential for automating complex tasks in software development, and this SDK provides a straightforward way to integrate Anthropic's advanced models, especially as the AI agent paradigm gains traction in enterprise and research applications. However, its proprietary nature and lack of a clear license may limit broader adoption. _themes: agents · sdk · ai-development · code-interaction_ #### [NVIDIA/cuQuantum](https://github.com/NVIDIA/cuQuantum) *Jupyter Notebook · ★475 · BSD-3-Clause · production · score:0.75 · hot:0.73 · rising:0.75 · durable:0.65 · board:rising · trend:up* Home for cuQuantum Python & NVIDIA cuQuantum SDK C++ samples **Why it matters.** The NVIDIA cuQuantum repository provides Python packages and C++ samples for GPU-accelerated quantum computing simulations, including tools for state vector simulations and tensor networks. It matters right now as quantum computing research demands efficient simulation capabilities, and NVIDIA's integration with CUDA offers high-performance alternatives to traditional CPU-based methods, potentially accelerating algorithm development in a field still dominated by experimental hardware. _themes: quantum-computing · cuda · simulation · gpu-acceleration_ #### [NVIDIA/NVFlare](https://github.com/NVIDIA/NVFlare) *Python · ★920 · Apache-2.0 · production · score:0.80 · hot:0.73 · rising:0.76 · durable:0.67 · board:rising · trend:up* NVIDIA Federated Learning Application Runtime Environment **Why it matters.** NVFlare is an open-source Python SDK that enables federated learning by allowing users to adapt existing ML and DL workflows for secure, privacy-preserving collaborations across distributed parties. It supports a range of algorithms and frameworks, making it easier to transition from research to production in scenarios involving sensitive data. This matters now due to increasing global regulations on data privacy and the growing demand for collaborative AI in fields like healthcare and finance, where data sharing is restricted. _themes: federated-learning · privacy · decentralized · ml_ #### [microsoft/qdk](https://github.com/microsoft/qdk) *Rust · ★908 · MIT · beta · score:0.70 · hot:0.72 · rising:0.74 · durable:0.63 · board:rising · trend:up* Microsoft Quantum Development Kit, including the Q# programming language, resource estimator, and Quantum Katas **Why it matters.** The Microsoft Quantum Development Kit provides tools for quantum programming, including the Q# language, a compiler, resource estimator, and educational resources like Quantum Katas, allowing developers to write and simulate quantum algorithms. It matters now as quantum computing research progresses and industries explore its potential for solving complex problems, though its practical applications remain limited due to the early stage of quantum hardware. _themes: quantum · programming · compiler · simulation_ #### [facebookresearch/projectaria_tools](https://github.com/facebookresearch/projectaria_tools) *C++ · ★762 · Apache-2.0 · production · score:0.70 · hot:0.72 · rising:0.75 · durable:0.65 · board:rising · trend:up* projectaria_tools is an C++/Python open-source toolkit to interact with Project Aria data. **Why it matters.** Project Aria Tools is an open-source C++/Python toolkit that provides utilities for processing and interacting with Meta's Project Aria device data, supporting both Gen1 and Gen2 hardware for AR, machine perception, and AI research. It matters now because the recent Gen2 release includes advanced sensors and on-device algorithms, enabling researchers to efficiently work with enhanced datasets for real-world applications in augmented reality. However, its niche focus on Meta's ecosystem limits broader appeal, requiring users to have access to specific hardware. _themes: ar · perception · ai · sensors_ #### [NVIDIA/cuda-quantum](https://github.com/NVIDIA/cuda-quantum) *C++ · ★1,006 · Apache-2.0 · beta · score:0.60 · hot:0.72 · rising:0.74 · durable:0.62 · board:rising · trend:up* C++ and Python support for the CUDA Quantum programming model for heterogeneous quantum-classical workflows **Why it matters.** CUDA-Quantum offers C++ and Python tools for developing hybrid quantum-classical workflows, including compilers and runtimes that integrate quantum processing units with GPUs and CPUs. This matters now because quantum computing is evolving quickly, and NVIDIA's involvement could standardize hybrid computing practices, though its adoption is still limited by the niche quantum hardware landscape. However, it provides practical backends for testing, which is critical for advancing quantum algorithms in research. _themes: quantum · hybrid · programming · gpu_ #### [openai/openai-ruby](https://github.com/openai/openai-ruby) *Ruby · ★422 · Apache-2.0 · production · score:0.70 · hot:0.72 · rising:0.74 · durable:0.67 · board:rising · trend:up* Official Ruby SDK for the OpenAI API **Why it matters.** This repository provides the official Ruby SDK for interacting with OpenAI's REST API, offering features like streaming responses, pagination handling, and type support for Ruby 3.2.0+. It matters right now as developers building AI-integrated applications in Ruby need reliable, official tools to access OpenAI's capabilities efficiently, especially with the increasing demand for language-specific SDKs amid widespread AI adoption. _themes: api · sdk · inference · openai_ #### [NVIDIA/NVTX](https://github.com/NVIDIA/NVTX) *C++ · ★528 · NOASSERTION · production · score:0.70 · hot:0.71 · rising:0.73 · durable:0.66 · board:rising · trend:up* The NVIDIA® Tools Extension SDK (NVTX) is a C-based Application Programming Interface (API) for annotating events, code ranges, and resources in your applications. **Why it matters.** NVTX is a C-based API that enables developers to add annotations to code for profiling and visualization in compatible tools, primarily for NVIDIA environments, helping to analyze performance bottlenecks in applications. It matters right now as AI and GPU-accelerated workloads demand precise optimization to handle increasing complexity and scale, making such instrumentation essential for efficient development and debugging. _themes: profiling · instrumentation · performance · gpu_ #### [NVIDIA/cuda-samples](https://github.com/NVIDIA/cuda-samples) *C · ★9,095 · NOASSERTION · production · score:0.85 · hot:0.68 · rising:0.72 · durable:0.74 · board:durable · trend:up* Samples for CUDA Developers which demonstrates features in CUDA Toolkit **Why it matters.** This repository provides sample code demonstrating features of NVIDIA's CUDA Toolkit, helping developers learn and implement GPU programming techniques. It matters right now because CUDA is essential for optimizing performance in AI, machine learning, and scientific computing, especially as demand for efficient GPU acceleration grows with advancing AI models. _themes: cuda · gpu-computing · parallel-programming · samples_ #### [xai-org/xai-sdk-python](https://github.com/xai-org/xai-sdk-python) *Python · ★411 · Apache-2.0 · production · score:0.60 · hot:0.66 · rising:0.69 · durable:0.66 · board:rising · trend:stable* The official Python SDK for the xAI API **Why it matters.** This repository offers the official Python SDK for xAI's API, allowing developers to interact with xAI's services for tasks like text and image generation using gRPC-based synchronous and asynchronous clients. It matters now as xAI, a relatively new AI company, provides an alternative to established players, potentially appealing to developers seeking diverse AI options, though its modest 409 stars suggest limited adoption and unproven widespread utility. _themes: ai-sdk · grpc · async · api-integration_ #### [microsoft/skills](https://github.com/microsoft/skills) *TypeScript · ★2,091 · MIT · experimental · score:0.70 · hot:0.66 · rising:0.66 · durable:0.61 · board:rising · trend:stable* Skills, MCP servers, Custom Agents, Agents.md for SDKs to ground Coding Agents **Why it matters.** This repository provides a collection of skills, custom agents, and configurations for AI coding agents that integrate with Azure SDKs and Microsoft AI Foundry, allowing developers to enhance agent capabilities for tasks like interacting with cloud services. It matters now because the growing adoption of AI agents in software development demands reusable, pre-built components to streamline workflows and boost productivity, especially within Microsoft ecosystems, though it's still in active development and not fully mature. _themes: agents · sdk · azure · skills_ #### [openai/GABRIEL](https://github.com/openai/GABRIEL) *Jupyter Notebook · ★367 · Apache-2.0 · beta · score:0.60 · hot:0.66 · rising:0.67 · durable:0.67 · board:durable · trend:stable* An official OpenAI toolkit for social scientists and data scientists to measure quantitative attributes in text, images, or audio using the GPT API. **Why it matters.** GABRIEL is an OpenAI toolkit that streamlines the use of GPT for extracting quantitative attributes from text, images, and audio, primarily for social and data scientists, by handling tasks like prompting, batching, and error management. It matters now because the growing adoption of LLMs in research demands reliable tools to convert qualitative data into analyzable datasets, though its niche focus on social science applications limits broader appeal amid increasing competition in LLM wrappers. _themes: inference · multimodal · prompting · research_ #### [microsoft/Windows-driver-samples](https://github.com/microsoft/Windows-driver-samples) *C · ★7,692 · MS-PL · production · score:0.80 · hot:0.66 · rising:0.74 · durable:0.74 · board:rising · trend:up* This repo contains driver samples prepared for use with Microsoft Visual Studio and the Windows Driver Kit (WDK). It contains both Universal Windows Driver and desktop-only driver samples. **Why it matters.** This repository contains official code samples for developing Windows drivers, specifically for Windows 11, using Microsoft Visual Studio and the Windows Driver Kit (WDK), covering both universal and desktop drivers. It matters because driver development is essential for hardware compatibility in the Windows ecosystem, particularly with ongoing updates to Windows 11 that introduce new features and requirements for devices like cameras and Bluetooth. However, its relevance is niche, primarily benefiting experienced developers who need to integrate hardware with Windows, and it requires specific tools like Visual Studio 2022, limiting broader adoption. _themes: drivers · windows · hardware · development_ #### [NVIDIA/cuda-tile](https://github.com/NVIDIA/cuda-tile) *C++ · ★940 · NOASSERTION · beta · score:0.75 · hot:0.63 · rising:0.65 · durable:0.68 · board:durable · trend:stable* CUDA Tile IR is an MLIR-based intermediate representation and compiler infrastructure for CUDA kernel optimization, focusing on tile-based computation patterns and optimizations targeting NVIDIA tensor core units. **Why it matters.** CUDA Tile IR is an MLIR-based intermediate representation and compiler infrastructure designed to optimize CUDA kernels for tile-based computations on NVIDIA GPUs, particularly targeting tensor core units to improve performance in high-compute workloads. It matters now because the growing demand for efficient GPU acceleration in AI and machine learning requires tools that simplify kernel development and memory management, potentially reducing optimization overhead for developers working on NVIDIA hardware. _themes: gpu · optimization · compiler · mlir_ #### [NVIDIA/cuda-q-academic](https://github.com/NVIDIA/cuda-q-academic) *Jupyter Notebook · ★314 · NOASSERTION · beta · score:0.60 · hot:0.62 · rising:0.63 · durable:0.55 · board:rising · trend:stable* This repo contains CUDA-Q Academic materials, including self-paced Jupyter notebook modules for building and optimizing hybrid quantum-classical algorithms using CUDA-Q. **Why it matters.** This repository offers Jupyter notebooks for learning CUDA-Q, a platform for developing hybrid quantum-classical algorithms, primarily aimed at educational purposes. It matters because it provides free resources from NVIDIA to build skills in emerging quantum computing, which could intersect with AI/ML, but its niche focus and lack of formal releases limit immediate practical adoption for most users. _themes: quantum · education · hybrid-algorithms · notebooks_ #### [huggingface/mcp-course](https://github.com/huggingface/mcp-course) *MDX · ★895 · Apache-2.0 · experimental · score:0.50 · hot:0.61 · rising:0.62 · durable:0.52 · board:rising · trend:stable* **Why it matters.** This repository provides a free course on the Model Context Protocol (MCP), aimed at teaching users how to integrate AI models with external data and tools through basic to advanced implementations. It matters because standardizing such integrations could address growing needs in AI development for more interactive and robust applications, though its value is currently limited as the course content is not yet available. With HuggingFace's influence, MCP might become a key standard if it gains traction. _themes: mcp · ai-integration · sdk · agents_ #### [xai-org/xai-cookbook](https://github.com/xai-org/xai-cookbook) *TypeScript · ★371 · NOASSERTION · beta · score:0.60 · hot:0.60 · rising:0.62 · durable:0.52 · board:rising · trend:stable* A collection of pragmatic, real-world examples guiding you from basic to advanced use of xAI's Grok APIs. **Why it matters.** The xai-cookbook repo provides a collection of practical, real-world examples for using xAI's Grok APIs, covering basic to advanced implementations through Jupyter notebooks. It matters right now because xAI is an emerging AI provider, and this resource helps developers quickly learn and apply Grok APIs amid growing interest in alternative AI models. However, its lack of formal releases and unspecified license may limit its reliability for production use. _themes: api · examples · notebooks · grok_ #### [NVIDIA/aerial-cuda-accelerated-ran](https://github.com/NVIDIA/aerial-cuda-accelerated-ran) *C++ · ★333 · NOASSERTION · beta · score:0.70 · hot:0.60 · rising:0.62 · durable:0.55 · board:rising · trend:stable* An SDK (Software Development Kit) for building commercial-grade, AI-native, 3GPP, and O-RAN compliant 5G/6G gNB software on NVIDIA-accelerated computing platforms. **Why it matters.** This SDK provides tools for building AI-native 5G and 6G radio access network software using NVIDIA's CUDA acceleration, including GPU-accelerated components for physical layer processing, scheduling, and AI integration. It matters now as telecommunications are advancing towards AI-driven optimizations for 5G deployment and 6G development, potentially improving network efficiency and enabling new applications in a rapidly evolving industry. _themes: ai · 5g · cuda · gpu_ #### [google-deepmind/gemini-robotics-sdk](https://github.com/google-deepmind/gemini-robotics-sdk) *Python · ★573 · Apache-2.0 · beta · score:0.60 · hot:0.59 · rising:0.60 · durable:0.56 · board:rising · trend:stable* **Why it matters.** This SDK provides tools for managing Google DeepMind's Gemini Robotics models, including model serving, evaluation on real and simulated robots, fine-tuning, and data handling, primarily for advanced users in the field. It matters because it represents early access to cutting-edge AI for robotics applications, but its restricted availability through a trusted tester program limits its immediate impact and broader adoption. _themes: robotics · fine-tuning · inference · agents_ #### [google-deepmind/alphagenome](https://github.com/google-deepmind/alphagenome) *Python · ★1,873 · Apache-2.0 · beta · score:0.80 · hot:0.59 · rising:0.65 · durable:0.74 · board:durable · trend:stable* This API provides programmatic access to the AlphaGenome model developed by Google DeepMind. **Why it matters.** This repository provides an API for accessing AlphaGenome, a DeepMind model that predicts genomic features like gene expression and chromatin patterns from DNA sequences up to 1 million base pairs. It matters now because it offers state-of-the-art tools for genomic analysis at no cost for non-commercial use, potentially advancing research in biology and medicine amid growing AI applications in life sciences, though its query limits may restrict large-scale applications. _themes: genomics · inference · ai-biology · prediction_ #### [huggingface/AnyLanguageModel](https://github.com/huggingface/AnyLanguageModel) *Swift · ★835 · Apache-2.0 · beta · score:0.70 · hot:0.58 · rising:0.60 · durable:0.65 · board:durable · trend:stable* An API-compatible, drop-in replacement for Apple's Foundation Models framework with support for custom language model providers. **Why it matters.** AnyLanguageModel is a Swift library that acts as a drop-in replacement for Apple's Foundation Models, allowing developers to integrate custom language model providers and tools for tasks like generating responses with external data. It matters now because it addresses the limitations of Apple's closed ecosystem by enabling more flexible AI integration in iOS and macOS apps, amid increasing demand for customizable LLMs in a rapidly evolving AI landscape. This could help developers build more versatile applications without overhauling existing codebases. _themes: swift · llm · tools · agents_ #### [microsoft/dev-tunnels](https://github.com/microsoft/dev-tunnels) *C# · ★459 · MIT · beta · score:0.60 · hot:0.57 · rising:0.58 · durable:0.52 · board:rising · trend:stable* Dev Tunnels SDK **Why it matters.** The Dev Tunnels SDK allows developers to securely expose local web services to the internet for testing and debugging, with features like access control and language-specific support in C#, TypeScript, and others. It matters now as remote development workflows become standard, addressing the need for safe, controlled external access to local environments without relying on less secure alternatives like port forwarding. _themes: tunneling · security · debugging · devtools_ #### [openai/chatkit-python](https://github.com/openai/chatkit-python) *Python · ★377 · Apache-2.0 · production · score:0.40 · hot:0.56 · rising:0.60 · durable:0.63 · board:durable · trend:stable* **Why it matters.** This repository appears to be a Python library for interacting with OpenAI's chat-related APIs, likely providing utilities for building conversational AI applications, though it's unclear if it's an official project or a wrapper given the lack of detailed documentation in the provided excerpt. It matters right now because the proliferation of chatbots and AI assistants demands easy-to-use tools for developers, but its value is questionable since OpenAI's primary SDK already covers these features, potentially making this redundant unless it offers unique enhancements. _themes: chat · ai · sdk · inference_ #### [microsoft/DataConnectors](https://github.com/microsoft/DataConnectors) *PowerShell · ★780 · MIT · production · score:0.60 · hot:0.53 · rising:0.59 · durable:0.62 · board:durable · trend:stable* Data Connector SDK and samples for Power Query and Power BI **Why it matters.** This repository offers an SDK and samples for developing custom data connectors that integrate with Power Query and Power BI, allowing users to connect to various data sources for enhanced analytics. However, it appears to be superseded by a newer SDK in public preview, as indicated in the README, meaning future development and support have shifted elsewhere. As a result, while it remains a resource for existing implementations, its relevance is diminishing for new projects. _themes: data-connectors · powerquery · bi · sdk_ #### [microsoft/Windows-universal-samples](https://github.com/microsoft/Windows-universal-samples) *JavaScript · ★9,693 · MIT · production · score:0.60 · hot:0.51 · rising:0.63 · durable:0.71 · board:durable · trend:stable* API samples for the Universal Windows Platform. **Why it matters.** This repository offers a collection of code samples demonstrating API usage for the Universal Windows Platform (UWP), enabling developers to build apps that run across Windows devices like desktop and mobile. It matters now because UWP remains a foundational part of Windows app development, especially for maintaining legacy apps on Windows 10 and 11, though it's being overshadowed by newer frameworks like WinUI, making these samples valuable for developers updating or learning existing UWP codebases. _themes: uwp · windows-api · app-samples · cross-platform_ #### [NVIDIA/DLSS](https://github.com/NVIDIA/DLSS) *C · ★1,299 · NOASSERTION · production · score:0.70 · hot:0.50 · rising:0.57 · durable:0.70 · board:durable · trend:stable* NVIDIA DLSS is a new and improved deep learning neural network that boosts frame rates and generates beautiful, sharp images for your games **Why it matters.** NVIDIA DLSS is an AI-driven upscaling technology that uses deep learning neural networks to enhance game graphics by increasing frame rates and image sharpness, integrated via an SDK for developers. It matters right now because it addresses performance bottlenecks in modern gaming, especially on NVIDIA hardware, amid growing demand for efficient real-time rendering solutions as AI accelerates in graphics applications. _themes: upscaling · deep-learning · rendering · ai_ #### [NVIDIA/ACE](https://github.com/NVIDIA/ACE) *HCL · ★310 · NOASSERTION · production · score:0.65 · hot:0.45 · rising:0.52 · durable:0.65 · board:durable · trend:down* NVIDIA ACE samples, workflows, and resources **Why it matters.** This repository provides samples, workflows, and resources for NVIDIA's ACE technologies, which are microservices for creating digital humans using generative AI, focusing on applications like speech recognition and text translation; however, the actual microservices require a separate evaluation license from NVIDIA AI Enterprise, making it more of a supplementary guide than a standalone solution. It matters now amid the rise of AI-driven digital interactions in gaming and customer service, but its dependency on proprietary tools and modest adoption (310 stars) may hinder accessibility for non-NVIDIA users. _themes: generative-ai · speech-recognition · inference · digital-humans_ #### [google-gemini/deprecated-generative-ai-python](https://github.com/google-gemini/deprecated-generative-ai-python) *Python · ★2,287 · Apache-2.0 · archived · score:0.30 · hot:0.44 · rising:0.48 · durable:0.54 · board:durable · trend:stable* This SDK is now deprecated, use the new unified Google GenAI SDK. **Why it matters.** This repository is a deprecated Python SDK for interacting with Google's Gemini API, which has been superseded by a new unified SDK for better features and maintenance. It matters right now because developers currently using it must migrate to avoid potential disruptions, as this version is only receiving critical bug fixes and will soon reach end-of-life. _themes: genai · inference · sdk · migration_ #### [openai/openai-realtime-embedded](https://github.com/openai/openai-realtime-embedded) *? · ★1,581 · MIT · beta · score:0.70 · hot:0.44 · rising:0.52 · durable:0.64 · board:durable · trend:stable* Instructions on how to use the Realtime API on Microcontrollers and Embedded Platforms **Why it matters.** This repository offers instructions and examples for integrating OpenAI's Realtime API with WebRTC on microcontrollers like ESP32, focusing on real-time AI interactions in embedded environments. It matters now because the push for edge computing in AI is growing, enabling low-latency applications in IoT and wearables, but its lack of official releases and maintenance updates raises questions about long-term reliability and broader applicability. _themes: realtime · embedded · webrtc · inference_ #### [NVIDIA/open-gpu-doc](https://github.com/NVIDIA/open-gpu-doc) *C · ★1,330 · MIT · production · score:0.60 · hot:0.44 · rising:0.52 · durable:0.63 · board:durable · trend:stable* Documentation of NVIDIA chip/hardware interfaces **Why it matters.** This repository provides open documentation for NVIDIA's GPU hardware interfaces, detailing low-level chip operations for developers working with NVIDIA technology. It matters now because understanding these interfaces is essential for optimizing AI, machine learning, and high-performance computing applications on NVIDIA hardware, though its utility is hampered by the lack of formal releases and reliance on GitHub commits for updates. _themes: gpu · hardware · interfaces · documentation_ #### [openai/openai-realtime-api-beta](https://github.com/openai/openai-realtime-api-beta) *JavaScript · ★1,013 · MIT · beta · score:0.70 · hot:0.43 · rising:0.49 · durable:0.59 · board:durable · trend:stable* Node.js + JavaScript reference client for the Realtime API (beta) **Why it matters.** This repository offers a JavaScript reference client for OpenAI's Realtime API in beta, allowing developers to prototype conversational applications with real-time voice and text interactions using Node.js or browser environments. It matters now because real-time AI features are essential for emerging applications like voice assistants, and OpenAI's ecosystem is influential, but its beta status means it could change or have bugs, making it suitable only for early experimentation rather than production. _themes: realtime · api · conversational-ai · voice_ ### tool (161) #### [langgenius/dify](https://github.com/langgenius/dify) *TypeScript · ★138,311 · NOASSERTION · production · score:0.85 · hot:0.90 · rising:0.89 · durable:0.87 · board:hot · trend:up* Production-ready platform for agentic workflow development. **Why it matters.** Dify is an open-source platform that simplifies building and deploying LLM-based applications with features like agentic workflows, RAG pipelines, and observability tools, allowing users to move from prototype to production quickly. It matters now because the rapid growth of AI agents and generative AI demands accessible, low-code solutions for developers to orchestrate complex workflows efficiently, amid increasing enterprise adoption of tools like this. _themes: agents · rag · workflow · llm_ #### [unslothai/unsloth](https://github.com/unslothai/unsloth) *Python · ★62,476 · Apache-2.0 · beta · score:0.80 · hot:0.89 · rising:0.90 · durable:0.81 · board:rising · trend:up* Web UI for training and running open models like Gemma 4, Qwen3.5, DeepSeek, gpt-oss locally. **Why it matters.** Unsloth provides a web-based user interface for training and running open-source AI models like Gemma, Qwen, and DeepSeek directly on local machines, simplifying fine-tuning and inference without relying on cloud services. It matters now because the growing emphasis on privacy, cost efficiency, and on-device AI processing makes local model handling increasingly relevant, especially as open models proliferate and users seek accessible tools for experimentation and development. _themes: fine-tuning · inference · local-ai · ui_ #### [ollama/ollama](https://github.com/ollama/ollama) *Go · ★169,714 · MIT · production · score:0.90 · hot:0.89 · rising:0.94 · durable:0.85 · board:rising · trend:up* Get up and running with Kimi-K2.5, GLM-5, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models. **Why it matters.** Ollama is an open-source tool written in Go that simplifies running and managing various large language models like Llama, Gemma, and Qwen on local machines with easy installation and integrations. It matters right now due to the growing emphasis on privacy, offline AI capabilities, and cost-effective alternatives to cloud-based services, making it essential for developers experimenting with LLMs amid regulatory scrutiny and hardware advancements. This accessibility lowers the entry barrier for building AI applications without proprietary dependencies. _themes: inference · llm · local-ai · open-source_ #### [open-webui/open-webui](https://github.com/open-webui/open-webui) *Python · ★132,658 · NOASSERTION · production · score:0.90 · hot:0.89 · rising:0.90 · durable:0.87 · board:rising · trend:up* User-friendly AI Interface (Supports Ollama, OpenAI API, ...) **Why it matters.** Open WebUI is a self-hosted web interface for interacting with various large language models, including Ollama and OpenAI-compatible APIs, featuring built-in support for retrieval-augmented generation (RAG) and easy deployment via Docker or Kubernetes. It matters now because it addresses the growing need for privacy-focused, offline AI tools amid increasing regulatory scrutiny and the democratization of AI, allowing users to customize and secure their AI workflows without relying on cloud services. _themes: rag · llm · inference · self-hosted_ #### [Mintplex-Labs/anything-llm](https://github.com/Mintplex-Labs/anything-llm) *JavaScript · ★58,596 · MIT · production · score:0.80 · hot:0.89 · rising:0.88 · durable:0.86 · board:hot · trend:up* The all-in-one AI productivity accelerator. On device and privacy first with no annoying setup or configuration. **Why it matters.** AnythingLLM is an all-in-one AI tool that allows users to chat with their documents, build and deploy custom AI agents, and manage multi-user setups, all running locally with minimal configuration for enhanced privacy. It matters now because it addresses the growing demand for accessible, privacy-focused AI solutions amid increasing data security concerns and the proliferation of LLMs, making advanced AI capabilities available to non-experts without the typical setup hurdles. _themes: rag · agents · local-llm · privacy_ #### [thedotmack/claude-mem](https://github.com/thedotmack/claude-mem) *TypeScript · ★63,091 · NOASSERTION · production · score:0.75 · hot:0.87 · rising:0.88 · durable:0.82 · board:rising · trend:up* A Claude Code plugin that automatically captures everything Claude does during your coding sessions, compresses it with AI (using Claude's agent-sdk), and injects relevant context back into future sessions. **Why it matters.** This repository offers a plugin for Anthropic's Claude that records and compresses coding session activities using AI to create a persistent memory, allowing relevant context to be reused in subsequent sessions, which helps maintain continuity in AI-assisted development. It matters right now because the growing adoption of AI agents in coding workflows demands better memory management to enhance efficiency and reduce repetition, especially as models like Claude gain popularity for complex tasks. However, its reliance on proprietary tools like Claude's SDK may limit broader applicability. _themes: agents · memory · embeddings · ai_ #### [firecrawl/firecrawl](https://github.com/firecrawl/firecrawl) *TypeScript · ★110,773 · AGPL-3.0 · production · score:0.85 · hot:0.87 · rising:0.90 · durable:0.86 · board:rising · trend:up* 🔥 The API to search, scrape, and interact with the web for AI **Why it matters.** Firecrawl is an API that enables AI agents to search, scrape, and extract clean web data, including handling JavaScript-heavy pages and outputting in LLM-friendly formats like markdown. It matters now because the growing demand for real-time web interaction in AI applications requires reliable, low-latency tools that abstract away complexities like proxies and rate limits, allowing developers to build more efficient agents amid the AI boom. _themes: agents · web-scraping · llm · data-extraction_ #### [FlowiseAI/Flowise](https://github.com/FlowiseAI/Flowise) *TypeScript · ★52,057 · NOASSERTION · production · score:0.80 · hot:0.87 · rising:0.87 · durable:0.80 · board:rising · trend:up* Build AI Agents, Visually **Why it matters.** Flowise is an open-source platform that allows users to visually build and deploy AI agents, chatbots, and workflows using a low-code interface, integrating with tools like Langchain and OpenAI for tasks such as RAG and multi-agent systems. It matters now because the growing adoption of LLMs and agentic AI requires accessible tools for non-experts, enabling faster prototyping and deployment in applications like customer support and automation, though its reliance on external APIs may introduce dependency risks. _themes: agents · rag · low-code · chatbots_ #### [langfuse/langfuse](https://github.com/langfuse/langfuse) *TypeScript · ★25,538 · NOASSERTION · production · score:0.80 · hot:0.86 · rising:0.86 · durable:0.76 · board:hot · trend:up* 🪢 Open source LLM engineering platform: LLM Observability, metrics, evals, prompt management, playground, datasets. Integrates with OpenTelemetry, Langchain, OpenAI SDK, LiteLLM, and more. 🍊YC W23 **Why it matters.** Langfuse is an open-source platform that provides observability, monitoring, evaluation, and prompt management for large language model applications, integrating with tools like Langchain and OpenAI. It matters right now because the growing complexity of AI deployments demands robust debugging and performance tracking to prevent issues like hallucinations or inefficiencies, especially as teams rush to productionize LLMs. However, its reliance on integrations means it may not suit standalone use without those ecosystems. _themes: llm-observability · evaluation · prompt-management · monitoring_ #### [oobabooga/textgen](https://github.com/oobabooga/textgen) *Python · ★46,843 · AGPL-3.0 · production · score:0.85 · hot:0.86 · rising:0.92 · durable:0.81 · board:rising · trend:up* The original local LLM interface. Text, vision, tool-calling, training. UI + API, 100% offline and private. **Why it matters.** Oobabooga/textgen provides a comprehensive, offline interface for running large language models locally, including text generation, vision processing, tool-calling, and training via a user-friendly UI and API. It matters right now as it addresses privacy and cost concerns in AI by enabling fully offline operations, making it a practical alternative for developers avoiding cloud dependencies amid increasing regulatory scrutiny. Its support for multiple backends and easy setup lowers barriers for local LLM experimentation and deployment. _themes: inference · local-llm · tool-calling · ui_ #### [n8n-io/n8n](https://github.com/n8n-io/n8n) *TypeScript · ★184,707 · NOASSERTION · production · score:0.90 · hot:0.86 · rising:0.90 · durable:0.83 · board:rising · trend:up* Fair-code workflow automation platform with native AI capabilities. Combine visual building with custom code, self-host or cloud, 400+ integrations. **Why it matters.** n8n is an open-source workflow automation platform that allows users to build, automate, and integrate processes using a visual interface combined with custom code, supporting over 400 integrations and native AI features for tasks like AI agent workflows. It matters now because businesses are increasingly seeking flexible, self-hosted automation tools amid growing data privacy concerns and the need to incorporate AI without vendor lock-in, making it a practical choice for efficient, customizable operations. _themes: automation · integrations · ai · low-code_ #### [netdata/netdata](https://github.com/netdata/netdata) *C · ★78,508 · GPL-3.0 · production · score:0.85 · hot:0.86 · rising:0.90 · durable:0.81 · board:rising · trend:up* The fastest path to AI-powered full stack observability, even for lean teams. **Why it matters.** Netdata is an open-source, real-time infrastructure monitoring platform that provides per-second metrics, visualizations, and AI-powered anomaly detection with minimal setup. It matters now because organizations are dealing with increasingly complex distributed systems where immediate insights are crucial for maintaining reliability and preventing downtime, especially for lean teams in DevOps environments. _themes: monitoring · observability · ai · real-time_ #### [santifer/career-ops](https://github.com/santifer/career-ops) *JavaScript · ★36,369 · MIT · production · score:0.60 · hot:0.85 · rising:0.88 · durable:0.81 · board:rising · trend:up* AI-powered job search system built on Claude Code. 14 skill modes, Go dashboard, PDF generation, batch processing. **Why it matters.** Career-Ops is an AI-powered tool that automates job search tasks like evaluating offers with a scoring system, generating tailored ATS-optimized resumes, and scanning job portals via CLI, but it requires users to provide personal context for accurate results. It matters now in a saturated job market where AI is used for candidate screening, as it helps individuals filter high-quality opportunities efficiently, though its reliance on Anthropic's Claude may limit accessibility and accuracy without fine-tuning. However, the system's batch processing and tracking features could save time, but early evaluations might be suboptimal, making it more of a supplementary aid than a reliable standalone solution. _themes: agents · automation · resume · job-search_ #### [iOfficeAI/AionUi](https://github.com/iOfficeAI/AionUi) *TypeScript · ★22,198 · Apache-2.0 · production · score:0.80 · hot:0.85 · rising:0.85 · durable:0.80 · board:rising · trend:up* Free, local, open-source 24/7 Cowork app and OpenClaw for Gemini CLI, Claude Code, Codex, OpenCode, Qwen Code, Goose CLI, Auggie, and more | 🌟 Star if you like it! **Why it matters.** AionUi is an open-source web UI that integrates multiple AI agents like Gemini CLI and Claude Code, allowing them to perform tasks such as file manipulation, code generation, and automation on your local machine with user oversight. It matters now because it addresses growing demands for privacy-focused, multi-model AI tools amid increasing adoption of LLMs for productivity, though its reliance on external APIs and potential complexity may limit accessibility for non-technical users. _themes: agents · llm · automation · webui_ #### [upstash/context7](https://github.com/upstash/context7) *TypeScript · ★53,161 · MIT · beta · score:0.80 · hot:0.84 · rising:0.85 · durable:0.81 · board:rising · trend:up* Context7 Platform -- Up-to-date code documentation for LLMs and AI code editors **Why it matters.** Context7 is a tool that integrates real-time, version-specific code documentation into LLM prompts, helping developers avoid outdated information and API hallucinations for more accurate AI-assisted coding. This matters now because the rapid pace of library updates and the increasing reliance on LLMs in development workflows demand reliable, up-to-date resources to enhance productivity and reduce errors. As AI code editors gain popularity, tools like this address a critical gap in ensuring contextually relevant documentation. _themes: rag · agents · llm · documentation_ #### [mindsdb/mindsdb](https://github.com/mindsdb/mindsdb) *Python · ★39,013 · NOASSERTION · production · score:0.80 · hot:0.84 · rising:0.85 · durable:0.80 · board:rising · trend:up* Query Engine for AI Analytics: Build self-reasoning agents across all your live data **Why it matters.** MindsDB is an open-source query engine that enables building AI agents for analyzing live data across various sources using natural language queries, without requiring ETL processes. It integrates SQL-compatible constructs for semantic search and agent workflows, making it suitable for conversational analytics. This matters now because the growing demand for AI-driven business intelligence requires tools that seamlessly connect LLMs with enterprise data, though its effectiveness depends on specific use cases and data complexity. _themes: agents · rag · analytics · llms_ #### [microsoft/TypeScript](https://github.com/microsoft/TypeScript) *TypeScript · ★108,586 · Apache-2.0 · production · score:1.00 · hot:0.84 · rising:0.90 · durable:0.79 · board:rising · trend:up* TypeScript is a superset of JavaScript that compiles to clean JavaScript output. **Why it matters.** TypeScript is a statically typed superset of JavaScript that compiles to plain JavaScript, helping developers catch errors early and maintain larger codebases with better tooling. It matters now because the growing complexity of web and enterprise applications demands type safety to improve productivity and reduce bugs, especially as JavaScript remains dominant in frontend and backend development. _themes: static-typing · javascript · transpiler · type-safety_ #### [microsoft/vscode](https://github.com/microsoft/vscode) *TypeScript · ★184,021 · MIT · production · score:0.90 · hot:0.83 · rising:0.89 · durable:0.79 · board:rising · trend:up* Visual Studio Code **Why it matters.** Visual Studio Code is an open-source code editor built with Electron and TypeScript, offering features like code editing, debugging, and extensibility for various programming languages. It matters right now as it's a dominant tool in software development, used by millions for its efficiency and community-driven updates, though it faces competition from lighter alternatives and potential performance issues on resource-constrained devices. _themes: editor · debugging · extensions · typescript_ #### [CherryHQ/cherry-studio](https://github.com/CherryHQ/cherry-studio) *TypeScript · ★43,837 · AGPL-3.0 · production · score:0.80 · hot:0.83 · rising:0.86 · durable:0.78 · board:rising · trend:up* AI productivity studio with smart chat, autonomous agents, and 300+ assistants. Unified access to frontier LLMs **Why it matters.** Cherry Studio is a desktop application that provides a unified interface for accessing multiple large language models from providers like OpenAI and Anthropic, along with features for managing over 300 pre-configured AI assistants and handling document processing. It matters right now because it simplifies AI integration for everyday tasks amid the growing adoption of LLMs, potentially improving productivity for users without requiring custom development, though its AGPL license may limit enterprise use. _themes: agents · llm-access · productivity · assistants_ #### [googleworkspace/cli](https://github.com/googleworkspace/cli) *Rust · ★25,037 · Apache-2.0 · beta · score:0.75 · hot:0.83 · rising:0.84 · durable:0.79 · board:rising · trend:up* Google Workspace CLI — one command-line tool for Drive, Gmail, Calendar, Sheets, Docs, Chat, Admin, and more. Dynamically built from Google Discovery Service. Includes AI agent skills. **Why it matters.** The Google Workspace CLI is a command-line tool that dynamically generates commands for interacting with Google services like Drive, Gmail, and Calendar based on Google's Discovery Service, eliminating the need for manual API boilerplate and including built-in AI agent skills for automation. It matters now because it simplifies integration and automation for developers and AI systems in an era of increasing productivity demands, though it's unofficial and under active development, potentially filling gaps left by Google's own tools amid growing AI adoption in enterprise workflows. _themes: automation · cli · ai-agents · google-api_ #### [jackwener/OpenCLI](https://github.com/jackwener/OpenCLI) *JavaScript · ★16,432 · Apache-2.0 · production · score:0.80 · hot:0.83 · rising:0.85 · durable:0.80 · board:rising · trend:up* Make Any Website & Tool Your CLI. A universal CLI Hub and AI-native runtime. Transform any website, Electron app, or local binary into a standardized command-line interface. Built for AI Agents to discover, learn, and execute tools seamlessly via a unified AGENT.md integration. **Why it matters.** OpenCLI transforms websites, Electron apps, and local binaries into standardized CLI interfaces, allowing for easy automation and integration with AI agents through a unified runtime. This matters now as AI agents increasingly require seamless access to diverse tools for tasks like web navigation and command execution, addressing a key gap in AI workflows by enabling deterministic interactions without exposing credentials. However, its reliance on browser sessions may introduce stability issues in production environments. _themes: agents · cli · automation · ai-tools_ #### [microsoft/terminal](https://github.com/microsoft/terminal) *C++ · ★102,797 · MIT · production · score:0.95 · hot:0.83 · rising:0.89 · durable:0.82 · board:rising · trend:up* The new Windows Terminal and the original Windows console host, all in the same place! **Why it matters.** The Microsoft Terminal repository contains the source code for the Windows Terminal, a modern terminal emulator that supports multiple tabs, panes, and shells like PowerShell and WSL, while also maintaining the legacy Windows console host. It matters now because it's the default terminal for Windows users, improving productivity with features like Unicode support and GPU acceleration, especially as remote development and command-line tools become essential in daily workflows. _themes: terminal · command-line · windows · open-source_ #### [LostRuins/koboldcpp](https://github.com/LostRuins/koboldcpp) *C++ · ★10,247 · AGPL-3.0 · production · score:0.80 · hot:0.82 · rising:0.84 · durable:0.73 · board:rising · trend:up* Run GGUF models easily with a KoboldAI UI. One File. Zero Install. **Why it matters.** Koboldcpp is a standalone executable that simplifies running GGUF and GGML language models with a KoboldAI-inspired UI, offering features like text generation, image processing, and voice recognition without any installation. It matters now because it enables easy, local execution of AI models on personal hardware, addressing privacy concerns and the growing demand for offline AI tools amid increasing model availability, though its reliance on external libraries may introduce subtle dependencies. With over 10,000 stars, it appeals to users seeking accessible AI experimentation but requires verification for stability in production use. _themes: inference · text-generation · multimodal · local-ai_ #### [farion1231/cc-switch](https://github.com/farion1231/cc-switch) *Rust · ★47,189 · MIT · production · score:0.70 · hot:0.82 · rising:0.85 · durable:0.79 · board:rising · trend:up* A cross-platform desktop All-in-One assistant tool for Claude Code, Codex, OpenCode, openclaw & Gemini CLI. **Why it matters.** CC Switch is a cross-platform desktop application built with Rust and Tauri that manages and integrates multiple AI coding assistants like Claude Code, Codex, Gemini CLI, OpenCode, and OpenClaw, allowing users to switch providers, manage skills, and streamline workflows. It matters right now because the growing ecosystem of AI models demands efficient tools for seamless integration and management to enhance developer productivity, though its heavy reliance on sponsored content raises questions about impartiality and long-term sustainability. _themes: ai-tools · provider-management · inference · productivity_ #### [activepieces/activepieces](https://github.com/activepieces/activepieces) *TypeScript · ★21,777 · NOASSERTION · beta · score:0.75 · hot:0.82 · rising:0.82 · durable:0.72 · board:hot · trend:up* AI Agents & MCPs & AI Workflow Automation • (~400 MCP servers for AI agents) • AI Automation / AI Agent with MCPs • AI Workflows & AI Agents • MCPs for AI Agents **Why it matters.** Activepieces is an open-source workflow automation platform that allows users to build and deploy AI agents and integrations using a TypeScript-based framework, serving as an alternative to tools like Zapier. It matters now because it provides a community-driven, extensible solution for AI automation amid rising demand for cost-effective, customizable tools, though its reliance on community contributions may lead to inconsistent quality and coverage. _themes: agents · automation · workflow · ai-framework_ #### [saturndec/waoowaoo](https://github.com/saturndec/waoowaoo) *TypeScript · ★11,503 · no-license · beta · score:0.70 · hot:0.82 · rising:0.82 · durable:0.73 · board:rising · trend:up* 首家工业级全流程 AI 影视生产平台。Industry-first professional AI Agent platform for controllable film & video production. From shorts to live-action with Hollywood-standard workflows. **Why it matters.** This repo offers an AI-based tool for generating short videos and comics from text inputs, automating processes like script analysis, character/scene creation, and video synthesis, aimed at film production. It matters now as generative AI is transforming creative industries, potentially democratizing video making, but it's in an early, buggy stage developed by one person, so its reliability and scalability are questionable for professional use. _themes: agents · generative-ai · video-generation · automation_ #### [hugohe3/ppt-master](https://github.com/hugohe3/ppt-master) *Python · ★6,449 · MIT · beta · score:0.75 · hot:0.81 · rising:0.82 · durable:0.76 · board:rising · trend:up* AI generates natively editable PPTX from any document — real PowerPoint shapes, not images — no design skills needed **Why it matters.** This repository offers a Python-based AI tool that converts documents like PDFs, DOCXs, URLs, or Markdown into natively editable PowerPoint files with real shapes and text, addressing the common frustration of uneditable AI-generated outputs. It matters now because it provides an open-source, cost-effective alternative for professionals creating presentations daily, though its output quality depends on the underlying AI models and may not always match professional design standards. However, it risks dependency on evolving AI capabilities and potential inaccuracies in complex document parsing. _themes: ai-agent · document-processing · presentation-generation · productivity_ #### [microsoft/PowerToys](https://github.com/microsoft/PowerToys) *C# · ★131,852 · MIT · production · score:0.90 · hot:0.81 · rising:0.87 · durable:0.79 · board:rising · trend:up* Microsoft PowerToys is a collection of utilities that supercharge productivity and customization on Windows **Why it matters.** Microsoft PowerToys is a set of utilities for Windows that provide features like window management, color picking, and advanced pasting to enhance productivity and customization. It matters right now because it fills gaps in the Windows operating system, offering reliable tools for everyday users amid ongoing OS updates, though some utilities may overlap with built-in features or require careful configuration to avoid conflicts. _themes: productivity · windows · utilities · customization_ #### [JuliaLang/julia](https://github.com/JuliaLang/julia) *Julia · ★48,601 · MIT · production · score:0.90 · hot:0.81 · rising:0.86 · durable:0.74 · board:rising · trend:up* The Julia Programming Language **Why it matters.** Julia is a high-performance dynamic programming language designed for technical computing, particularly in numerical analysis, data science, and machine learning, offering a balance of ease-of-use and speed that rivals compiled languages. It matters right now because the growing demands in AI research and scientific simulations require efficient tools that reduce development time without sacrificing performance, though it still faces challenges in ecosystem maturity compared to established languages like Python. Adoption is increasing due to its just-in-time compilation and parallel computing capabilities, making it relevant for high-performance computing tasks. _themes: hpc · machine-learning · numerical-computing · science_ #### [microsoft/playwright-mcp](https://github.com/microsoft/playwright-mcp) *TypeScript · ★31,094 · Apache-2.0 · beta · score:0.70 · hot:0.81 · rising:0.84 · durable:0.80 · board:rising · trend:up* Playwright MCP server **Why it matters.** Playwright MCP is a server that allows LLMs to perform browser automation using Playwright's accessibility snapshots, avoiding the need for visual processing and making it suitable for text-based AI interactions. It matters now as AI agents increasingly require efficient web interaction tools for tasks like automated testing and exploratory workflows, though it faces competition from more token-efficient alternatives like CLI-based approaches that reduce context overhead. _themes: agents · browser-automation · llm-tools · accessibility_ #### [jeecgboot/JeecgBoot](https://github.com/jeecgboot/JeecgBoot) *Java · ★45,885 · Apache-2.0 · production · score:0.75 · hot:0.81 · rising:0.84 · durable:0.84 · board:rising · trend:up* 一款 AI 驱动的低代码平台，提供"零代码"与"代码生成"双模式——零代码模式一句话搭建系统，代码生成模式自动输出前后端代码与建表 SQL，生成即可运行。平台内置 AI 聊天助手、AI大模型、知识库、AI流程编排、MCP 与插件体系，兼容主流大模型，支持一句话生成流程图、设计表单、聊天式业务操作，解决 Java 项目 80% 重复工作，高效且不失灵活。 **Why it matters.** JeecgBoot is an AI-driven low-code platform for Java development that automates code generation, system building via natural language, and includes features like AI chat assistants and workflow orchestration, reducing repetitive tasks by up to 80%. It matters now amid the growing demand for rapid application development tools that integrate AI, helping enterprises accelerate prototyping and deployment in a competitive market while maintaining flexibility. _themes: low-code · ai-generation · codegen · workflows_ #### [ItzCrazyKns/Vane](https://github.com/ItzCrazyKns/Vane) *TypeScript · ★33,852 · MIT · production · score:0.80 · hot:0.81 · rising:0.83 · durable:0.82 · board:rising · trend:up* Vane is an AI-powered answering engine. **Why it matters.** Vane is a self-hosted AI answering engine that integrates web search with local or cloud-based LLMs to provide cited answers while prioritizing user privacy, effectively combining tools like SearxNG for search and various AI models for response generation. It matters now amid rising privacy concerns and the proliferation of AI chatbots, offering a customizable alternative for users wary of data collection by commercial services, though it relies on existing technologies and may not introduce novel innovations. However, its popularity indicates practical utility for those seeking control over their AI interactions. _themes: rag · agents · search-engine · privacy_ #### [Panniantong/Agent-Reach](https://github.com/Panniantong/Agent-Reach) *Python · ★17,835 · MIT · beta · score:0.80 · hot:0.80 · rising:0.80 · durable:0.81 · board:durable · trend:up* Give your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees. **Why it matters.** Agent-Reach is a Python-based CLI tool that allows AI agents to scrape and access content from platforms like Twitter, Reddit, YouTube, and others without API fees, handling web restrictions and data cleaning automatically. It matters right now because AI agents are increasingly used for tasks requiring real-time web data, but developers face barriers like costs and IP blocks, which this tool addresses to streamline agent development. However, its reliance on scraping could lead to reliability issues if platforms enforce anti-scraping measures. _themes: agents · web-scraping · ai-tools · automation_ #### [openai/codex-plugin-cc](https://github.com/openai/codex-plugin-cc) *JavaScript · ★15,126 · Apache-2.0 · beta · score:0.70 · hot:0.80 · rising:0.83 · durable:0.76 · board:rising · trend:up* Use Codex from Claude Code to review code or delegate tasks. **Why it matters.** This repository offers a plugin for Claude Code that integrates OpenAI's Codex to perform code reviews and delegate tasks, allowing users to leverage AI assistance within their existing workflow. It matters now as AI-driven coding tools are gaining traction for productivity gains, but its utility is niche, limited to Claude Code users and dependent on OpenAI's API, which may incur costs and has usage limits. Critics might note that it adds another layer of dependency without addressing broader compatibility or privacy concerns. _themes: ai · code-review · plugins · automation_ #### [p-e-w/heretic](https://github.com/p-e-w/heretic) *Python · ★19,588 · AGPL-3.0 · beta · score:0.75 · hot:0.79 · rising:0.81 · durable:0.79 · board:rising · trend:up* Fully automatic censorship removal for language models **Why it matters.** Heretic is a Python tool that automatically removes censorship from transformer-based language models using directional ablation and optimization techniques, aiming to minimize refusals while preserving the original model's behavior; this matters now because it highlights vulnerabilities in AI safety alignments amid growing regulatory scrutiny and ethical debates, potentially enabling misuse but also advancing model customization. _themes: llm · abliteration · model-editing · optimization_ #### [microsoft/winget-cli](https://github.com/microsoft/winget-cli) *C++ · ★25,719 · MIT · production · score:0.85 · hot:0.79 · rising:0.83 · durable:0.74 · board:rising · trend:up* WinGet is the Windows Package Manager. This project includes a CLI (Command Line Interface), PowerShell modules, and a COM (Component Object Model) API (Application Programming Interface). **Why it matters.** WinGet is Microsoft's command-line tool for managing software packages on Windows, allowing users to install, update, and remove applications from sources like the Microsoft Store and community repositories. It matters now because it provides a standardized, efficient alternative to manual installations, which is increasingly important for developers and IT teams dealing with Windows 11 adoption and automation needs, though it's limited to Windows ecosystems and may face policy restrictions. _themes: cli · package-manager · windows · automation_ #### [f/prompts.chat](https://github.com/f/prompts.chat) *HTML · ★160,101 · NOASSERTION · production · score:0.90 · hot:0.79 · rising:0.82 · durable:0.80 · board:rising · trend:up* f.k.a. Awesome ChatGPT Prompts. Share, discover, and collect prompts from the community. Free and open source — self-host for your organization with complete privacy. **Why it matters.** This repository is a community-curated collection of prompts for AI chat models like ChatGPT, Claude, and Gemini, allowing users to share, discover, and adapt them for better AI interactions, with options for self-hosting to ensure privacy. It matters now because prompt engineering is essential for maximizing AI tool effectiveness amid the AI boom, and this resource has gained massive traction with over 160k stars, academic citations, and endorsements from industry leaders, making it a go-to reference for practical AI application. _themes: prompt-engineering · llm · ai · community_ #### [microsoft/vscode-copilot-chat](https://github.com/microsoft/vscode-copilot-chat) *TypeScript · ★9,830 · MIT · beta · score:0.85 · hot:0.79 · rising:0.82 · durable:0.73 · board:rising · trend:up* Copilot Chat extension for VS Code **Why it matters.** The microsoft/vscode-copilot-chat repo is a VS Code extension that integrates AI agents to autonomously handle coding tasks like planning, editing files, running commands, and self-correcting errors, building on GitHub Copilot's capabilities. It matters right now because AI-assisted coding is rapidly evolving to boost developer productivity amid widespread adoption of LLMs, but its reliance on a paid subscription may hinder accessibility for some users, potentially limiting its immediate impact. _themes: agents · coding-assistant · ai-programming · inference_ #### [microsoft/azure-devops-mcp](https://github.com/microsoft/azure-devops-mcp) *TypeScript · ★1,584 · MIT · beta · score:0.70 · hot:0.78 · rising:0.80 · durable:0.72 · board:rising · trend:up* The MCP server for Azure DevOps, bringing the power of Azure DevOps directly to your agents. **Why it matters.** This repo provides a local TypeScript server for interacting with Azure DevOps, enabling users to execute tasks like listing projects, builds, and repos directly from their code editor without relying on the cloud interface. It matters now because Microsoft is actively recommending migration to a remote version in public preview, indicating this local implementation may soon be deprecated, yet it remains valuable for developers seeking offline or low-latency access to Azure DevOps features. _themes: devops · integration · ci-cd · agents_ #### [microsoft/aspire](https://github.com/microsoft/aspire) *C# · ★5,834 · MIT · production · score:0.80 · hot:0.78 · rising:0.82 · durable:0.72 · board:rising · trend:up* Aspire is the tool for code-first, extensible, observable dev and deploy. **Why it matters.** Aspire is a code-first toolchain from Microsoft for defining, running, and deploying distributed applications using languages like C# and TypeScript, integrating services, containers, databases, and observability via OpenTelemetry. It simplifies local development and cloud deployment but may add complexity for non-.NET users due to its ecosystem ties, making it relevant amid growing microservices adoption, though its multi-language support is still emerging and not as mature as dedicated IaC tools. _themes: cloud · dotnet · observability · deployment_ #### [microsoft/agent-governance-toolkit](https://github.com/microsoft/agent-governance-toolkit) *Python · ★1,116 · MIT · beta · score:0.80 · hot:0.78 · rising:0.79 · durable:0.72 · board:rising · trend:up* AI Agent Governance Toolkit — Policy enforcement, zero-trust identity, execution sandboxing, and reliability engineering for autonomous AI agents. Covers 10/10 OWASP Agentic Top 10. **Why it matters.** The Agent Governance Toolkit provides runtime governance for AI agents, including deterministic policy enforcement, zero-trust identity, execution sandboxing, and reliability engineering, while covering all 10 OWASP Agentic risks with extensive testing. It matters right now because the rapid adoption of autonomous AI agents heightens security and compliance risks, and this toolkit offers a practical way for organizations to enforce policies and mitigate vulnerabilities in production environments before they lead to real-world issues. _themes: agents · security · governance · ai-safety_ #### [microsoft/pyright](https://github.com/microsoft/pyright) *Python · ★15,385 · NOASSERTION · production · score:0.85 · hot:0.78 · rising:0.81 · durable:0.73 · board:rising · trend:up* Static Type Checker for Python **Why it matters.** Pyright is a static type checker for Python that analyzes code for type-related errors without execution, supporting large codebases with high performance and standards compliance. It matters now because Python's increasing use in complex, production environments demands better type safety to reduce bugs and improve maintainability, especially as type annotations become more prevalent in modern development workflows. Its integration with tools like VS Code makes it practical for everyday use in professional settings. _themes: type-checking · static-analysis · python · code-quality_ #### [microsoft/vcpkg](https://github.com/microsoft/vcpkg) *CMake · ★26,913 · MIT · production · score:0.80 · hot:0.78 · rising:0.83 · durable:0.73 · board:rising · trend:up* C++ Library Manager for Windows, Linux, and MacOS **Why it matters.** vcpkg is a package manager for C++ libraries that simplifies dependency management across Windows, Linux, and macOS by handling downloads, builds, and integration with build systems like CMake. It matters right now because C++ projects often face challenges with cross-platform compatibility and library versioning, and vcpkg provides a robust, community-driven solution that integrates seamlessly into modern development workflows, especially for enterprise and open-source projects. _themes: c++ · package-management · cmake · dependencies_ #### [microsoft/data-formulator](https://github.com/microsoft/data-formulator) *TypeScript · ★15,226 · MIT · experimental · score:0.70 · hot:0.77 · rising:0.80 · durable:0.74 · board:rising · trend:up* 🪄 Create rich visualizations with AI **Why it matters.** Data Formulator is an AI-powered tool that uses agents to generate and explore data visualizations from user inputs, incorporating features like chat integration and a unified agent for recommendations. It matters right now because it leverages AI to make data visualization more accessible to non-experts amid growing demand for AI-driven analytics, but its alpha status raises concerns about reliability and full functionality in real-world applications. _themes: agents · visualization · ai-data · chat_ #### [deepfakes/faceswap](https://github.com/deepfakes/faceswap) *Python · ★55,186 · GPL-3.0 · production · score:0.80 · hot:0.76 · rising:0.84 · durable:0.81 · board:rising · trend:up* Deepfakes Software For All **Why it matters.** This repository offers an open-source tool for swapping faces in images and videos using deep learning techniques, requiring users to handle their own data extraction, training, and conversion processes. It matters now due to the increasing prevalence of deepfake technology in media manipulation, raising ethical concerns about misinformation, while also serving as a benchmark for AI ethics and generative model development in research and applications. _themes: deep-learning · face-swap · generative-ai · ethics_ #### [openinterpreter/open-interpreter](https://github.com/openinterpreter/open-interpreter) *Python · ★63,264 · AGPL-3.0 · beta · score:0.80 · hot:0.76 · rising:0.82 · durable:0.74 · board:rising · trend:up* A natural language interface for computers **Why it matters.** Open Interpreter is a tool that enables large language models to execute code locally in languages like Python, JavaScript, and Shell through a natural language interface, allowing users to perform tasks such as data analysis, file editing, and browser automation with user approval for safety. It matters right now because it bridges the gap between conversational AI and practical computing, empowering users to automate everyday tasks amid the growing adoption of AI agents, though it requires caution due to potential security risks from code execution. _themes: agents · llm · automation · code-execution_ #### [microsoft/typespec](https://github.com/microsoft/typespec) *Java · ★5,696 · MIT · production · score:0.80 · hot:0.76 · rising:0.79 · durable:0.70 · board:rising · trend:up* **Why it matters.** TypeSpec is a language for defining cloud service APIs and generating related assets like code, documentation, and schemas from a single source, supporting protocols such as REST, OpenAPI, and gRPC. It matters now because it addresses the growing need for standardized, reusable API designs in cloud-native development, helping teams enforce best practices and reduce inconsistencies amid rapid API proliferation. This Microsoft-backed tool promotes efficiency in API ecosystems where interoperability and maintainability are critical. _themes: api-design · code-generation · cloud-services · linter_ #### [lukasmasuch/best-of-ml-python](https://github.com/lukasmasuch/best-of-ml-python) *? · ★23,429 · CC-BY-SA-4.0 · production · score:0.85 · hot:0.76 · rising:0.81 · durable:0.82 · board:durable · trend:up* 🏆 A ranked list of awesome machine learning Python libraries. Updated weekly. **Why it matters.** This repository curates and ranks over 900 open-source machine learning Python libraries across various categories, using a quality score based on GitHub metrics, and updates weekly to keep the list current amid rapid ML advancements. It matters because it helps developers and researchers efficiently discover reliable tools in a crowded ecosystem, though its rankings depend on potentially subjective metrics and may not always reflect the latest innovations accurately. _themes: machine-learning · python · curation · discovery_ #### [microsoft/vscode-js-debug](https://github.com/microsoft/vscode-js-debug) *TypeScript · ★1,937 · MIT · production · score:0.85 · hot:0.75 · rising:0.78 · durable:0.68 · board:rising · trend:up* A DAP-compatible JavaScript debugger. Used in VS Code, VS, + more **Why it matters.** microsoft/vscode-js-debug is a Debug Adapter Protocol (DAP)-based debugger for JavaScript, supporting environments like Node.js, Chrome, and React Native, and serving as the default debugger in VS Code and Visual Studio. It matters because effective debugging is crucial for modern web and app development, and this tool's integration into widely used IDEs enhances productivity for developers working on complex JavaScript projects, especially with its active updates and standalone usability. _themes: debug · javascript · vscode · dap_ #### [microsoft/playwright-cli](https://github.com/microsoft/playwright-cli) *TypeScript · ★8,838 · Apache-2.0 · beta · score:0.70 · hot:0.75 · rising:0.77 · durable:0.73 · board:rising · trend:up* CLI for common Playwright actions. Record and generate Playwright code, inspect selectors and take screenshots. **Why it matters.** Playwright CLI is a command-line tool that simplifies common Playwright actions like recording code, inspecting selectors, and taking screenshots, specifically designed for integration with AI coding agents to enhance efficiency. It matters now because the growing use of LLMs in development workflows demands token-efficient tools to avoid context overload, enabling smoother automation in resource-constrained environments. However, its early version and narrow focus on agents mean it's not a universal solution yet. _themes: agents · cli · automation · playwright_ #### [microsoft/kiota](https://github.com/microsoft/kiota) *C# · ★3,708 · MIT · production · score:0.80 · hot:0.75 · rising:0.78 · durable:0.69 · board:rising · trend:up* OpenAPI based HTTP Client code generator **Why it matters.** Kiota is a command-line tool that generates strongly typed HTTP client code from OpenAPI specifications for languages like C#, Go, Java, and others, aiming to reduce reliance on multiple API SDKs. It matters now because the growing adoption of OpenAPI in API-driven development makes automated, consistent client generation essential for efficiency and maintainability, especially in enterprise environments where Microsoft tools are prevalent. _themes: openapi · codegen · api-client · http_ #### [sansan0/TrendRadar](https://github.com/sansan0/TrendRadar) *Python · ★52,322 · GPL-3.0 · beta · score:0.75 · hot:0.75 · rising:0.79 · durable:0.76 · board:rising · trend:up* ⭐AI-driven public opinion & trend monitor with multi-platform aggregation, RSS, and smart alerts.🎯 告别信息过载，你的 AI 舆情监控助手与热点筛选工具！聚合多平台热点 + RSS 订阅，支持关键词精准筛选。AI 智能筛选新闻 + AI 翻译 + AI 分析简报直推手机，也支持接入 MCP 架构，赋能 AI 自然语言对话分析、情感洞察与趋势预测等。支持 Docker ，数据本地/云端自持。集成微信/飞书/钉钉/Telegram/邮件/ntfy/bark/slack 等渠道智能推送。 **Why it matters.** TrendRadar is an AI-driven tool that aggregates news from multiple platforms, uses AI for filtering, translation, and analysis, and sends alerts via various channels like WeChat and Telegram, making it easier to monitor public opinion and trends. It matters now because information overload is a growing issue in the digital age, and this tool provides accessible, customizable monitoring with AI enhancements, though its reliance on external APIs like newsnow could pose stability risks. However, its high star count suggests strong community interest, but the lack of formal releases raises questions about long-term maintenance. _themes: ai-analysis · rss-aggregation · alerts · trending_ #### [Snailclimb/JavaGuide](https://github.com/Snailclimb/JavaGuide) *Java · ★155,067 · Apache-2.0 · production · score:0.80 · hot:0.75 · rising:0.80 · durable:0.79 · board:rising · trend:up* Java 面试 & 后端通用面试指南，覆盖计算机基础、数据库、分布式、高并发、系统设计与 AI 应用开发 **Why it matters.** This repository provides detailed guides and resources for Java backend interviews, covering foundational topics like JVM, databases, and distributed systems, as well as emerging areas such as AI agents, RAG, and prompt engineering. It matters right now because the job market demands strong backend skills combined with AI knowledge, offering practical preparation for developers amid rapid AI adoption in enterprises, though its content is more tutorial-based and less innovative. _themes: interview · java · agents · system-design_ #### [microsoft/Conversation-Knowledge-Mining-Solution-Accelerator](https://github.com/microsoft/Conversation-Knowledge-Mining-Solution-Accelerator) *Python · ★439 · MIT · production · score:0.70 · hot:0.75 · rising:0.77 · durable:0.68 · board:rising · trend:up* This solution accelerator leverages Microsoft Foundry, Azure Content Understanding, Azure OpenAI Service, and Foundry IQ to enable organizations to derive insights from volumes of conversational data using generative AI. It offers key phrase extraction, topic modeling, and interactive chat experiences through an intuitive web interface. **Why it matters.** This repository offers a solution accelerator that uses Azure services and generative AI to extract insights from conversational data, including key phrase extraction, topic modeling, and interactive chat interfaces, helping organizations analyze unstructured dialogue for better decision-making. It matters now as businesses face growing volumes of conversational data from customer interactions, and tools like this enable efficient pattern detection and insight generation amid the AI boom, though it relies heavily on Microsoft ecosystems which may limit broader adoption. _themes: generative-ai · nlp · topic-modeling · chat-interface_ #### [microsoft/apm](https://github.com/microsoft/apm) *Python · ★1,881 · MIT · beta · score:0.70 · hot:0.75 · rising:0.74 · durable:0.66 · board:hot · trend:up* Agent Package Manager **Why it matters.** APM is an open-source tool that manages dependencies for AI agents, allowing developers to define configurations in a YAML file for easy setup, portability, and reproducibility, similar to traditional package managers. It matters now because the rapid adoption of AI coding assistants like GitHub Copilot requires standardized ways to handle agent contexts, skills, and plugins, reducing manual setup time and improving collaboration in AI-driven development projects. _themes: agents · package-manager · ai-dependencies · prompt-engineering_ #### [NVIDIA/nvidia-container-toolkit](https://github.com/NVIDIA/nvidia-container-toolkit) *Go · ★4,262 · Apache-2.0 · production · score:0.85 · hot:0.74 · rising:0.78 · durable:0.72 · board:rising · trend:up* Build and run containers leveraging NVIDIA GPUs **Why it matters.** The NVIDIA Container Toolkit enables users to build and run containers that leverage NVIDIA GPUs by providing a runtime library and utilities for automatic configuration, making it easier to integrate GPU acceleration into containerized workflows. It matters right now because the growing demand for AI, machine learning, and high-performance computing in cloud-native environments requires efficient GPU utilization, and this toolkit addresses compatibility and performance challenges in production setups. _themes: containers · gpu-acceleration · nvidia · cloud-native_ #### [huggingface/chat-ui](https://github.com/huggingface/chat-ui) *TypeScript · ★10,660 · Apache-2.0 · beta · score:0.80 · hot:0.74 · rising:0.77 · durable:0.74 · board:rising · trend:up* The open source codebase powering HuggingChat **Why it matters.** HuggingFace's chat-ui repo provides an open-source SvelteKit-based frontend for building chat interfaces with large language models, specifically supporting OpenAI-compatible APIs for services like Hugging Face's router or Ollama. It matters now because it simplifies creating AI chat applications amid growing demand for customizable LLMs, but its limitations to OpenAI protocols and removal of legacy features may require users to adapt for broader integrations, potentially reducing its versatility compared to more comprehensive tools. _themes: llm · chat · inference · frontend_ #### [microsoft/vcpkg-tool](https://github.com/microsoft/vcpkg-tool) *C++ · ★577 · MIT · beta · score:0.70 · hot:0.74 · rising:0.75 · durable:0.62 · board:rising · trend:up* Components of microsoft/vcpkg's binary. **Why it matters.** Vcpkg-tool is a component of Microsoft's vcpkg ecosystem that handles C and C++ library artifacts via manifests, integrating into shells for dependency management on Windows, Linux, and MacOS. It matters now as it aims to streamline C++ development workflows in a preview state, potentially reducing build errors and improving portability, but its evolving nature means users should expect changes based on feedback. _themes: c++ · package-management · dependencies · build-tools_ #### [microsoft/Web-Dev-For-Beginners](https://github.com/microsoft/Web-Dev-For-Beginners) *JavaScript · ★95,626 · MIT · production · score:0.90 · hot:0.74 · rising:0.82 · durable:0.80 · board:rising · trend:up* 24 Lessons, 12 Weeks, Get Started as a Web Developer **Why it matters.** This repository offers a 12-week curriculum with 24 lessons on web development fundamentals, including HTML, CSS, and JavaScript, through hands-on projects and quizzes, aimed at absolute beginners. It matters right now because the demand for basic web development skills is growing in the job market, and this free, structured resource from Microsoft provides an accessible entry point, though its lack of updates since the last release might limit its relevance to emerging web technologies. _themes: education · web-dev · tutorials · beginners_ #### [huggingface/hf-mount](https://github.com/huggingface/hf-mount) *Rust · ★674 · Apache-2.0 · beta · score:0.70 · hot:0.73 · rising:0.73 · durable:0.66 · board:hot · trend:up* Mount Hugging Face Buckets and repos as local filesystems. No download, no copy, no waiting. **Why it matters.** hf-mount is a Rust-based tool that mounts Hugging Face Buckets and repositories as local filesystems using FUSE or NFS, allowing on-demand data access without prior downloads, which saves time and storage. It matters now because the increasing size of AI models and datasets demands efficient remote access solutions, enabling seamless integration for developers working with Hugging Face resources in resource-constrained environments. This lazy-loading approach enhances productivity by supporting standard file operations without complex APIs. _themes: storage · filesystem · inference · lazy-loading_ #### [microsoft/component-detection](https://github.com/microsoft/component-detection) *C# · ★540 · MIT · production · score:0.70 · hot:0.73 · rising:0.74 · durable:0.64 · board:rising · trend:up* Scans your project to determine what components you use **Why it matters.** Component Detection is a C#-based tool that scans software projects to identify and catalog dependencies across various ecosystems, generating a graph-based output for Software Bill of Materials (SBOMs). It matters now due to increasing regulatory requirements for supply chain security, such as those from the US Executive Order on cybersecurity, helping developers and organizations mitigate risks from third-party components. _themes: sbom · dependencies · static-analysis · supply-chain_ #### [microsoft/OmniParser](https://github.com/microsoft/OmniParser) *Jupyter Notebook · ★24,663 · CC-BY-4.0 · beta · score:0.80 · hot:0.71 · rising:0.77 · durable:0.78 · board:durable · trend:up* A simple screen parsing tool towards pure vision based GUI agent **Why it matters.** OmniParser is a tool that parses screenshots of user interfaces into structured elements, enabling vision-based AI agents like GPT-4V to accurately interact with GUIs by grounding actions in specific regions. It matters now because the growing demand for autonomous AI agents in real-world applications requires reliable vision-to-action pipelines, and this tool addresses limitations in multimodal AI by integrating with various LLMs and providing utilities like OmniTool for Windows VM control, though its documentation is still evolving. _themes: agents · vision · inference · gui-parsing_ #### [microsoft/prompty](https://github.com/microsoft/prompty) *TypeScript · ★1,199 · MIT · beta · score:0.70 · hot:0.71 · rising:0.72 · durable:0.72 · board:rising · trend:stable* Prompty makes it easy to create, manage, debug, and evaluate LLM prompts for your AI applications. Prompty is an asset class and format for LLM prompts designed to enhance observability, understandability, and portability for developers. **Why it matters.** Prompty provides a markdown-based file format for defining, managing, and executing LLM prompts, enabling easier debugging and evaluation across tools like VS Code, Python, and TypeScript. It matters now because prompt engineering is a growing bottleneck in AI development, and this tool addresses portability and observability issues, though it's still in early stages and may not fully resolve real-world complexities. However, its standardization could help developers build more maintainable applications amid the rapid adoption of LLMs. _themes: promptengineering · llms · evaluation_ #### [microsoft/edit](https://github.com/microsoft/edit) *Rust · ★13,733 · MIT · production · score:0.70 · hot:0.70 · rising:0.76 · durable:0.77 · board:durable · trend:up* We all edit. **Why it matters.** Microsoft Edit is a simple terminal-based text editor written in Rust, inspired by classic editors like MS-DOS Editor, offering a modern interface for basic text editing tasks accessible to beginners. It matters now as it fills a niche for lightweight, easy-to-use tools in an era of feature-heavy IDEs, potentially appealing to users in educational or introductory contexts, though it doesn't introduce groundbreaking innovations. _themes: editor · terminal · rust · accessibility_ #### [microsoft/avml](https://github.com/microsoft/avml) *Rust · ★1,074 · MIT · production · score:0.60 · hot:0.70 · rising:0.73 · durable:0.67 · board:rising · trend:up* AVML - Acquire Volatile Memory for Linux **Why it matters.** AVML is a Rust-based tool for acquiring volatile memory on Linux systems, enabling users to capture memory dumps from sources like /dev/crash or /proc/kcore without needing prior knowledge of the OS distribution, making it portable and easy to deploy as a static binary. It matters now due to the growing emphasis on memory forensics in cybersecurity for incident response and malware analysis, especially as Linux is prevalent in enterprise and cloud environments, though its utility is limited by kernel lockdown features. _themes: memory-forensics · linux-security · rust_ #### [Developer-Y/cs-video-courses](https://github.com/Developer-Y/cs-video-courses) *? · ★80,316 · no-license · production · score:0.85 · hot:0.70 · rising:0.76 · durable:0.73 · board:rising · trend:up* List of Computer Science courses with video lectures. **Why it matters.** This repository compiles a curated list of university-level computer science courses with video lectures, focusing on topics like algorithms, machine learning, and systems programming, while excluding basic tutorials or advertisements. It matters now as accessible, high-quality educational resources are essential for self-learners in a rapidly evolving tech landscape, though its lack of a license and reliance on external links may pose maintenance and accessibility issues over time. _themes: education · cs-courses · video-lectures · open-source_ #### [anthropics/knowledge-work-plugins](https://github.com/anthropics/knowledge-work-plugins) *Python · ★11,361 · Apache-2.0 · beta · score:0.80 · hot:0.70 · rising:0.75 · durable:0.71 · board:rising · trend:up* Open source repository of plugins primarily intended for knowledge workers to use in Claude Cowork **Why it matters.** This repository provides open-source Python plugins for Anthropic's Claude AI, enabling knowledge workers to integrate it with tools like Slack and Notion for tasks in productivity, sales, and customer support, effectively turning the AI into a customized assistant for specific roles. It matters now because businesses are rapidly adopting AI for workflow automation, but the lack of a formal latest release raises questions about maintenance and stability, potentially limiting its immediate reliability compared to more polished alternatives. _themes: agents · plugins · productivity · integrations_ #### [khoj-ai/khoj](https://github.com/khoj-ai/khoj) *Python · ★34,158 · AGPL-3.0 · beta · score:0.75 · hot:0.70 · rising:0.71 · durable:0.79 · board:durable · trend:stable* Your AI second brain. Self-hostable. Get answers from the web or your docs. Build custom agents, schedule automations, do deep research. Turn any online or local LLM into your personal, autonomous AI (gpt, claude, gemini, llama, qwen, mistral). Get started - free. **Why it matters.** Khoj is a self-hostable AI assistant that integrates with various LLMs to enable chatting, document search, custom agent creation, and automation for personal productivity and research. It matters now due to increasing interest in privacy-focused, offline-capable AI tools amid rising data concerns, allowing users to leverage advanced RAG and semantic search without depending on proprietary cloud services. With its support for local and online models, it empowers users to build autonomous AI workflows in an era of rapid LLM proliferation. _themes: rag · agents · inference · productivity_ #### [anthropics/claude-code-action](https://github.com/anthropics/claude-code-action) *TypeScript · ★7,160 · MIT · beta · score:0.75 · hot:0.70 · rising:0.74 · durable:0.71 · board:rising · trend:up* **Why it matters.** This repository provides a GitHub action that integrates Anthropic's Claude AI to automate code-related tasks in pull requests and issues, such as answering questions, reviewing code, and implementing changes, with intelligent context-based activation. It matters now because it enhances developer productivity in an era of widespread AI adoption for coding, offering seamless integration with existing workflows amid growing demand for AI-assisted tools in software development. _themes: agents · github-actions · code-review · ai-assistants_ #### [anthropics/claude-plugins-official](https://github.com/anthropics/claude-plugins-official) *Python · ★17,304 · no-license · production · score:0.80 · hot:0.69 · rising:0.73 · durable:0.67 · board:rising · trend:up* Official, Anthropic-managed directory of high quality Claude Code Plugins. **Why it matters.** This repository serves as an official, Anthropic-managed directory for high-quality plugins that extend the capabilities of Claude Code, including custom commands, agents, and skills for AI-assisted development. It matters right now because as AI models like Claude become integral to application building, a curated plugin ecosystem provides trusted integrations that enhance productivity and security, though users must remain cautious of potential risks from third-party contributions. _themes: plugins · ai-assistants · skills · agents_ #### [microsoft/winget-pkgs](https://github.com/microsoft/winget-pkgs) *? · ★10,494 · MIT · production · score:0.80 · hot:0.69 · rising:0.75 · durable:0.66 · board:rising · trend:up* The Microsoft community Windows Package Manager manifest repository **Why it matters.** This repository hosts community-submitted manifest files for the Windows Package Manager, which is a command-line tool for installing and managing software on Windows by defining package details. It matters now because it standardizes and simplifies software distribution on Windows, addressing the need for efficient package management in enterprise and developer environments, though it's limited by its support for only specific installer types and lacks features like script-based installations. _themes: package-management · windows · community_ #### [NVIDIA/enroot](https://github.com/NVIDIA/enroot) *Shell · ★928 · Apache-2.0 · production · score:0.70 · hot:0.69 · rising:0.71 · durable:0.64 · board:rising · trend:up* A simple yet powerful tool to turn traditional container/OS images into unprivileged sandboxes. **Why it matters.** Enroot is a tool from NVIDIA that converts traditional container images into lightweight, unprivileged sandboxes using Linux kernel features like user and mount namespaces, reducing isolation overhead for high-performance computing environments. It matters now because as AI and ML workloads increasingly rely on efficient container management in HPC and GPU-accelerated setups, Enroot provides a simple, fast alternative to heavier tools, enabling better portability and reproducibility without unnecessary performance hits. _themes: containers · hpc · sandbox · gpu_ #### [microsoft/vscode-docs](https://github.com/microsoft/vscode-docs) *Markdown · ★6,466 · NOASSERTION · production · score:0.85 · hot:0.69 · rising:0.74 · durable:0.65 · board:rising · trend:up* Public documentation for Visual Studio Code **Why it matters.** This repository contains the official documentation for Visual Studio Code, a widely used code editor, covering setup, features, and extensions in Markdown format. It matters because clear documentation is crucial for developers to effectively use and contribute to VS Code, especially as AI integration in coding tools evolves, ensuring accessibility for a broad user base. However, as a docs repo, its impact is derivative of the main VS Code product and may not introduce new innovations. _themes: documentation · code-editor · open-source_ #### [NVIDIA/NemoClaw](https://github.com/NVIDIA/NemoClaw) *TypeScript · ★19,471 · Apache-2.0 · experimental · score:0.60 · hot:0.69 · rising:0.72 · durable:0.64 · board:rising · trend:up* Run OpenClaw more securely inside NVIDIA OpenShell with managed inference **Why it matters.** NVIDIA NemoClaw is a reference stack that enhances the security and management of OpenClaw AI assistants by integrating them with NVIDIA OpenShell, providing features like state management and routed inference. It matters right now as AI agent security is a growing concern amid increasing adoption of autonomous systems, but its alpha status means it's primarily useful for early feedback and experimentation rather than practical deployment. With no official release yet, it represents a timely but unproven effort in safe AI agent execution. _themes: agents · security · inference · runtime_ #### [anthropics/financial-services-plugins](https://github.com/anthropics/financial-services-plugins) *Python · ★7,661 · Apache-2.0 · production · score:0.70 · hot:0.69 · rising:0.75 · durable:0.71 · board:rising · trend:up* **Why it matters.** This repository provides Python-based plugins for Anthropic's Claude AI, enabling it to handle specialized financial services workflows like investment banking and equity research by integrating with data sources and custom processes. It matters because financial institutions are increasingly adopting AI to automate analysis and reduce errors in a highly regulated environment, but these plugins may require significant customization and could inherit risks from AI dependencies like hallucinations or data privacy issues. _themes: agents · plugins · finance · workflows_ #### [huggingface/smol-course](https://github.com/huggingface/smol-course) *Jupyter Notebook · ★6,633 · Apache-2.0 · beta · score:0.80 · hot:0.68 · rising:0.73 · durable:0.68 · board:rising · trend:up* A course on aligning smol models. **Why it matters.** This repository provides a practical, hands-on course for aligning small language models like SmolLM3 and SmolVLM2, covering topics from instruction tuning to evaluation and preference alignment, all runnable on local machines with minimal GPU needs. It matters now because it lowers the barrier to AI development amid growing interest in efficient, resource-light models, enabling users to adapt LLMs for specific applications without relying on expensive infrastructure, especially as alignment techniques become essential for safe and customized AI deployment. _themes: fine-tuning · alignment · evaluation · small-models_ #### [anthropics/claude-code-base-action](https://github.com/anthropics/claude-code-base-action) *TypeScript · ★807 · MIT · beta · score:0.60 · hot:0.68 · rising:0.71 · durable:0.67 · board:rising · trend:up* This repo is a mirror of the contents of base-action in https://github.com/anthropics/claude-code-action. **Why it matters.** This repository is a mirror of a GitHub Action that integrates Anthropic's Claude Code AI into workflows, allowing users to run AI-assisted coding tasks via prompts and tools within CI/CD pipelines. It matters now because AI-driven automation in development is increasingly essential for efficiency, though its value is limited as it's just a mirror and not the primary development source. Being in beta, it offers practical utility for developers experimenting with AI integrations but requires directing contributions to the main repo. _themes: ai · github-actions · code-assistant · automation_ #### [microsoft/typescript-go](https://github.com/microsoft/typescript-go) *Go · ★24,824 · Apache-2.0 · experimental · score:0.70 · hot:0.67 · rising:0.69 · durable:0.68 · board:rising · trend:up* Staging repo for development of native port of TypeScript **Why it matters.** This repository is developing a native Go implementation of TypeScript, aiming to replicate its type checking and compilation features for potentially better performance in Go-centric environments. It matters right now because it could enhance TypeScript's usability in high-performance applications, such as backend services, but it's still experimental and lacks full feature parity, so users should approach with caution. _themes: typescript · go · compiler · performance_ #### [microsoft/AI-For-Beginners](https://github.com/microsoft/AI-For-Beginners) *Jupyter Notebook · ★46,626 · MIT · production · score:0.85 · hot:0.67 · rising:0.75 · durable:0.77 · board:durable · trend:up* 12 Weeks, 24 Lessons, AI for All! **Why it matters.** This repository provides a 12-week curriculum with 24 lessons on artificial intelligence for beginners, including practical labs, quizzes, and coverage of tools like TensorFlow and PyTorch, as well as AI ethics. It matters right now because the rapid growth of AI demands accessible, free educational resources to build foundational skills, and its multi-language support enhances global reach amid increasing AI adoption in various fields. _themes: ai · machine-learning · deep-learning · education_ #### [huggingface/course](https://github.com/huggingface/course) *MDX · ★3,854 · Apache-2.0 · production · score:0.85 · hot:0.67 · rising:0.71 · durable:0.65 · board:rising · trend:up* The Hugging Face course on Transformers **Why it matters.** This repository contains the free, open-source course materials from Hugging Face on using Transformers for NLP and other tasks, covering their ecosystem libraries like Transformers and Datasets. It matters right now because Transformers are foundational to modern AI applications, and accessible educational resources are essential amid the growing demand for AI skills, helping beginners and professionals quickly upskill in a rapidly evolving field. _themes: transformers · nlp · deep-learning · fine-tuning_ #### [meta-llama/PurpleLlama](https://github.com/meta-llama/PurpleLlama) *Python · ★4,125 · NOASSERTION · experimental · score:0.70 · hot:0.67 · rising:0.68 · durable:0.60 · board:rising · trend:up* Set of tools to assess and improve LLM security. **Why it matters.** PurpleLlama provides a collection of tools and evaluations for assessing and improving the security of large language models, with an initial focus on cyber security and input/output safeguards. This is relevant now as the rapid adoption of generative AI increases risks like misuse and vulnerabilities, making responsible development tools essential for the community. However, its lack of a formal release and unclear licensing details may limit immediate usability. _themes: llm-security · evals · ai-safety · tools_ #### [apache/casbin-gateway](https://github.com/apache/casbin-gateway) *Go · ★557 · Apache-2.0 · production · score:0.60 · hot:0.67 · rising:0.67 · durable:0.62 · board:rising · trend:stable* Casbin AI & MCP security gateway for HTTP, online demo: https://door.caswaf.com **Why it matters.** Apache Casbin Gateway is an AI-enhanced web application firewall (WAF) for HTTP traffic, built on Casbin for access control and integrating tools like ModSecurity and LLMs for threat detection and authentication. It matters now as AI-driven applications grow, exposing new security vulnerabilities, but its implementation may add complexity without proving superior to established WAFs in all scenarios. _themes: waf · ai · llm · security_ #### [frankbria/ralph-claude-code](https://github.com/frankbria/ralph-claude-code) *Shell · ★8,742 · MIT · beta · score:0.70 · hot:0.66 · rising:0.69 · durable:0.68 · board:rising · trend:up* Autonomous AI development loop for Claude Code with intelligent exit detection **Why it matters.** Ralph for Claude Code automates iterative AI-driven development loops using Claude, enabling projects to improve autonomously with safeguards like intelligent exit detection and rate limiting to prevent infinite loops and API overuse. It matters now as AI-assisted coding tools are increasingly adopted to boost developer productivity amid the generative AI boom, though its beta status means it's still refining core features. This could streamline workflows but requires caution due to ongoing development. _themes: agents · ai-development · automation · cli-tools_ #### [microsoft/RustTraining](https://github.com/microsoft/RustTraining) *Rust · ★13,897 · MIT · production · score:0.80 · hot:0.66 · rising:0.72 · durable:0.75 · board:durable · trend:stable* Beginner, advanced, expert level Rust training material **Why it matters.** This repository offers a collection of training materials for Rust programming, covering beginner to expert levels with courses on various backgrounds, async programming, and advanced patterns. It matters because Rust is increasingly vital for safe and efficient systems programming in industries like web development and embedded systems, and Microsoft's involvement provides high-quality, free educational resources that can accelerate learning. However, as it's not an official reference, users must cross-check with authoritative sources to ensure accuracy. _themes: rust · education · programming · async_ #### [microsoft/calculator](https://github.com/microsoft/calculator) *C++ · ★30,905 · MIT · production · score:0.30 · hot:0.66 · rising:0.72 · durable:0.63 · board:rising · trend:up* Windows Calculator: A simple yet powerful calculator that ships with Windows **Why it matters.** The Microsoft Calculator repo is an open-source implementation of the Windows Calculator app, offering features like standard, scientific, and programmer modes, along with unit conversions and infinite precision arithmetic. It matters right now as a practical example of C++ and C# in UWP development, providing insights for Windows app developers, though it's primarily a utility rather than an innovative tool in the AI/ML space. _themes: uwp · windows · c++ · csharp_ #### [allenai/olmocr](https://github.com/allenai/olmocr) *Python · ★17,149 · Apache-2.0 · beta · score:0.80 · hot:0.65 · rising:0.71 · durable:0.79 · board:durable · trend:stable* Toolkit for linearizing PDFs for LLM datasets/training **Why it matters.** This repo provides a toolkit for converting PDFs and image-based documents into clean, readable Markdown text, handling complex elements like equations and tables while maintaining a natural reading order, which is essential for preparing high-quality datasets for LLM training. It matters now because the proliferation of LLMs demands efficient, accurate document processing to overcome limitations in existing OCR tools, especially for multi-column layouts and figures. However, its reliance on a GPU and a specific 7B parameter model may limit accessibility for some users. _themes: ocr · document-processing · llm · inference_ #### [NVIDIA/pyxis](https://github.com/NVIDIA/pyxis) *C · ★428 · Apache-2.0 · production · score:0.70 · hot:0.65 · rising:0.67 · durable:0.62 · board:rising · trend:stable* Container plugin for Slurm Workload Manager **Why it matters.** Pyxis is a SPANK plugin for Slurm that enables unprivileged users to run containerized tasks on HPC clusters using the srun command, leveraging NVIDIA's Enroot for container management. It matters now because it simplifies container usage in resource-constrained environments like AI and scientific computing, where reproducibility and isolation are critical, especially with the growing adoption of GPU-accelerated workloads in Slurm-based systems. _themes: containers · hpc · slurm · mpi_ #### [huggingface/blog](https://github.com/huggingface/blog) *Jupyter Notebook · ★3,385 · no-license · production · score:0.70 · hot:0.65 · rising:0.68 · durable:0.59 · board:rising · trend:up* Public repo for HF blog posts **Why it matters.** This repository serves as the official GitHub home for Hugging Face's blog posts, allowing community members to contribute articles on AI topics via Markdown files and specific guidelines. It matters for disseminating practical AI knowledge and fostering open-source collaboration, especially through events like Hacktoberfest, but its lack of a license hinders broader reuse and integration, and the reliance on external image hosting adds unnecessary complexity. _themes: ai · blogging · community · open-source_ #### [NVIDIA/nvbandwidth](https://github.com/NVIDIA/nvbandwidth) *C++ · ★672 · Apache-2.0 · beta · score:0.60 · hot:0.65 · rising:0.65 · durable:0.60 · board:hot · trend:stable* A tool for bandwidth measurements on NVIDIA GPUs. **Why it matters.** NVIDIA/nvbandwidth is a C++ tool for measuring bandwidth of memory copy operations on NVIDIA GPUs, using methods like copy engines or kernel copies, which helps in system tuning for better performance. It matters for developers working with GPU-intensive applications, as bandwidth optimization is crucial in AI and high-performance computing, but its value is limited by the need for specific CUDA versions and additional dependencies, making it less accessible for casual users. _themes: gpu · performance · benchmark · cuda_ #### [microsoft/rushstack](https://github.com/microsoft/rushstack) *TypeScript · ★6,451 · NOASSERTION · production · score:0.80 · hot:0.64 · rising:0.67 · durable:0.59 · board:rising · trend:up* Monorepo for tools developed by the Rush Stack community **Why it matters.** Rushstack is a monorepo containing tools like Rush for build orchestration, Heft for task running, and API Extractor for TypeScript API management, designed to streamline development in large-scale TypeScript projects. It matters now because managing complex monorepos is a growing challenge in enterprise software, where efficient builds, dependency resolution, and documentation are essential for productivity, especially amid the increasing adoption of TypeScript and Node.js ecosystems. _themes: monorepo · typescript · build · toolchain_ #### [huggingface/gpu-fryer](https://github.com/huggingface/gpu-fryer) *Rust · ★380 · Apache-2.0 · beta · score:0.75 · hot:0.64 · rising:0.65 · durable:0.67 · board:durable · trend:stable* Where GPUs get cooked 👩‍🍳🔥 **Why it matters.** GPU Fryer is a Rust-based tool for stress testing NVIDIA GPUs to identify thermal throttling and performance degradation, particularly for ML inference and training workloads on multi-GPU systems. It matters now because reliable GPU performance is critical in scaling AI infrastructure amid growing demands for efficient training and inference, helping prevent costly downtime in production environments like those at Hugging Face. _themes: gpu · monitoring · performance · inference_ #### [microsoft/fabric-toolbox](https://github.com/microsoft/fabric-toolbox) *PowerShell · ★742 · MIT · beta · score:0.70 · hot:0.64 · rising:0.67 · durable:0.57 · board:rising · trend:up* Fabric toolbox is a repository of tools, accelerators, scripts, and samples to accelerate your success with Microsoft Fabric, brought to you by Fabric CAT. **Why it matters.** This repository provides a collection of PowerShell scripts, tools, and samples for monitoring and managing Microsoft Fabric, including cost analysis and deployment accelerators. It matters because Microsoft Fabric is an emerging platform for data analytics, and these resources can help users address practical challenges like cost optimization and reliability, though they are presented as examples without formal releases, potentially requiring customization. _themes: monitoring · deployment · data-management · scripts_ #### [huggingface/dataset-viewer](https://github.com/huggingface/dataset-viewer) *Python · ★856 · Apache-2.0 · production · score:0.70 · hot:0.64 · rising:0.68 · durable:0.65 · board:rising · trend:stable* Backend that powers the dataset viewer on Hugging Face dataset pages through a public API. **Why it matters.** This repository provides the backend API for Hugging Face's dataset viewer, enabling users to browse, filter, and analyze datasets on the platform. It matters now because the growing volume of machine learning datasets requires efficient tools for exploration, and this API supports the Hugging Face ecosystem by making data more accessible for research and development. _themes: datasets · api · machine-learning · data-visualization_ #### [openai/transformer-debugger](https://github.com/openai/transformer-debugger) *Python · ★4,108 · MIT · experimental · score:0.70 · hot:0.64 · rising:0.66 · durable:0.64 · board:rising · trend:stable* **Why it matters.** Transformer Debugger is a tool from OpenAI that helps researchers investigate and debug behaviors in small language models by allowing interventions in the forward pass and providing explanations of components like neurons and attention heads. It matters now because AI interpretability and safety are critical amid growing concerns about model alignment, offering practical methods to understand and mitigate issues in transformer-based systems without extensive coding. However, its lack of a formal release and focus on small models limit its immediate applicability to larger, production-scale models. _themes: interpretability · debugging · transformers · autoencoders_ #### [huggingface/hub-docs](https://github.com/huggingface/hub-docs) *Handlebars · ★541 · Apache-2.0 · production · score:0.60 · hot:0.64 · rising:0.68 · durable:0.57 · board:rising · trend:up* Docs of the Hugging Face Hub **Why it matters.** This repository contains the documentation for the Hugging Face Hub, a platform for sharing and collaborating on machine learning models, including guides on usage and contributions. It matters because high-quality docs are essential for onboarding users in the rapidly growing ML community, but as a straightforward docs repo, it doesn't introduce new innovations and primarily serves as a maintenance tool for existing users. _themes: documentation · machine-learning · open-source · community_ #### [huggingface/autotrain-advanced](https://github.com/huggingface/autotrain-advanced) *Python · ★4,569 · Apache-2.0 · archived · score:0.20 · hot:0.64 · rising:0.64 · durable:0.53 · board:rising · trend:up* 🤗 AutoTrain Advanced **Why it matters.** Hugging Face's AutoTrain Advanced is a no-code tool designed to simplify the training and deployment of machine learning models, particularly for LLM fine-tuning tasks like SFT, ORPO, and DPO. It matters less now because the project is no longer maintained, with no new features or bug fixes, directing users to alternatives for reliable ongoing use in the evolving ML field. _themes: fine-tuning · llm · no-code · huggingface_ #### [google-research/arxiv-latex-cleaner](https://github.com/google-research/arxiv-latex-cleaner) *Python · ★6,796 · Apache-2.0 · production · score:0.75 · hot:0.63 · rising:0.68 · durable:0.74 · board:durable · trend:stable* arXiv LaTeX Cleaner: Easily clean the LaTeX code of your paper to submit to arXiv **Why it matters.** This repository offers a Python-based tool that automates the cleaning of LaTeX code by removing auxiliary files, comments, and other unnecessary elements, making it easier to prepare papers for arXiv submission. It matters now because LaTeX remains a staple in academic publishing, and with the growing volume of research submissions, tools like this save time and reduce errors for authors dealing with complex documents. _themes: latex · automation · publishing · scripting_ #### [microsoft/Mastering-GitHub-Copilot-for-Paired-Programming](https://github.com/microsoft/Mastering-GitHub-Copilot-for-Paired-Programming) *Python · ★7,791 · MIT · production · score:0.70 · hot:0.63 · rising:0.70 · durable:0.72 · board:durable · trend:stable* A multi-module course teaching everything you need to know about using GitHub Copilot as an AI Peer Programming resource. **Why it matters.** This repository offers a multi-module course that teaches developers how to effectively use GitHub Copilot for paired programming, covering topics like code generation, problem-solving, and workflow automation across languages such as Python, C#, and JavaScript. It matters right now because AI coding assistants like Copilot are increasingly integrated into development workflows, helping to boost productivity amid the rapid adoption of generative AI tools, and this resource provides practical, hands-on training to leverage these capabilities. _themes: ai-assistants · coding · tutorials · pair-programming_ #### [microsoft/SqlNexus](https://github.com/microsoft/SqlNexus) *C# · ★416 · MIT · production · score:0.60 · hot:0.63 · rising:0.66 · durable:0.60 · board:rising · trend:stable* SQL Nexus is a tool that helps you identify the root cause of SQL Server performance issues. It loads and analyzes performance data collected by SQL LogScout, SQLDiag or PSSDiag. It can dramatically reduce the amount of time you spend manually analyzing data. **Why it matters.** SQL Nexus is a Microsoft-developed tool that loads and analyzes performance data from SQL Server diagnostics like SQL LogScout or PSSDiag to identify root causes of performance issues, reducing manual analysis time. It matters now because database performance problems are increasingly common in enterprise environments with growing data volumes, and this tool provides a structured, efficient approach for IT professionals. However, its reliance on specific Microsoft tools and potential limitations in handling diverse data sources may require users to adapt it carefully. _themes: sql-server · performance · analysis · diagnostics_ #### [microsoft/amplifier](https://github.com/microsoft/amplifier) *Shell · ★3,056 · MIT · experimental · score:0.60 · hot:0.63 · rising:0.65 · durable:0.60 · board:rising · trend:stable* **Why it matters.** Amplifier is an AI-powered CLI tool that provides modular assistance for development tasks, allowing users to extend its architecture for custom AI experiences. It matters now as it explores AI integration in command-line workflows amid growing interest in AI coding tools, though its early, unstable state means it's more of a research artifact than a reliable solution, requiring users to handle security and supervision themselves. _themes: agents · cli · ai-tools · modular_ #### [huggingface/audio-transformers-course](https://github.com/huggingface/audio-transformers-course) *MDX · ★491 · Apache-2.0 · production · score:0.70 · hot:0.62 · rising:0.65 · durable:0.59 · board:rising · trend:stable* The Hugging Face Course on Transformers for Audio **Why it matters.** This repository contains the free, open-source content for Hugging Face's course on applying Transformers to audio and speech processing tasks, including tutorials and examples. It matters now because Transformers are increasingly used in audio applications like speech recognition and generation, providing accessible learning resources amid growing demand for AI skills in this area, though it lacks formal releases or extensive updates. _themes: audio · transformers · deep-learning · course_ #### [microsoft/agentrc](https://github.com/microsoft/agentrc) *TypeScript · ★785 · MIT · experimental · score:0.70 · hot:0.62 · rising:0.62 · durable:0.55 · board:rising · trend:stable* Get your repo ready for AI. **Why it matters.** AgentRC is a TypeScript-based tool that analyzes repositories to generate and maintain context files for AI coding agents, helping them better understand codebases, conventions, and dependencies. It matters right now because AI assistants are increasingly used in development, but they often lack specific repo knowledge, leading to suboptimal outputs; this tool addresses that by automating context creation and evaluation, potentially improving accuracy and efficiency in AI-driven coding workflows. _themes: agents · code-analysis · ai-readiness · eval_ #### [huggingface/aisheets](https://github.com/huggingface/aisheets) *TypeScript · ★1,627 · Apache-2.0 · beta · score:0.70 · hot:0.61 · rising:0.62 · durable:0.65 · board:durable · trend:stable* Build, enrich, and transform datasets using AI models with no code **Why it matters.** AI Sheets is a no-code tool from Hugging Face that allows users to build, enrich, and transform datasets using AI models, including LLMs, by integrating with the Hugging Face Hub or local deployments. It matters now because the growing need for accessible data preparation in AI projects, especially for synthetic data and LLM evaluation, makes it relevant amid the no-code trend, though its lack of a formal release raises questions about stability and readiness. _themes: nocode · llms · synthetic-data · inference_ #### [meta-llama/prompt-ops](https://github.com/meta-llama/prompt-ops) *Python · ★800 · MIT · beta · score:0.70 · hot:0.61 · rising:0.63 · durable:0.59 · board:rising · trend:stable* An open-source tool for LLM prompt optimization. **Why it matters.** Prompt-ops is a Python tool that automatically optimizes prompts for Llama models using data-driven methods and a new technique called Prompt Duel Optimizer, which improves performance on benchmarks without manual tweaking. It matters now because prompt engineering is a critical bottleneck in LLM development, and this tool offers efficient, label-free optimization amid growing demand for reliable AI applications, potentially saving time for developers working with large-scale models. _themes: prompt-optimization · llm · inference_ #### [huggingface/ai-deadlines](https://github.com/huggingface/ai-deadlines) *TypeScript · ★336 · MIT · beta · score:0.70 · hot:0.61 · rising:0.63 · durable:0.60 · board:rising · trend:stable* ⏰ AI conference deadline countdowns **Why it matters.** This repository provides a web application that displays countdowns for submission deadlines of top AI conferences like NeurIPS and ICLR, helping researchers track and manage their paper submissions efficiently. It matters now because the AI research community is in the midst of frequent conference cycles, where missing deadlines can hinder publication opportunities, and this tool offers an up-to-date alternative to unmaintained resources. The app automates data fetching via AI agents, ensuring reliability in a fast-paced environment. _themes: deadlines · conferences · AI-research · automation_ #### [NVIDIA/hpc-container-maker](https://github.com/NVIDIA/hpc-container-maker) *Python · ★512 · Apache-2.0 · beta · score:0.70 · hot:0.61 · rising:0.62 · durable:0.54 · board:rising · trend:stable* HPC Container Maker **Why it matters.** HPC Container Maker is a Python tool that generates Dockerfiles or Singularity definition files for high-performance computing from high-level recipes, abstracting away low-level details and incorporating best practices. It matters now because containerization in HPC is increasingly essential for reproducible workflows and resource efficiency, but its lack of recent releases raises questions about ongoing maintenance and adaptability to evolving container ecosystems. _themes: hpc · containers · automation · python_ #### [microsoft/language-server-protocol](https://github.com/microsoft/language-server-protocol) *HTML · ★12,749 · CC-BY-4.0 · production · score:0.80 · hot:0.60 · rising:0.66 · durable:0.65 · board:rising · trend:stable* Defines a common protocol for language servers. **Why it matters.** The Language Server Protocol (LSP) defines a standardized JSON-RPC based interface for code editors and IDEs to interact with language-specific servers, enabling features like code completion, diagnostics, and refactoring without tying them to a particular editor. It matters because it promotes interoperability in developer tools, allowing efficient reuse of language services across platforms, though its relevance persists amid evolving AI-assisted coding tools that could build upon or extend it. However, as a mature specification, it lacks active releases, potentially limiting adaptations to new trends. _themes: protocols · ide · development · standards_ #### [zyddnys/manga-image-translator](https://github.com/zyddnys/manga-image-translator) *Python · ★9,756 · GPL-3.0 · beta · score:0.75 · hot:0.60 · rising:0.63 · durable:0.72 · board:durable · trend:stable* Translate manga/image 一键翻译各类图片内文字 https://cotrans.touhou.ai/ (no longer working) **Why it matters.** This repository provides a tool for one-click translation of text in images, particularly manga, using OCR, machine translation, and image inpainting to detect, translate, and repair text; it matters now because the rising popularity of global anime and manga content demands accessible translation tools, though its beta status means it's not yet reliable for professional use. _themes: ocr · machine-translation · image-processing · deep-learning_ #### [openai/openai-security-bots](https://github.com/openai/openai-security-bots) *Python · ★383 · MIT · beta · score:0.50 · hot:0.60 · rising:0.61 · durable:0.57 · board:rising · trend:stable* **Why it matters.** This repository offers Slack bots that integrate OpenAI APIs to automate security workflows, such as incident triage, response, and SDLC tasks, primarily for security teams. It matters because it showcases AI-assisted automation in cybersecurity, which could enhance efficiency amid rising threats, but its lack of releases and modest 383 stars indicate it may not be actively maintained or widely adopted. _themes: agents · security · automation · ai_ #### [huggingface/llm-ls](https://github.com/huggingface/llm-ls) *Rust · ★865 · Apache-2.0 · experimental · score:0.60 · hot:0.59 · rising:0.59 · durable:0.61 · board:durable · trend:stable* LSP server leveraging LLMs for code completion (and more?) **Why it matters.** Huggingface/llm-ls is an experimental LSP server that uses large language models for code completion by integrating with various backends like Hugging Face and OpenAI, while handling prompt context and AST parsing to improve suggestions. It matters now because AI-assisted coding is gaining traction, but this project is still in early development with potential bugs, making it a promising yet unreliable option for developers seeking lightweight LLM integrations in IDEs. _themes: llm · code-completion · inference · lsp_ #### [huggingface/Mongoku](https://github.com/huggingface/Mongoku) *Svelte · ★1,408 · MIT · production · score:0.75 · hot:0.59 · rising:0.64 · durable:0.67 · board:durable · trend:stable* 🔥The Web-scale GUI for MongoDB **Why it matters.** Mongoku is a web-based GUI for MongoDB that allows users to query and manage databases directly from their browser, with features like fast operations on large datasets and document mappings. It matters now as a straightforward alternative for teams handling big data, given its use in production at Hugging Face, but it doesn't introduce groundbreaking innovations beyond existing tools and may face competition in usability and integration. _themes: mongodb · gui · database · admin_ #### [huggingface/optimum-benchmark](https://github.com/huggingface/optimum-benchmark) *Python · ★336 · Apache-2.0 · beta · score:0.70 · hot:0.59 · rising:0.59 · durable:0.69 · board:durable · trend:stable* 🏋️ A unified multi-backend utility for benchmarking Transformers, Timm, PEFT, Diffusers and Sentence-Transformers with full support of Optimum's hardware optimizations & quantization schemes. **Why it matters.** Hugging Face's optimum-benchmark provides a unified tool for benchmarking various AI libraries like Transformers and Diffusers, supporting multiple backends and hardware optimizations for inference and training. It matters now because optimizing model performance amid rising computational costs is critical for efficient AI deployment, especially with features like quantization and new backends for tools like vLLM enhancing practical usability. _themes: benchmark · inference · optimization · quantization_ #### [EgoAlpha/prompt-in-context-learning](https://github.com/EgoAlpha/prompt-in-context-learning) *Jupyter Notebook · ★2,222 · MIT · beta · score:0.80 · hot:0.59 · rising:0.60 · durable:0.66 · board:durable · trend:stable* Awesome resources for in-context learning and prompt engineering: Mastery of the LLMs such as ChatGPT, GPT-3, and FlanT5, with up-to-date and cutting-edge updates. **Why it matters.** This repository curates resources, papers, and guides for in-context learning and prompt engineering with LLMs like ChatGPT and GPT-3, including practical examples and experimentation tools. It matters now because prompt engineering is a critical skill for optimizing AI models in real-world applications, especially as LLMs become more prevalent and accessible, helping users adapt quickly to ongoing AI advancements. _themes: in-context-learning · prompt-engineering · llms · agents_ #### [microsoft/work-iq](https://github.com/microsoft/work-iq) *PowerShell · ★740 · NOASSERTION · beta · score:0.60 · hot:0.58 · rising:0.59 · durable:0.53 · board:rising · trend:stable* MCP Server and CLI for accessing Work IQ **Why it matters.** Microsoft Work IQ is a plugin marketplace that extends GitHub Copilot with tools to access Microsoft 365 data, requiring a CLI and server for setup and administrative consent. It matters now as enterprises seek to integrate AI with productivity suites for better data-driven workflows, but its public preview status and dependency on admin privileges limit immediate adoption and reliability. _themes: plugins · ai-integration · microsoft-365 · copilot_ #### [huggingface/meshgen](https://github.com/huggingface/meshgen) *Python · ★859 · MIT · beta · score:0.70 · hot:0.56 · rising:0.59 · durable:0.66 · board:durable · trend:stable* Use AI Agents directly in Blender. **Why it matters.** MeshGen is a Blender addon that integrates AI agents to control 3D modeling tasks via natural language, supporting local and remote inference backends for flexibility. It matters now because the growing adoption of AI in creative tools could streamline workflows for 3D artists, though its reliance on external models and hardware limits widespread accessibility amid the AI agent boom. _themes: agents · inference · 3d_ #### [microsoft/Ontology-Playground](https://github.com/microsoft/Ontology-Playground) *TypeScript · ★419 · MIT · beta · score:0.60 · hot:0.55 · rising:0.57 · durable:0.57 · board:durable · trend:stable* Free, open-source web app for learning about ontologies and Microsoft Fabric IQ. Explore a catalogue of pre-built ontologies, design your own visually, export as RDF/XML, and share interactive diagrams. Zero backend, fully static. **Why it matters.** Ontology Playground is a static web app that enables users to explore pre-built ontologies, design their own visually using a graph editor, and export them as RDF/XML, all without any backend dependencies. It matters now because ontologies are crucial for semantic web and AI applications like Microsoft Fabric IQ, providing an accessible entry point for learning and experimentation amid growing interest in knowledge graphs and data integration. However, as a preview with limited features and no official releases, it may not yet be robust for production use. _themes: ontology · semantic-web · visualization · rdf_ #### [openai/codex-action](https://github.com/openai/codex-action) *TypeScript · ★931 · Apache-2.0 · beta · score:0.70 · hot:0.54 · rising:0.56 · durable:0.61 · board:durable · trend:stable* **Why it matters.** This repo provides a GitHub Action for running OpenAI's Codex in workflows, focusing on secure integration via API keys and a proxy, which is useful for tasks like automated code reviews. It matters now as AI-assisted development tools are increasingly adopted, but this one is limited to OpenAI's ecosystem and lacks a formal release, potentially making it less reliable for production use compared to broader alternatives. _themes: github-actions · ai-integration · code-review · security_ #### [openai/model_spec](https://github.com/openai/model_spec) *? · ★779 · CC0-1.0 · production · score:0.80 · hot:0.53 · rising:0.59 · durable:0.68 · board:durable · trend:stable* The OpenAI Model Spec **Why it matters.** The OpenAI Model Spec is a document outlining the desired behaviors and guidelines for OpenAI's AI models to ensure safety, alignment, and ethical use. It matters right now as AI regulations and transparency demands intensify, providing a reference for developers and researchers to understand and critique OpenAI's approach. This repo archives the spec's source and versions, enabling public scrutiny but offering limited direct utility beyond documentation. _themes: ai-ethics · model-alignment · safety · documentation_ #### [huggingface/llm-vscode](https://github.com/huggingface/llm-vscode) *TypeScript · ★1,315 · Apache-2.0 · beta · score:0.60 · hot:0.53 · rising:0.57 · durable:0.65 · board:durable · trend:stable* LLM powered development for VSCode **Why it matters.** Huggingface/llm-vscode is a VSCode extension that integrates Large Language Models for features like code completion and attribution, relying on Hugging Face's backend for HTTP-based requests to models. It matters now as developers increasingly seek open-source alternatives to proprietary AI coding assistants like Copilot, but its early version (0.1.0) and free-tier limitations may hinder reliability for production use. While it promotes model flexibility and context-aware prompts, the moderate star count (1315) suggests it's not yet widely adopted or battle-tested. _themes: inference · code-completion · llm · attribution_ #### [huggingface/search-and-learn](https://github.com/huggingface/search-and-learn) *Python · ★1,130 · Apache-2.0 · experimental · score:0.70 · hot:0.52 · rising:0.54 · durable:0.59 · board:durable · trend:stable* Recipes to scale inference-time compute of open models **Why it matters.** This repository provides recipes for scaling inference-time compute in open models, allowing LLMs to improve performance on complex tasks by dynamically allocating more resources during inference, rather than relying on expensive retraining. It matters now because the rising costs of model pretraining make test-time optimization a practical alternative for enhancing capabilities, as demonstrated by models like OpenAI's o1, potentially enabling more efficient AI development amid resource constraints. However, the incomplete documentation and lack of formal releases suggest it may not be fully polished for immediate use. _themes: inference · scaling · llms · search_ #### [mckaywrigley/chatbot-ui](https://github.com/mckaywrigley/chatbot-ui) *TypeScript · ★33,183 · MIT · beta · score:0.80 · hot:0.52 · rising:0.61 · durable:0.73 · board:durable · trend:stable* AI chat for any model. **Why it matters.** This repository offers an open-source UI for interacting with various AI models, enabling users to build and deploy chatbots without proprietary dependencies. It matters now because the AI chat market is expanding rapidly, and this project provides a customizable, community-driven alternative that addresses deployment challenges, though its lack of a tagged release may hinder reliability for production use. _themes: chat-ui · ai-models · inference · deployment_ #### [google-research/tuning_playbook](https://github.com/google-research/tuning_playbook) *? · ★30,035 · NOASSERTION · beta · score:0.85 · hot:0.52 · rising:0.59 · durable:0.72 · board:durable · trend:stable* A playbook for systematically maximizing the performance of deep learning models. **Why it matters.** This repository provides a comprehensive playbook for systematically tuning deep learning models, covering topics like architecture selection, optimizer choices, and experimental design strategies. It matters right now because optimizing models is critical amid growing AI complexity and resource constraints, helping practitioners achieve better performance efficiently in an era of rapid model development and deployment. _themes: tuning · hyperparameter-optimization · deep-learning · optimization_ #### [deepseek-ai/awesome-deepseek-integration](https://github.com/deepseek-ai/awesome-deepseek-integration) *? · ★36,313 · CC0-1.0 · beta · score:0.70 · hot:0.51 · rising:0.61 · durable:0.73 · board:durable · trend:stable* Integrate the DeepSeek API into popular software **Why it matters.** This repository curates a list of integrations for the DeepSeek API into various popular software and frameworks, including AI agents, RAG systems, browser extensions, and more, providing developers with ready-to-use examples and guides. It matters right now because the proliferation of AI APIs demands seamless integration tools, and with DeepSeek's rapid adoption, this resource helps accelerate development of AI-enhanced applications amid growing competition in the AI space. _themes: api-integration · ai-agents · rag · frameworks_ #### [allenai/codescientist](https://github.com/allenai/codescientist) *Python · ★325 · Apache-2.0 · experimental · score:0.60 · hot:0.51 · rising:0.54 · durable:0.65 · board:durable · trend:stable* CodeScientist: An automated scientific discovery system for code-based experiments **Why it matters.** CodeScientist automates the generation, implementation, and analysis of scientific experiments by using LLMs to mutate existing code and articles, then runs them in containers and produces reports, with options for human oversight. It matters now amid growing AI automation in research, as it could potentially speed up experiment design, but its reliance on LLMs for critical mutations raises concerns about accuracy and reproducibility, especially given its early stage and moderate adoption. _themes: llm · automation · agents · research_ #### [huggingface/llm.nvim](https://github.com/huggingface/llm.nvim) *Lua · ★1,154 · Apache-2.0 · beta · score:0.70 · hot:0.51 · rising:0.52 · durable:0.56 · board:durable · trend:stable* LLM powered development for Neovim **Why it matters.** llm.nvim is a Neovim plugin that provides AI-powered code completion using Large Language Models from Hugging Face, allowing developers to integrate LLM features into their workflow. It matters now amid the growing adoption of AI coding assistants, as it offers an alternative for Neovim users who want Hugging Face's capabilities without switching editors, but its reliance on external APIs and lack of official releases may lead to instability and rate-limiting issues. _themes: llm · inference · code-completion · neovim_ #### [NVIDIA/DCGM](https://github.com/NVIDIA/DCGM) *C++ · ★708 · Apache-2.0 · production · score:0.75 · hot:0.50 · rising:0.53 · durable:0.58 · board:durable · trend:stable* NVIDIA Data Center GPU Manager (DCGM) is a project for gathering telemetry and measuring the health of NVIDIA GPUs **Why it matters.** NVIDIA DCGM is a suite of tools for monitoring and managing NVIDIA GPUs in data centers, providing telemetry, health checks, diagnostics, and integration with systems like Kubernetes. It matters now because the growing scale of AI and machine learning workloads demands reliable GPU oversight to prevent downtime and optimize resource use, especially in enterprise environments where GPU failures can be costly. _themes: monitoring · gpu · infrastructure · telemetry_ #### [facebookresearch/nougat](https://github.com/facebookresearch/nougat) *Python · ★9,921 · MIT · beta · score:0.80 · hot:0.50 · rising:0.58 · durable:0.74 · board:durable · trend:stable* Implementation of Nougat Neural Optical Understanding for Academic Documents **Why it matters.** Nougat is a Python-based tool for parsing academic PDFs, using neural networks to accurately extract and interpret LaTeX math expressions and tables. It matters now as the growing need for automated document processing in research could improve efficiency in handling scientific literature, though its early release version suggests potential limitations in reliability and feature completeness. _themes: ocr · pdf-parsing · neural-networks · academic-documents_ #### [huggingface/hf-agents](https://github.com/huggingface/hf-agents) *Shell · ★389 · Apache-2.0 · experimental · score:0.70 · hot:0.50 · rising:0.51 · durable:0.59 · board:durable · trend:stable* HF CLI extension to run local coding agent powered by llmfit and llama.cpp **Why it matters.** This repo is a Hugging Face CLI extension that automates hardware detection using llmfit, recommends compatible models, and launches a local coding agent via llama.cpp and Pi, all in one command. It matters now because it lowers the barrier for developers to run AI agents locally, addressing privacy and resource constraints in an era of rising on-device AI experimentation. However, its reliance on external tools and lack of a formal release may limit immediate adoption. _themes: agents · inference · local · hardware_ #### [openai/openai-openapi](https://github.com/openai/openai-openapi) *? · ★2,363 · MIT · production · score:0.80 · hot:0.50 · rising:0.57 · durable:0.71 · board:durable · trend:stable* OpenAPI specification for the OpenAI API **Why it matters.** This repository provides the official OpenAPI specification for the OpenAI API, allowing developers to generate documentation, clients, and tools for integration. It matters because it standardizes access to OpenAI's evolving services, helping developers maintain compatibility amid rapid AI advancements, though its manual updates may introduce delays in reflecting the latest changes. _themes: api-spec · openai · integration · documentation_ #### [microsoft/cascadia-code](https://github.com/microsoft/cascadia-code) *Python · ★27,641 · NOASSERTION · production · score:0.75 · hot:0.48 · rising:0.55 · durable:0.75 · board:durable · trend:stable* This is a fun, new monospaced font that includes programming ligatures and is designed to enhance the modern look and feel of the Windows Terminal. **Why it matters.** Cascadia Code is a monospaced font optimized for coding, featuring ligatures that improve readability of symbols in editors and terminals. It matters now as it's the default font in Windows Terminal and Visual Studio, enhancing developer productivity amid the growing use of modern IDEs and remote development tools, though its lack of a clear license might limit adoption. _themes: fonts · coding · ligatures · ui-enhancement_ #### [google-deepmind/alphatensor](https://github.com/google-deepmind/alphatensor) *Python · ★2,828 · Apache-2.0 · experimental · score:0.85 · hot:0.47 · rising:0.53 · durable:0.65 · board:durable · trend:stable* **Why it matters.** AlphaTensor uses reinforcement learning to discover faster matrix multiplication algorithms, which are fundamental to machine learning and high-performance computing, potentially improving efficiency in AI workloads. This repo provides code for exploring, benchmarking, and verifying these algorithms, making it relevant for ongoing efforts to optimize computational resources amid growing demands for AI training. It builds on a 2022 Nature publication, offering practical tools that could influence future hardware and software designs. _themes: reinforcement-learning · optimization · algorithms · matrix-multiplication_ #### [huggingface/llm_training_handbook](https://github.com/huggingface/llm_training_handbook) *Python · ★558 · CC-BY-SA-4.0 · beta · score:0.75 · hot:0.47 · rising:0.50 · durable:0.65 · board:durable · trend:down* An open collection of methodologies to help with successful training of large language models. **Why it matters.** This repository provides an open collection of practical methodologies, scripts, and copy-paste commands for training large language models, focusing on technical aspects like parallelism, throughput optimization, and debugging for engineers. It matters now because the rapid growth of LLMs requires accessible resources to handle scaling and performance challenges in real-world training scenarios, especially as organizations push for efficient AI development. However, its incomplete status and lack of formal releases mean it's still evolving and not a comprehensive solution yet. _themes: llm-training · performance · scalability · debugging_ #### [anthropics/life-sciences](https://github.com/anthropics/life-sciences) *Python · ★314 · no-license · beta · score:0.65 · hot:0.46 · rising:0.51 · durable:0.62 · board:durable · trend:down* Repo for the Claude Code Marketplace to use with the Claude for Life Sciences Launch. This will continue to host the marketplace.json long-term, but not the actual MCP servers. **Why it matters.** This repository hosts plugins for Anthropic's Claude AI to integrate life sciences tools, such as PubMed and BioRender, allowing users to access external databases and perform specialized analyses directly within the AI interface. It matters now because AI is rapidly being adopted in research to enhance productivity, but this setup is limited to Claude users and may face issues like dependency on proprietary systems and lack of a specified license, potentially hindering broader adoption. _themes: ai-plugins · life-sciences · research · integration_ #### [anthropics/claude-code-security-review](https://github.com/anthropics/claude-code-security-review) *Python · ★4,315 · MIT · beta · score:0.70 · hot:0.46 · rising:0.52 · durable:0.65 · board:durable · trend:stable* An AI-powered security review GitHub Action using Claude to analyze code changes for security vulnerabilities. **Why it matters.** This repository offers a GitHub Action that leverages Anthropic's Claude AI to scan code changes in pull requests for security vulnerabilities, providing automated comments and analysis based on semantic understanding. It matters now because the increasing adoption of AI in software development highlights the need for efficient, context-aware security tools to mitigate risks in rapidly evolving codebases, though its reliance on a proprietary API and lack of hardening may limit widespread adoption. _themes: ai · security · code-review · vulnerability-detection_ #### [xai-org/grok-prompts](https://github.com/xai-org/grok-prompts) *Jinja · ★4,051 · AGPL-3.0 · production · score:0.60 · hot:0.46 · rising:0.53 · durable:0.64 · board:durable · trend:stable* Prompts for our Grok chat assistant and the `@grok` bot on X. **Why it matters.** This repository contains system prompts for xAI's Grok AI models, used to configure chat assistants on grok.com and X, providing templates in Jinja format for various features and model versions. It matters for developers and researchers interested in prompt engineering, as it offers transparency into how xAI structures AI behavior, though its value is limited without access to the underlying models and may not introduce novel techniques beyond standard practices. However, with growing emphasis on AI ethics and open-sourcing, it serves as a practical reference for adapting similar prompts. _themes: prompts · llm · ai-assistants · system-prompts_ #### [QwenLM/Qwen3-ASR-Toolkit](https://github.com/QwenLM/Qwen3-ASR-Toolkit) *Python · ★936 · MIT · production · score:0.70 · hot:0.45 · rising:0.53 · durable:0.70 · board:durable · trend:down* Official Python toolkit for the Qwen3-ASR API. Parallel high‑throughput calls, robust long‑audio transcription, multi‑sample‑rate support. **Why it matters.** This repository provides a Python toolkit for the Qwen3-ASR API, focusing on handling long audio files by splitting them using voice activity detection and processing chunks in parallel, which addresses common limitations in speech recognition workflows. It matters now as ASR demands grow in applications like transcription services, and this tool offers practical enhancements for developers working with multilingual audio, though its reliance on a specific API may limit broader adoption compared to fully open-source alternatives. _themes: asr · speech-recognition · parallel-processing · audio_ #### [google-deepmind/dramatron](https://github.com/google-deepmind/dramatron) *Jupyter Notebook · ★1,075 · Apache-2.0 · experimental · score:0.70 · hot:0.45 · rising:0.51 · durable:0.62 · board:durable · trend:down* Dramatron uses large language models to generate coherent scripts and screenplays. **Why it matters.** Dramatron is a Jupyter Notebook-based tool from Google DeepMind that uses large language models to generate scripts and screenplays through hierarchical story building, requiring users to integrate their own LLM for functionality. It matters now as AI-assisted creative writing gains interest amid advancements in LLMs, but its experimental nature and need for custom implementation limit its immediate practicality for widespread adoption. _themes: llm · text-generation · creative-ai · story-generation_ #### [microsoft/AI](https://github.com/microsoft/AI) *Python · ★2,495 · MIT · beta · score:0.40 · hot:0.45 · rising:0.52 · durable:0.60 · board:durable · trend:stable* Microsoft AI **Why it matters.** This repository appears to be a basic introductory overview of artificial intelligence, covering definitions, applications, and historical context, but it lacks substantial code, tools, or practical implementations based on the provided excerpt. It matters primarily for educational purposes in a field where foundational knowledge is essential, though its generic content offers little new value in the current era of advanced AI developments like large language models and practical frameworks. _themes: ai · introduction · history · machine-learning_ #### [microsoft/AI-Red-Teaming-Playground-Labs](https://github.com/microsoft/AI-Red-Teaming-Playground-Labs) *TypeScript · ★1,910 · MIT · beta · score:0.75 · hot:0.45 · rising:0.51 · durable:0.64 · board:durable · trend:stable* AI Red Teaming playground labs to run AI Red Teaming trainings including infrastructure. **Why it matters.** This repository provides labs and challenges for AI red teaming, including infrastructure based on Chat Copilot, to help users practice identifying vulnerabilities like prompt injection in AI systems. It matters now due to the growing risks in AI deployment and the need for hands-on security training, as evidenced by its use in events like Black Hat and integration with tools like PyRIT for automated testing. _themes: red-teaming · ai-security · prompt-injection · ai-testing_ #### [deepseek-ai/awesome-deepseek-coder](https://github.com/deepseek-ai/awesome-deepseek-coder) *? · ★775 · no-license · experimental · score:0.60 · hot:0.45 · rising:0.49 · durable:0.58 · board:durable · trend:stable* A curated list of open-source projects related to DeepSeek Coder **Why it matters.** This repository is a curated list of open-source projects related to DeepSeek Coder, a series of AI models designed for coding tasks, including base and instruct variants available on Hugging Face. It matters right now as it aggregates resources for emerging code-focused LLMs, potentially aiding developers and researchers exploring alternatives to established models like CodeLlama, though its value is limited to being a simple directory without original contributions or updates. _themes: llm · coding · inference · models_ #### [huggingface/awesome-huggingface](https://github.com/huggingface/awesome-huggingface) *? · ★1,069 · Apache-2.0 · production · score:0.70 · hot:0.45 · rising:0.51 · durable:0.64 · board:durable · trend:down* 🤗 A list of wonderful open-source projects & applications integrated with Hugging Face libraries. **Why it matters.** This repository curates a list of open-source projects and applications integrated with Hugging Face libraries, serving as a centralized resource for discovering tools in machine learning, NLP, and transformers. It matters now because the Hugging Face ecosystem is growing rapidly amid the AI boom, helping developers and researchers quickly find and adopt relevant projects to accelerate their work in an increasingly competitive field. _themes: nlp · transformers · machine-learning · curation_ #### [NVIDIA/nvidia-settings](https://github.com/NVIDIA/nvidia-settings) *C · ★334 · GPL-2.0 · production · score:0.40 · hot:0.45 · rising:0.48 · durable:0.49 · board:durable · trend:stable* NVIDIA driver control panel **Why it matters.** NVIDIA/nvidia-settings is an open-source tool that provides a command-line and graphical interface for configuring NVIDIA graphics driver settings on Linux, such as overclocking, fan control, and display options. It matters for users managing NVIDIA hardware in production environments where fine-tuned GPU performance is needed, but its relevance is diminished by the lack of recent updates, potentially leaving it behind more modern alternatives in the fast-paced AI and computing sectors. _themes: gpu · configuration · monitoring · linux_ #### [EleutherAI/cookbook](https://github.com/EleutherAI/cookbook) *Python · ★837 · Apache-2.0 · beta · score:0.65 · hot:0.44 · rising:0.48 · durable:0.59 · board:durable · trend:down* Deep learning for dummies. All the practical details and useful utilities that go into working with real models. **Why it matters.** This repository compiles practical utilities, benchmarks, and reading lists for beginners working with deep learning models, particularly transformers and LLMs, focusing on real-world calculations and educational resources. It matters now because accessible guides help demystify AI for newcomers amid rapid advancements, though its value is tempered by the lack of formal releases and modest community engagement with only 837 stars. _themes: deep-learning · transformers · benchmarks · education_ #### [google-deepmind/funsearch](https://github.com/google-deepmind/funsearch) *Jupyter Notebook · ★1,043 · Apache-2.0 · experimental · score:0.60 · hot:0.44 · rising:0.49 · durable:0.59 · board:durable · trend:down* **Why it matters.** FunSearch is a system from Google DeepMind that uses large language models to generate and evolve programs for solving mathematical problems, leading to discoveries in areas like cap sets and bin packing as published in a 2023 Nature paper. It matters because it showcases AI's potential in automated mathematical discovery, which could accelerate research, but the repo's partial implementation and lack of updates limit its practical utility for broader adoption. The absence of a latest release and full infrastructure suggests it's more of a proof-of-concept than a ready-to-use tool. _themes: llm · program-search · mathematical-discovery · evolutionary-algorithm_ #### [anthropics/claudes-c-compiler](https://github.com/anthropics/claudes-c-compiler) *Rust · ★2,634 · CC0-1.0 · experimental · score:0.60 · hot:0.44 · rising:0.48 · durable:0.58 · board:durable · trend:stable* Claude Opus 4.6 wrote a dependency-free C compiler in Rust, with backends targeting x86 (64- and 32-bit), ARM, and RISC-V, capable of compiling a booting Linux kernel. **Why it matters.** This repository features a C compiler written entirely by an AI (Claude Opus 4.6) in Rust, targeting architectures like x86-64, ARM, and RISC-V, and capable of compiling a booting Linux kernel without external dependencies. It serves as a demonstration of AI's potential in generating complex software, but its unvalidated code and lack of reliability make it unsuitable for practical use, highlighting current limitations in AI-assisted development. This is relevant now as it contributes to ongoing debates about AI's role in coding, especially with rapid advancements in large language models. _themes: ai-generated · compiler · rust · codegen_ #### [meta-llama/synthetic-data-kit](https://github.com/meta-llama/synthetic-data-kit) *Python · ★1,573 · MIT · experimental · score:0.70 · hot:0.44 · rising:0.47 · durable:0.57 · board:durable · trend:down* Tool for generating high quality Synthetic datasets **Why it matters.** Synthetic Data Kit is a CLI tool from Meta that generates and curates synthetic datasets for fine-tuning LLMs, using simple commands to handle data ingestion, creation, and formatting. It addresses the common challenge of preparing unstructured data for LLM training, which is relevant amid growing needs for custom models, but its lack of a formal release and reliance on external LLMs may limit reliability and adoption. While it could streamline workflows for developers, the tool's effectiveness hinges on the quality of generated data, which varies based on the base LLM used. _themes: fine-tuning · synthetic-data · llm · data-generation_ #### [huggingface/education-toolkit](https://github.com/huggingface/education-toolkit) *Jupyter Notebook · ★386 · Apache-2.0 · beta · score:0.70 · hot:0.43 · rising:0.48 · durable:0.57 · board:durable · trend:down* Educational materials for universities **Why it matters.** This repository provides a collection of tutorials, resources, and educational materials focused on Hugging Face's ecosystem, including guides for exploring models, building ML demos with Gradio, and hosting them on Hugging Face Spaces, aimed at simplifying ML education for workshops and classes. It matters now because the growing demand for accessible ML training materials helps bridge the skills gap in a rapidly evolving field, though its lack of formal releases and modest star count (386) suggests potential maintenance issues that could affect long-term reliability. _themes: tutorials · education · huggingface · ml-demos_ #### [google-deepmind/barkour_robot](https://github.com/google-deepmind/barkour_robot) *C++ · ★340 · NOASSERTION · experimental · score:0.70 · hot:0.43 · rising:0.46 · durable:0.57 · board:durable · trend:down* Barkour Robot: Agile Quadruped Robots by Google DeepMind **Why it matters.** This repository provides design assets, assembly instructions, and core software like firmware for Google DeepMind's Barkour quadruped robots, enabling researchers to build and experiment with agile robotic systems. It matters now because advancements in AI-driven robotics are accelerating, with potential applications in autonomous navigation and physical AI interactions, and DeepMind's resources could foster innovation in a field that's gaining momentum due to real-world demands like warehouse automation and disaster response. _themes: robotics · quadruped · firmware · hardware_ #### [openai/automated-interpretability](https://github.com/openai/automated-interpretability) *Python · ★1,074 · no-license · experimental · score:0.70 · hot:0.42 · rising:0.46 · durable:0.56 · board:durable · trend:down* **Why it matters.** This repository provides code and tools for automatically explaining neuron behavior in language models like GPT-2 XL, based on a research paper, including datasets for neuron activations and explanations. It matters right now because AI interpretability is critical for addressing model transparency, safety, and ethical concerns amid growing adoption of large language models, though its lack of a license and official releases limits broader accessibility. _themes: interpretability · neurons · llms · datasets_ #### [openai/circuit_sparsity](https://github.com/openai/circuit_sparsity) *Python · ★515 · Apache-2.0 · experimental · score:0.65 · hot:0.42 · rising:0.46 · durable:0.59 · board:durable · trend:down* Open-source release accompanying Gao et al. 2025 **Why it matters.** This repo provides tools for inspecting and visualizing sparse circuits in neural networks, including a Streamlit dashboard for interactive exploration and code for running inference on lightweight GPT models, based on the Gao et al. 2025 paper. It matters right now because AI efficiency and interpretability are critical amid growing model sizes and computational costs, offering practical ways to prune and understand transformers without hype-driven overclaims. _themes: sparsity · visualization · inference · neural-networks_ #### [allenai/tango](https://github.com/allenai/tango) *Python · ★572 · Apache-2.0 · beta · score:0.70 · hot:0.42 · rising:0.47 · durable:0.62 · board:durable · trend:down* Organize your experiments into discrete steps that can be cached and reused throughout the lifetime of your research project. **Why it matters.** Tango is a Python library that structures machine learning experiments into cachable steps, helping to manage and reuse components in research projects, which can reduce redundancy and improve reproducibility. It matters now as AI research involves increasingly complex workflows, and tools like this address the need for efficient experiment organization amid growing computational demands. However, its adoption is niche and may not fully integrate with all popular ML frameworks. _themes: experiment-management · caching · ml-workflows · pytorch_ #### [microsoft/live-share](https://github.com/microsoft/live-share) *? · ★2,374 · CC-BY-4.0 · production · score:0.60 · hot:0.42 · rising:0.48 · durable:0.57 · board:durable · trend:stable* Real-time collaborative development from the comfort of your favorite tools **Why it matters.** Microsoft Live Share is a tool that enables real-time collaborative coding sessions directly within Visual Studio and VS Code, allowing developers to share screens, edit code together, and debug remotely. It matters now because remote work continues to dominate, making seamless pair-programming essential for distributed teams, though its integration is limited to Microsoft's ecosystem and may not be as feature-rich as newer alternatives. _themes: collaboration · pair-programming · ide · real-time_ #### [allenai/awesome-open-source-lms](https://github.com/allenai/awesome-open-source-lms) *? · ★363 · no-license · beta · score:0.70 · hot:0.42 · rising:0.46 · durable:0.59 · board:durable · trend:down* Friends of OLMo and their links. **Why it matters.** This repository curates a list of open-source language models and related resources, originating from a 2024 NeurIPS tutorial by Allen AI, focusing on the full pipeline of language model development. It matters now because it addresses the growing need for transparency in AI amid the dominance of proprietary models, providing researchers with accessible tools and insights to study and build open alternatives, which is crucial for advancing ethical AI research and mitigating biases. _themes: open-source · language-models · tutorials · AI-pipeline_ #### [QwenLM/AutoIF](https://github.com/QwenLM/AutoIF) *Python · ★330 · Apache-2.0 · experimental · score:0.70 · hot:0.41 · rising:0.45 · durable:0.58 · board:durable · trend:down* **Why it matters.** AutoIF is a tool for automatically generating and verifying instruction-following data for large language models through self-play and code execution feedback, addressing the challenges of manual data annotation in LLM training. It matters now because the rapid growth of AI applications demands efficient, scalable methods to improve model reliability, especially as fine-tuning datasets become a bottleneck in research and development. However, its experimental nature and reliance on specific dependencies limit immediate widespread adoption. _themes: llm · fine-tuning · data-generation · verification_ #### [allenai/python-package-template](https://github.com/allenai/python-package-template) *Python · ★539 · Apache-2.0 · production · score:0.70 · hot:0.41 · rising:0.47 · durable:0.60 · board:durable · trend:down* A template repo for Python packages **Why it matters.** This repository provides a template for setting up Python packages with built-in CI/CD, testing, documentation, and best practices, helping developers avoid boilerplate code. It matters now as the Python ecosystem grows, enabling faster project starts and maintaining high standards, especially for open-source contributors looking to streamline their workflows. _themes: python · packaging · ci-cd · templates_ #### [huggingface/gpt-oss-recipes](https://github.com/huggingface/gpt-oss-recipes) *Jupyter Notebook · ★503 · Apache-2.0 · experimental · score:0.60 · hot:0.41 · rising:0.45 · durable:0.56 · board:durable · trend:down* Collection of scripts and notebooks for OpenAI's latest GPT OSS models **Why it matters.** This repository provides a collection of scripts and notebooks for optimizing and fine-tuning OpenAI's GPT-OSS models, specifically the 20B and 120B parameter versions, focusing on techniques like tensor parallelism and flash attention. It matters now because the release of these open-source models democratizes access to large-scale AI, but their size demands efficient handling, making these recipes valuable for advanced users; however, the lack of formal releases and limited scope may introduce instability or obsolescence. _themes: fine-tuning · inference · parallelism · optimization_ #### [microsoft/DbgShell](https://github.com/microsoft/DbgShell) *C# · ★697 · MIT · experimental · score:0.60 · hot:0.40 · rising:0.44 · durable:0.50 · board:durable · trend:down* A PowerShell front-end for the Windows debugger engine. **Why it matters.** DbgShell provides a PowerShell interface to the Windows debugger engine, aiming to simplify automation of debugging tasks that are otherwise cumbersome with tools like WinDbg. However, it's an experimental project with no official support, frequent breaking changes, and is not suitable for production use, making it relevant only for enthusiasts or developers exploring Windows debugging automation but not a reliable solution right now. _themes: debugging · automation · powershell · windows_ #### [allenai/pawls](https://github.com/allenai/pawls) *Python · ★428 · Apache-2.0 · experimental · score:0.60 · hot:0.38 · rising:0.42 · durable:0.50 · board:durable · trend:down* Software that makes labeling PDFs easy. **Why it matters.** PAWLS is a Python tool for annotating PDFs by extracting structures and enabling label collection via a UI, primarily designed for academic papers in the Semantic Scholar corpus. It matters now as the demand for annotated datasets in AI-driven document analysis grows, though its lack of official releases and niche focus limit its immediate broad applicability. _themes: pdf-annotation · document-labeling · data-annotation · extraction_ #### [allenai/writing-code-for-nlp-research-emnlp2018](https://github.com/allenai/writing-code-for-nlp-research-emnlp2018) *Python · ★554 · Apache-2.0 · archived · score:0.40 · hot:0.35 · rising:0.39 · durable:0.50 · board:durable · trend:down* A companion repository for the "Writing code for NLP Research" Tutorial at EMNLP 2018 **Why it matters.** This repository serves as a companion to a 2018 EMNLP tutorial, offering code examples and best practices for writing effective Python code in NLP research, likely including topics like data processing and model implementation. It matters today primarily as a historical resource for understanding foundational coding approaches in NLP, though its relevance is limited due to advancements in modern frameworks since 2018, making it more suitable for educational reference than practical application. _themes: nlp · coding · tutorial · research_ #### [EleutherAI/vqgan-clip](https://github.com/EleutherAI/vqgan-clip) *Jupyter Notebook · ★354 · MIT · experimental · score:0.60 · hot:0.35 · rising:0.40 · durable:0.54 · board:durable · trend:down* **Why it matters.** This repo provides an implementation of VQGAN-CLIP for text-to-image generation, allowing users to create and edit images based on textual prompts using pre-trained models. It matters as an early open-source experiment in generative AI, enabling accessible exploration of semantic image synthesis, but its lack of updates and formal releases makes it less relevant compared to more advanced, actively maintained alternatives. _themes: image-generation · text-to-image · vqgan · clip_ #### [allenai/scholarphi](https://github.com/allenai/scholarphi) *Python · ★429 · Apache-2.0 · beta · score:0.65 · hot:0.34 · rising:0.40 · durable:0.52 · board:durable · trend:down* An interactive PDF reader. **Why it matters.** Scholarphi is an augmented PDF reader that extracts entities and their bounding boxes from academic papers, providing an interactive interface for better navigation and analysis. It matters now as the volume of scientific literature grows, offering researchers a practical tool for efficient document interaction in fields like AI, though its lack of formal releases and modest adoption (429 stars) suggests it's not yet widely polished or essential. _themes: nlp · pdf-processing · entity-extraction · interactive-ui_ ### uncategorized (1536) #### [labring/FastGPT](https://github.com/labring/FastGPT) *TypeScript · ★27,770 · NOASSERTION · — · score:0.00 · hot:0.80 · rising:0.78 · durable:0.62 · board:hot · trend:up* FastGPT is a knowledge-based platform built on the LLMs, offers a comprehensive suite of out-of-the-box capabilities such as data processing, RAG retrieval, and visual AI workflow orchestration, letting you easily develop and deploy complex question-answering systems without the need for extensive setup or configuration. #### [sickn33/antigravity-awesome-skills](https://github.com/sickn33/antigravity-awesome-skills) *Python · ★33,957 · MIT · — · score:0.00 · hot:0.79 · rising:0.82 · durable:0.67 · board:rising · trend:up* Installable GitHub library of 1,400+ agentic skills for Claude Code, Cursor, Codex CLI, Gemini CLI, Antigravity, and more. Includes installer CLI, bundles, workflows, and official/community skill collections. #### [simstudioai/sim](https://github.com/simstudioai/sim) *TypeScript · ★27,828 · Apache-2.0 · — · score:0.00 · hot:0.79 · rising:0.78 · durable:0.65 · board:hot · trend:up* Build, deploy, and orchestrate AI agents. Sim is the central intelligence layer for your AI workforce. #### [ZhuLinsen/daily_stock_analysis](https://github.com/ZhuLinsen/daily_stock_analysis) *Python · ★30,506 · MIT · — · score:0.00 · hot:0.79 · rising:0.80 · durable:0.68 · board:rising · trend:up* LLM驱动的 A/H/美股智能分析器：多数据源行情 + 实时新闻 + LLM决策仪表盘 + 多渠道推送，零成本定时运行，纯白嫖. LLM-powered stock analysis system for A/H/US markets. #### [code-yeongyu/oh-my-openagent](https://github.com/code-yeongyu/oh-my-openagent) *TypeScript · ★52,763 · NOASSERTION · — · score:0.00 · hot:0.79 · rising:0.78 · durable:0.63 · board:hot · trend:up* omo; the best agent harness - previously oh-my-opencode #### [Tencent/WeKnora](https://github.com/Tencent/WeKnora) *Go · ★13,933 · NOASSERTION · — · score:0.00 · hot:0.79 · rising:0.75 · durable:0.61 · board:hot · trend:up* LLM-powered framework for deep document understanding, semantic retrieval, and context-aware answers using RAG paradigm. #### [NirDiamant/RAG_Techniques](https://github.com/NirDiamant/RAG_Techniques) *Jupyter Notebook · ★26,868 · NOASSERTION · — · score:0.00 · hot:0.78 · rising:0.76 · durable:0.64 · board:hot · trend:up* This repository showcases various advanced techniques for Retrieval-Augmented Generation (RAG) systems. Each technique has a detailed notebook tutorial. #### [1Panel-dev/MaxKB](https://github.com/1Panel-dev/MaxKB) *Python · ★20,762 · GPL-3.0 · — · score:0.00 · hot:0.78 · rising:0.77 · durable:0.65 · board:hot · trend:up* 🔥 MaxKB is an open-source platform for building enterprise-grade agents. 强大易用的开源企业级智能体平台。 #### [onyx-dot-app/onyx](https://github.com/onyx-dot-app/onyx) *Python · ★27,640 · NOASSERTION · — · score:0.00 · hot:0.78 · rising:0.77 · durable:0.60 · board:hot · trend:up* Open Source AI Platform - AI Chat with advanced features that works with every LLM #### [emcie-co/parlant](https://github.com/emcie-co/parlant) *Python · ★17,981 · Apache-2.0 · — · score:0.00 · hot:0.77 · rising:0.77 · durable:0.64 · board:hot · trend:up* The interaction control harness for customer-facing AI agents - optimized for building controlled, consistent, and predictable customer interactions with LLMs. #### [xtekky/gpt4free](https://github.com/xtekky/gpt4free) *Python · ★66,028 · GPL-3.0 · — · score:0.00 · hot:0.77 · rising:0.82 · durable:0.66 · board:rising · trend:up* The official gpt4free repository | various collection of powerful language models | opus 4.6 gpt 5.3 kimi 2.5 deepseek v3.2 gemini 3 #### [google-gemini/gemini-cli](https://github.com/google-gemini/gemini-cli) *TypeScript · ★101,773 · Apache-2.0 · — · score:0.00 · hot:0.77 · rising:0.79 · durable:0.62 · board:rising · trend:up* An open-source AI agent that brings the power of Gemini directly into your terminal. #### [topoteretes/cognee](https://github.com/topoteretes/cognee) *Python · ★16,441 · Apache-2.0 · — · score:0.00 · hot:0.77 · rising:0.77 · durable:0.63 · board:hot · trend:up* Knowledge Engine for AI Agent Memory in 6 lines of code #### [deepset-ai/haystack](https://github.com/deepset-ai/haystack) *MDX · ★24,902 · Apache-2.0 · — · score:0.00 · hot:0.77 · rising:0.75 · durable:0.64 · board:hot · trend:up* Open-source AI orchestration framework for building context-engineered, production-ready LLM applications. Design modular pipelines and agent workflows with explicit control over retrieval, routing, memory, and generation. Built for scalable agents, RAG, multimodal applications, semantic search, and conversational systems. #### [OpenBMB/UltraRAG](https://github.com/OpenBMB/UltraRAG) *Python · ★5,504 · Apache-2.0 · — · score:0.00 · hot:0.77 · rising:0.74 · durable:0.63 · board:hot · trend:up* [GitHub Trending #2] A Low-Code MCP Framework for Building Complex and Innovative RAG Pipelines #### [screenpipe/screenpipe](https://github.com/screenpipe/screenpipe) *Rust · ★18,260 · NOASSERTION · — · score:0.00 · hot:0.77 · rising:0.76 · durable:0.61 · board:hot · trend:up* Run agents that work for you based on what you do. AI finally knows what you are doing #### [router-for-me/CLIProxyAPI](https://github.com/router-for-me/CLIProxyAPI) *Go · ★27,250 · MIT · — · score:0.00 · hot:0.76 · rising:0.78 · durable:0.63 · board:rising · trend:up* Wrap Gemini CLI, Antigravity, ChatGPT Codex, Claude Code as an OpenAI/Gemini/Claude/Codex compatible API service, allowing you to enjoy the free Gemini 2.5 Pro, GPT 5, Claude model through API #### [safishamsi/graphify](https://github.com/safishamsi/graphify) *Python · ★30,456 · MIT · — · score:0.00 · hot:0.76 · rising:0.79 · durable:0.64 · board:rising · trend:up* AI coding assistant skill (Claude Code, Codex, OpenCode, Cursor, Gemini CLI, GitHub Copilot CLI, OpenClaw, Factory Droid, Trae, Google Antigravity). Turn any folder of code, docs, papers, images, or videos into a queryable knowledge graph #### [dyad-sh/dyad](https://github.com/dyad-sh/dyad) *TypeScript · ★20,145 · NOASSERTION · — · score:0.00 · hot:0.76 · rising:0.75 · durable:0.60 · board:hot · trend:up* Local, open-source AI app builder for power users ✨ v0 / Lovable / Replit / Bolt alternative 🌟 Star if you like it! #### [ComposioHQ/composio](https://github.com/ComposioHQ/composio) *TypeScript · ★27,825 · MIT · — · score:0.00 · hot:0.76 · rising:0.77 · durable:0.64 · board:rising · trend:up* Composio powers 1000+ toolkits, tool search, context management, authentication, and a sandboxed workbench to help you build AI agents that turn intent into action. #### [PrefectHQ/fastmcp](https://github.com/PrefectHQ/fastmcp) *Python · ★24,649 · Apache-2.0 · — · score:0.00 · hot:0.76 · rising:0.77 · durable:0.63 · board:rising · trend:up* 🚀 The fast, Pythonic way to build MCP servers and clients. #### [ray-project/ray](https://github.com/ray-project/ray) *Python · ★42,203 · Apache-2.0 · — · score:0.00 · hot:0.76 · rising:0.76 · durable:0.57 · board:hot · trend:up* Ray is an AI compute engine. Ray consists of a core distributed runtime and a set of AI Libraries for accelerating ML workloads. #### [aden-hive/hive](https://github.com/aden-hive/hive) *Python · ★10,096 · Apache-2.0 · — · score:0.00 · hot:0.76 · rising:0.77 · durable:0.58 · board:rising · trend:up* Multi-Agent Harness for Production AI #### [volcengine/OpenViking](https://github.com/volcengine/OpenViking) *Python · ★22,575 · AGPL-3.0 · — · score:0.00 · hot:0.76 · rising:0.77 · durable:0.63 · board:rising · trend:up* OpenViking is an open-source context database designed specifically for AI Agents(such as openclaw). OpenViking unifies the management of context (memory, resources, and skills) that Agents need through a file system paradigm, enabling hierarchical context delivery and self-evolving. #### [google/langextract](https://github.com/google/langextract) *Python · ★35,679 · Apache-2.0 · — · score:0.00 · hot:0.76 · rising:0.79 · durable:0.64 · board:rising · trend:up* A Python library for extracting structured information from unstructured text using LLMs with precise source grounding and interactive visualization. #### [Upsonic/Upsonic](https://github.com/Upsonic/Upsonic) *Python · ★7,831 · MIT · — · score:0.00 · hot:0.76 · rising:0.74 · durable:0.63 · board:hot · trend:up* Build autonomous AI agents in Python. #### [yamadashy/repomix](https://github.com/yamadashy/repomix) *TypeScript · ★23,676 · MIT · — · score:0.00 · hot:0.76 · rising:0.74 · durable:0.64 · board:hot · trend:up* 📦 Repomix is a powerful tool that packs your entire repository into a single, AI-friendly file. Perfect for when you need to feed your codebase to Large Language Models (LLMs) or other AI tools like Claude, ChatGPT, DeepSeek, Perplexity, Gemini, Gemma, Llama, Grok, and more. #### [AstrBotDevs/AstrBot](https://github.com/AstrBotDevs/AstrBot) *Python · ★30,264 · AGPL-3.0 · — · score:0.00 · hot:0.76 · rising:0.73 · durable:0.58 · board:hot · trend:up* AI Agent Assistant that integrates lots of IM platforms, LLMs, plugins and AI feature, and can be your openclaw alternative. ✨ #### [D4Vinci/Scrapling](https://github.com/D4Vinci/Scrapling) *Python · ★37,911 · BSD-3-Clause · — · score:0.00 · hot:0.76 · rising:0.79 · durable:0.64 · board:rising · trend:up* 🕷️ An adaptive Web Scraping framework that handles everything from a single request to a full-scale crawl! #### [olimorris/codecompanion.nvim](https://github.com/olimorris/codecompanion.nvim) *Lua · ★6,474 · Apache-2.0 · — · score:0.00 · hot:0.75 · rising:0.73 · durable:0.62 · board:hot · trend:up* ✨ AI Coding, Vim Style #### [mlflow/mlflow](https://github.com/mlflow/mlflow) *Python · ★25,445 · Apache-2.0 · — · score:0.00 · hot:0.75 · rising:0.75 · durable:0.56 · board:hot · trend:up* The open source AI engineering platform for agents, LLMs, and ML models. MLflow enables teams of all sizes to debug, evaluate, monitor, and optimize production-quality AI applications while controlling costs and managing access to models and data. #### [JuliusBrussee/caveman](https://github.com/JuliusBrussee/caveman) *Python · ★38,671 · MIT · — · score:0.00 · hot:0.75 · rising:0.77 · durable:0.64 · board:rising · trend:up* 🪨 why use many token when few token do trick — Claude Code skill that cuts 65% of tokens by talking like caveman #### [tw93/Pake](https://github.com/tw93/Pake) *Rust · ★47,913 · MIT · — · score:0.00 · hot:0.75 · rising:0.80 · durable:0.64 · board:rising · trend:up* 🤱🏻 Turn any webpage into a desktop app with one command. #### [vxcontrol/pentagi](https://github.com/vxcontrol/pentagi) *Go · ★15,200 · MIT · — · score:0.00 · hot:0.75 · rising:0.76 · durable:0.64 · board:rising · trend:up* Fully autonomous AI Agents system capable of performing complex penetration testing tasks #### [rtk-ai/rtk](https://github.com/rtk-ai/rtk) *Rust · ★29,832 · Apache-2.0 · — · score:0.00 · hot:0.75 · rising:0.75 · durable:0.60 · board:rising · trend:up* CLI proxy that reduces LLM token consumption by 60-90% on common dev commands. Single Rust binary, zero dependencies #### [langchain-ai/deepagents](https://github.com/langchain-ai/deepagents) *Python · ★21,263 · MIT · — · score:0.00 · hot:0.75 · rising:0.79 · durable:0.63 · board:rising · trend:up* Agent harness built with LangChain and LangGraph. Equipped with a planning tool, a filesystem backend, and the ability to spawn subagents - well-equipped to handle complex agentic tasks. #### [AI4Finance-Foundation/FinGPT](https://github.com/AI4Finance-Foundation/FinGPT) *Jupyter Notebook · ★19,616 · MIT · — · score:0.00 · hot:0.75 · rising:0.79 · durable:0.62 · board:rising · trend:up* FinGPT: Open-Source Financial Large Language Models! Revolutionize 🔥 We release the trained model on HuggingFace. #### [jamiepine/voicebox](https://github.com/jamiepine/voicebox) *TypeScript · ★20,964 · MIT · — · score:0.00 · hot:0.75 · rising:0.78 · durable:0.61 · board:rising · trend:up* The open-source voice synthesis studio #### [zeroclaw-labs/zeroclaw](https://github.com/zeroclaw-labs/zeroclaw) *Rust · ★30,337 · Apache-2.0 · — · score:0.00 · hot:0.75 · rising:0.78 · durable:0.62 · board:rising · trend:up* Fast, small, and fully autonomous AI personal assistant infrastructure, ANY OS, ANY PLATFORM — deploy anywhere, swap anything 🦀 #### [promptfoo/promptfoo](https://github.com/promptfoo/promptfoo) *TypeScript · ★20,288 · MIT · — · score:0.00 · hot:0.75 · rising:0.75 · durable:0.60 · board:rising · trend:up* Test your prompts, agents, and RAGs. Red teaming/pentesting/vulnerability scanning for AI. Compare performance of GPT, Claude, Gemini, Llama, and more. Simple declarative configs with command line and CI/CD integration. Used by OpenAI and Anthropic. #### [CoplayDev/unity-mcp](https://github.com/CoplayDev/unity-mcp) *C# · ★8,657 · MIT · — · score:0.00 · hot:0.75 · rising:0.73 · durable:0.63 · board:hot · trend:up* Unity MCP acts as a bridge, allowing AI assistants (like Claude, Cursor) to interact directly with your Unity Editor via a local MCP (Model Context Protocol) Client. Give your LLM tools to manage assets, control scenes, edit scripts, and automate tasks within Unity. #### [langchain4j/langchain4j](https://github.com/langchain4j/langchain4j) *Java · ★11,667 · Apache-2.0 · — · score:0.00 · hot:0.75 · rising:0.73 · durable:0.57 · board:hot · trend:up* LangChain4j is an idiomatic, open-source Java library for building LLM-powered applications on the JVM. It offers a unified API over popular LLM providers and vector stores, and makes implementing tool calling (including MCP support), agents and RAG easy. It integrates seamlessly with enterprise Java frameworks like Quarkus and Spring Boot. #### [xorbitsai/inference](https://github.com/xorbitsai/inference) *Python · ★9,243 · Apache-2.0 · — · score:0.00 · hot:0.74 · rising:0.74 · durable:0.61 · board:hot · trend:up* Swap GPT for any LLM by changing a single line of code. Xinference lets you run open-source, speech, and multimodal models on cloud, on-prem, or your laptop — all through one unified, production-ready inference API. #### [google/adk-python](https://github.com/google/adk-python) *Python · ★19,112 · Apache-2.0 · — · score:0.00 · hot:0.74 · rising:0.76 · durable:0.57 · board:rising · trend:up* An open-source, code-first Python toolkit for building, evaluating, and deploying sophisticated AI agents with flexibility and control. #### [lutzroeder/netron](https://github.com/lutzroeder/netron) *JavaScript · ★32,775 · MIT · — · score:0.00 · hot:0.74 · rising:0.77 · durable:0.60 · board:rising · trend:up* Visualizer for neural network, deep learning and machine learning models #### [github/github-mcp-server](https://github.com/github/github-mcp-server) *Go · ★29,074 · MIT · — · score:0.00 · hot:0.74 · rising:0.78 · durable:0.62 · board:rising · trend:up* GitHub's official MCP Server #### [holaboss-ai/holaOS](https://github.com/holaboss-ai/holaOS) *TypeScript · ★2,981 · MIT · — · score:0.00 · hot:0.74 · rising:0.74 · durable:0.60 · board:hot · trend:up* The agent environment for long-horizon work, continuity, and self-evolution. #### [ChromeDevTools/chrome-devtools-mcp](https://github.com/ChromeDevTools/chrome-devtools-mcp) *TypeScript · ★36,173 · Apache-2.0 · — · score:0.00 · hot:0.74 · rising:0.78 · durable:0.64 · board:rising · trend:up* Chrome DevTools for coding agents #### [paperless-ngx/paperless-ngx](https://github.com/paperless-ngx/paperless-ngx) *Python · ★38,712 · GPL-3.0 · — · score:0.00 · hot:0.74 · rising:0.78 · durable:0.62 · board:rising · trend:up* A community-supported supercharged document management system: scan, index and archive all your documents #### [academic/awesome-datascience](https://github.com/academic/awesome-datascience) *? · ★28,840 · MIT · — · score:0.00 · hot:0.74 · rising:0.80 · durable:0.60 · board:rising · trend:up* :memo: An awesome Data Science repository to learn and apply for real world problems. #### [QuantumNous/new-api](https://github.com/QuantumNous/new-api) *Go · ★27,648 · AGPL-3.0 · — · score:0.00 · hot:0.74 · rising:0.76 · durable:0.59 · board:rising · trend:up* A unified AI model hub for aggregation & distribution. It supports cross-converting various LLMs into OpenAI-compatible, Claude-compatible, or Gemini-compatible formats. A centralized gateway for personal and enterprise model management. 🍥 #### [Arize-ai/phoenix](https://github.com/Arize-ai/phoenix) *Python · ★9,347 · NOASSERTION · — · score:0.00 · hot:0.74 · rising:0.70 · durable:0.53 · board:hot · trend:up* AI Observability & Evaluation #### [alibaba/nacos](https://github.com/alibaba/nacos) *Java · ★32,867 · Apache-2.0 · — · score:0.00 · hot:0.74 · rising:0.77 · durable:0.61 · board:rising · trend:up* an easy-to-use dynamic service discovery, configuration and service management platform for building AI cloud native applications. #### [vercel/next.js](https://github.com/vercel/next.js) *JavaScript · ★139,003 · MIT · — · score:0.00 · hot:0.74 · rising:0.79 · durable:0.59 · board:rising · trend:up* The React Framework #### [strands-agents/sdk-python](https://github.com/strands-agents/sdk-python) *Python · ★5,668 · Apache-2.0 · — · score:0.00 · hot:0.74 · rising:0.71 · durable:0.56 · board:hot · trend:up* A model-driven approach to building AI agents in just a few lines of code. #### [vercel/turborepo](https://github.com/vercel/turborepo) *Rust · ★30,234 · MIT · — · score:0.00 · hot:0.73 · rising:0.78 · durable:0.63 · board:rising · trend:up* Build system optimized for JavaScript and TypeScript, written in Rust #### [janhq/jan](https://github.com/janhq/jan) *TypeScript · ★41,885 · NOASSERTION · — · score:0.00 · hot:0.73 · rising:0.76 · durable:0.60 · board:rising · trend:up* Jan is an open source alternative to ChatGPT that runs 100% offline on your computer. #### [agentscope-ai/agentscope](https://github.com/agentscope-ai/agentscope) *Python · ★24,035 · Apache-2.0 · — · score:0.00 · hot:0.73 · rising:0.75 · durable:0.62 · board:rising · trend:up* Build and run agents you can see, understand and trust. #### [getzep/graphiti](https://github.com/getzep/graphiti) *Python · ★25,117 · Apache-2.0 · — · score:0.00 · hot:0.73 · rising:0.75 · durable:0.61 · board:rising · trend:up* Build Real-Time Knowledge Graphs for AI Agents #### [meilisearch/meilisearch](https://github.com/meilisearch/meilisearch) *Rust · ★57,214 · NOASSERTION · — · score:0.00 · hot:0.73 · rising:0.74 · durable:0.58 · board:rising · trend:up* A lightning-fast search engine API bringing AI-powered hybrid search to your sites and applications. #### [TauricResearch/TradingAgents](https://github.com/TauricResearch/TradingAgents) *Python · ★51,632 · Apache-2.0 · — · score:0.00 · hot:0.73 · rising:0.77 · durable:0.65 · board:rising · trend:up* TradingAgents: Multi-Agents LLM Financial Trading Framework #### [neuron-core/neuron-ai](https://github.com/neuron-core/neuron-ai) *PHP · ★1,844 · MIT · — · score:0.00 · hot:0.73 · rising:0.71 · durable:0.59 · board:hot · trend:up* The PHP Agentic Framework to build production-ready AI driven applications. Connect components (LLMs, vector DBs, memory) to agents that can interact with your data. #### [neuml/txtai](https://github.com/neuml/txtai) *Python · ★12,405 · Apache-2.0 · — · score:0.00 · hot:0.73 · rising:0.71 · durable:0.63 · board:hot · trend:up* 💡 All-in-one AI framework for semantic search, LLM orchestration and language model workflows #### [genkit-ai/genkit](https://github.com/genkit-ai/genkit) *TypeScript · ★5,812 · Apache-2.0 · — · score:0.00 · hot:0.73 · rising:0.70 · durable:0.56 · board:hot · trend:up* Open-source framework for building AI-powered apps in JavaScript, Go, and Python, built and used in production by Google #### [Agenta-AI/agenta](https://github.com/Agenta-AI/agenta) *TypeScript · ★4,036 · NOASSERTION · — · score:0.00 · hot:0.73 · rising:0.70 · durable:0.53 · board:hot · trend:up* The open-source LLMOps platform: prompt playground, prompt management, LLM evaluation, and LLM observability all in one place. #### [dbeaver/dbeaver](https://github.com/dbeaver/dbeaver) *Java · ★49,652 · Apache-2.0 · — · score:0.00 · hot:0.73 · rising:0.75 · durable:0.54 · board:rising · trend:up* Free universal database tool and SQL client #### [LMCache/LMCache](https://github.com/LMCache/LMCache) *Python · ★8,019 · Apache-2.0 · — · score:0.00 · hot:0.73 · rising:0.71 · durable:0.56 · board:hot · trend:up* Supercharge Your LLM with the Fastest KV Cache Layer #### [Kilo-Org/kilocode](https://github.com/Kilo-Org/kilocode) *TypeScript · ★18,305 · MIT · — · score:0.00 · hot:0.73 · rising:0.75 · durable:0.56 · board:rising · trend:up* Kilo is the all-in-one agentic engineering platform. Build, ship, and iterate faster with the most popular open source coding agent. #1 coding agent on OpenRouter. 1.5M+ Kilo Coders. 25T+ tokens processed #### [yichuan-w/LEANN](https://github.com/yichuan-w/LEANN) *Python · ★10,837 · MIT · — · score:0.00 · hot:0.73 · rising:0.73 · durable:0.63 · board:rising · trend:up* [MLsys2026]: RAG on Everything with LEANN. Enjoy 97% storage savings while running a fast, accurate, and 100% private RAG application on your personal device. #### [google-ai-edge/mediapipe](https://github.com/google-ai-edge/mediapipe) *C++ · ★34,815 · Apache-2.0 · — · score:0.00 · hot:0.72 · rising:0.76 · durable:0.58 · board:rising · trend:up* Cross-platform, customizable ML solutions for live and streaming media. #### [PaddlePaddle/FastDeploy](https://github.com/PaddlePaddle/FastDeploy) *Python · ★3,676 · Apache-2.0 · — · score:0.00 · hot:0.72 · rising:0.71 · durable:0.53 · board:hot · trend:up* High-performance Inference and Deployment Toolkit for LLMs and VLMs based on PaddlePaddle #### [roboflow/supervision](https://github.com/roboflow/supervision) *Python · ★38,150 · MIT · — · score:0.00 · hot:0.72 · rising:0.77 · durable:0.63 · board:rising · trend:up* We write your reusable computer vision tools. 💜 #### [pydantic/pydantic-ai](https://github.com/pydantic/pydantic-ai) *Python · ★16,467 · MIT · — · score:0.00 · hot:0.72 · rising:0.74 · durable:0.58 · board:rising · trend:up* AI Agent Framework, the Pydantic way #### [ag2ai/ag2](https://github.com/ag2ai/ag2) *Python · ★4,421 · Apache-2.0 · — · score:0.00 · hot:0.72 · rising:0.72 · durable:0.54 · board:hot · trend:up* AG2 (formerly AutoGen): The Open-Source AgentOS.Join us at: https://discord.gg/sNGSwQME3x #### [2noise/ChatTTS](https://github.com/2noise/ChatTTS) *Python · ★39,107 · AGPL-3.0 · — · score:0.00 · hot:0.72 · rising:0.74 · durable:0.66 · board:rising · trend:up* A generative speech model for daily dialogue. #### [langchain-ai/langchainjs](https://github.com/langchain-ai/langchainjs) *TypeScript · ★17,522 · MIT · — · score:0.00 · hot:0.72 · rising:0.77 · durable:0.58 · board:rising · trend:up* The agent engineering platform #### [1186258278/OpenClawChineseTranslation](https://github.com/1186258278/OpenClawChineseTranslation) *JavaScript · ★3,716 · NOASSERTION · — · score:0.00 · hot:0.72 · rising:0.71 · durable:0.54 · board:hot · trend:up* 🦞 OpenClaw (Clawdbot/Moltbot) 汉化版 - 开源个人 AI 助手中文版 | Claude/ChatGPT LLM 接入 | WhatsApp/Telegram/Discord 多平台 | 每小时自动同步 | CLI + Dashboard 全中文 | 全流程搭建教程，以及排错指南！ #### [openvinotoolkit/openvino](https://github.com/openvinotoolkit/openvino) *C++ · ★10,113 · Apache-2.0 · — · score:0.00 · hot:0.72 · rising:0.72 · durable:0.54 · board:rising · trend:up* OpenVINO™ is an open source toolkit for optimizing and deploying AI inference #### [modelscope/ms-swift](https://github.com/modelscope/ms-swift) *Python · ★13,798 · Apache-2.0 · — · score:0.00 · hot:0.72 · rising:0.72 · durable:0.55 · board:rising · trend:up* Use PEFT or Full-parameter to CPT/SFT/DPO/GRPO 600+ LLMs (Qwen3.6, DeepSeek-R1, GLM-5, InternLM3, Llama4, ...) and 300+ MLLMs (Qwen3-VL, Qwen3-Omni, InternVL3.5, Ovis2.5, GLM4.5v, Llava, Phi4, ...) (AAAI 2025). #### [deepspeedai/DeepSpeed](https://github.com/deepspeedai/DeepSpeed) *Python · ★42,148 · Apache-2.0 · — · score:0.00 · hot:0.72 · rising:0.74 · durable:0.56 · board:rising · trend:up* DeepSpeed is a deep learning optimization library that makes distributed training and inference easy, efficient, and effective. #### [presenton/presenton](https://github.com/presenton/presenton) *TypeScript · ★4,735 · Apache-2.0 · — · score:0.00 · hot:0.72 · rising:0.73 · durable:0.55 · board:rising · trend:up* Open-Source AI Presentation Generator and API (Gamma, Beautiful AI, Decktopus Alternative) #### [tensorzero/tensorzero](https://github.com/tensorzero/tensorzero) *Rust · ★11,247 · Apache-2.0 · — · score:0.00 · hot:0.72 · rising:0.71 · durable:0.55 · board:hot · trend:up* TensorZero is an open-source LLMOps platform that unifies an LLM gateway, observability, evaluation, optimization, and experimentation. #### [Wei-Shaw/sub2api](https://github.com/Wei-Shaw/sub2api) *Go · ★13,675 · LGPL-3.0 · — · score:0.00 · hot:0.72 · rising:0.74 · durable:0.55 · board:rising · trend:up* Sub2API-CRS2 一站式开源中转服务，让 Claude、Openai 、Gemini、Antigravity订阅统一接入，支持拼车共享，更高效分摊成本，原生工具无缝使用。 #### [web-infra-dev/midscene](https://github.com/web-infra-dev/midscene) *TypeScript · ★12,716 · MIT · — · score:0.00 · hot:0.72 · rising:0.75 · durable:0.59 · board:rising · trend:up* AI-powered, vision-driven UI automation for every platform. #### [Blaizzy/mlx-audio](https://github.com/Blaizzy/mlx-audio) *Python · ★6,761 · MIT · — · score:0.00 · hot:0.72 · rising:0.73 · durable:0.58 · board:rising · trend:up* A text-to-speech (TTS), speech-to-text (STT) and speech-to-speech (STS) library built on Apple's MLX framework, providing efficient speech analysis on Apple Silicon. #### [siteboon/claudecodeui](https://github.com/siteboon/claudecodeui) *TypeScript · ★9,998 · AGPL-3.0 · — · score:0.00 · hot:0.72 · rising:0.74 · durable:0.58 · board:rising · trend:up* Use Claude Code, Cursor CLI or Codex on mobile and web with CloudCLI (aka Claude Code UI). CloudCLI is a free open source webui/GUI that helps you manage your Claude Code session and projects remotely #### [strands-agents/tools](https://github.com/strands-agents/tools) *Python · ★1,022 · Apache-2.0 · — · score:0.00 · hot:0.72 · rising:0.68 · durable:0.53 · board:hot · trend:up* A set of tools that gives agents powerful capabilities. #### [tisfeng/Easydict](https://github.com/tisfeng/Easydict) *Swift · ★12,900 · GPL-3.0 · — · score:0.00 · hot:0.72 · rising:0.72 · durable:0.58 · board:rising · trend:up* 一个简洁优雅的词典翻译 macOS App。开箱即用，支持离线 OCR 识别，支持有道词典，🍎 苹果系统词典，🍎 苹果系统翻译，OpenAI，Gemini，DeepL，Google，Bing，腾讯，百度，阿里，小牛，彩云和火山翻译。A concise and elegant Dictionary and Translator macOS App for looking up words and translating text. #### [vespa-engine/vespa](https://github.com/vespa-engine/vespa) *Java · ★6,881 · Apache-2.0 · — · score:0.00 · hot:0.71 · rising:0.71 · durable:0.51 · board:hot · trend:up* AI + Data, online. https://vespa.ai #### [rockbenben/ChatGPT-Shortcut](https://github.com/rockbenben/ChatGPT-Shortcut) *TypeScript · ★8,387 · MIT · — · score:0.00 · hot:0.71 · rising:0.73 · durable:0.61 · board:rising · trend:up* 🚀💪Maximize your efficiency and productivity. The ultimate hub to manage, customize, and share prompts. (English/中文/Español/العربية). 让生产力加倍的 AI 快捷指令。更高效地管理提示词，在分享社区中发现适用于不同场景的灵感。 #### [SolaceLabs/solace-agent-mesh](https://github.com/SolaceLabs/solace-agent-mesh) *Python · ★3,219 · Apache-2.0 · — · score:0.00 · hot:0.71 · rising:0.70 · durable:0.55 · board:hot · trend:up* An event-driven framework designed to build and orchestrate multi-agent AI systems. It enables seamless integration of AI agents with real-world data sources and systems, facilitating complex, multi-step workflows. #### [memodb-io/Acontext](https://github.com/memodb-io/Acontext) *TypeScript · ★3,329 · Apache-2.0 · — · score:0.00 · hot:0.71 · rising:0.70 · durable:0.60 · board:hot · trend:up* Agent Skills as a Memory Layer #### [alirezarezvani/claude-skills](https://github.com/alirezarezvani/claude-skills) *Python · ★11,911 · MIT · — · score:0.00 · hot:0.71 · rising:0.72 · durable:0.64 · board:rising · trend:up* 232+ Claude Code skills & agent plugins for Claude Code, Codex, Gemini CLI, Cursor, and 8 more coding agents — engineering, marketing, product, compliance, C-level advisory. #### [Kiln-AI/Kiln](https://github.com/Kiln-AI/Kiln) *Python · ★4,761 · NOASSERTION · — · score:0.00 · hot:0.71 · rising:0.69 · durable:0.56 · board:hot · trend:up* Build, Evaluate, and Optimize AI Systems. Includes evals, RAG, agents, fine-tuning, synthetic data generation, dataset management, MCP, and more. #### [flipped-aurora/gin-vue-admin](https://github.com/flipped-aurora/gin-vue-admin) *Go · ★24,585 · Apache-2.0 · — · score:0.00 · hot:0.71 · rising:0.76 · durable:0.61 · board:rising · trend:up* 🚀Vite+Vue3+Gin拥有AI辅助的基础开发平台，企业级业务AI+开发解决方案，内置mcp辅助服务，内置skills管理，支持TS和JS混用。它集成了JWT鉴权、权限管理、动态路由、显隐可控组件、分页封装、多点登录拦截、资源权限、上传下载、代码生成器、表单生成器和可配置的导入导出等开发必备功能。 #### [bentoml/BentoML](https://github.com/bentoml/BentoML) *Python · ★8,590 · Apache-2.0 · — · score:0.00 · hot:0.71 · rising:0.70 · durable:0.56 · board:hot · trend:up* The easiest way to serve AI apps and models - Build Model Inference APIs, Job queues, LLM apps, Multi-model pipelines, and more! #### [embeddings-benchmark/mteb](https://github.com/embeddings-benchmark/mteb) *Python · ★3,219 · Apache-2.0 · — · score:0.00 · hot:0.71 · rising:0.69 · durable:0.52 · board:hot · trend:up* MTEB: Massive Text Embedding Benchmark #### [elizaOS/eliza](https://github.com/elizaOS/eliza) *TypeScript · ★18,206 · MIT · — · score:0.00 · hot:0.71 · rising:0.76 · durable:0.64 · board:rising · trend:up* Autonomous agents for everyone #### [jarrodwatts/claude-hud](https://github.com/jarrodwatts/claude-hud) *JavaScript · ★19,977 · MIT · — · score:0.00 · hot:0.71 · rising:0.73 · durable:0.62 · board:rising · trend:up* A Claude Code plugin that shows what's happening - context usage, active tools, running agents, and todo progress #### [Narcooo/inkos](https://github.com/Narcooo/inkos) *TypeScript · ★4,605 · AGPL-3.0 · — · score:0.00 · hot:0.71 · rising:0.72 · durable:0.54 · board:rising · trend:up* Autonomous novel writing AI Agent — agents write, audit, and revise novels with human review gates #### [LazyAGI/LazyLLM](https://github.com/LazyAGI/LazyLLM) *Python · ★3,806 · Apache-2.0 · — · score:0.00 · hot:0.71 · rising:0.72 · durable:0.58 · board:rising · trend:up* Easiest and laziest way for building multi-agent LLMs applications. #### [vercel/vercel](https://github.com/vercel/vercel) *TypeScript · ★15,325 · Apache-2.0 · — · score:0.00 · hot:0.71 · rising:0.75 · durable:0.53 · board:rising · trend:up* Develop. Preview. Ship. #### [vllm-project/semantic-router](https://github.com/vllm-project/semantic-router) *Go · ★3,733 · Apache-2.0 · — · score:0.00 · hot:0.71 · rising:0.71 · durable:0.56 · board:hot · trend:up* System Level Intelligent Router for Mixture-of-Models at Cloud, Data Center and Edge #### [gpustack/gpustack](https://github.com/gpustack/gpustack) *Python · ★4,862 · Apache-2.0 · — · score:0.00 · hot:0.71 · rising:0.68 · durable:0.55 · board:hot · trend:up* A GPU cluster manager that configures and orchestrates inference engines like vLLM and SGLang for high-performance AI model deployment. #### [78/xiaozhi-esp32](https://github.com/78/xiaozhi-esp32) *C++ · ★25,784 · MIT · — · score:0.00 · hot:0.71 · rising:0.75 · durable:0.59 · board:rising · trend:up* An MCP-based chatbot | 一个基于MCP的聊天机器人 #### [yzhao062/pyod](https://github.com/yzhao062/pyod) *Python · ★9,801 · BSD-2-Clause · — · score:0.00 · hot:0.71 · rising:0.71 · durable:0.53 · board:rising · trend:up* A Python library for anomaly detection across tabular, time series, graph, text, and image data. 60+ detectors, benchmark-backed ADEngine orchestration, and an agentic workflow for AI agents. #### [run-llama/llama-agents](https://github.com/run-llama/llama-agents) *Python · ★347 · MIT · — · score:0.00 · hot:0.71 · rising:0.72 · durable:0.54 · board:rising · trend:up* Llama Agents + Workflows are an event-driven, async-first, step-based way to control the execution flow of AI applications like agents. #### [modelcontextprotocol/python-sdk](https://github.com/modelcontextprotocol/python-sdk) *Python · ★22,691 · MIT · — · score:0.00 · hot:0.71 · rising:0.76 · durable:0.60 · board:rising · trend:up* The official Python SDK for Model Context Protocol servers and clients #### [lancedb/lancedb](https://github.com/lancedb/lancedb) *HTML · ★9,997 · Apache-2.0 · — · score:0.00 · hot:0.71 · rising:0.72 · durable:0.53 · board:rising · trend:up* Developer-friendly OSS embedded retrieval library for multimodal AI. Search More; Manage Less. #### [modelcontextprotocol/inspector](https://github.com/modelcontextprotocol/inspector) *TypeScript · ★9,491 · NOASSERTION · — · score:0.00 · hot:0.70 · rising:0.73 · durable:0.54 · board:rising · trend:up* Visual testing tool for MCP servers #### [databendlabs/databend](https://github.com/databendlabs/databend) *Rust · ★9,256 · NOASSERTION · — · score:0.00 · hot:0.70 · rising:0.70 · durable:0.49 · board:rising · trend:up* Data Agent Ready Warehouse : One for Analytics, Search, AI, Python Sandbox. — rebuilt from scratch. Unified architecture on your S3. #### [oceanbase/oceanbase](https://github.com/oceanbase/oceanbase) *C++ · ★10,067 · NOASSERTION · — · score:0.00 · hot:0.70 · rising:0.71 · durable:0.51 · board:rising · trend:up* The Fastest Distributed Database for Transactional, Analytical, and AI Workloads. #### [kvcache-ai/Mooncake](https://github.com/kvcache-ai/Mooncake) *C++ · ★5,136 · Apache-2.0 · — · score:0.00 · hot:0.70 · rising:0.71 · durable:0.53 · board:rising · trend:up* Mooncake is the serving platform for Kimi, a leading LLM service provided by Moonshot AI. #### [pydantic/logfire](https://github.com/pydantic/logfire) *Python · ★4,178 · MIT · — · score:0.00 · hot:0.70 · rising:0.68 · durable:0.55 · board:hot · trend:up* AI observability platform for production LLM and agent systems. #### [Donchitos/Claude-Code-Game-Studios](https://github.com/Donchitos/Claude-Code-Game-Studios) *Shell · ★13,189 · MIT · — · score:0.00 · hot:0.70 · rising:0.73 · durable:0.63 · board:rising · trend:up* Turn Claude Code into a full game dev studio — 49 AI agents, 72 workflow skills, and a complete coordination system mirroring real studio hierarchy. #### [modelcontextprotocol/registry](https://github.com/modelcontextprotocol/registry) *Go · ★6,698 · NOASSERTION · — · score:0.00 · hot:0.70 · rising:0.71 · durable:0.53 · board:rising · trend:up* A community driven registry service for Model Context Protocol (MCP) servers. #### [NVIDIA-NeMo/DataDesigner](https://github.com/NVIDIA-NeMo/DataDesigner) *Python · ★1,646 · Apache-2.0 · — · score:0.00 · hot:0.70 · rising:0.67 · durable:0.53 · board:hot · trend:up* 🎨 NeMo Data Designer: Generate high-quality synthetic data from scratch or from seed data. #### [modelcontextprotocol/servers](https://github.com/modelcontextprotocol/servers) *TypeScript · ★84,096 · NOASSERTION · — · score:0.00 · hot:0.70 · rising:0.77 · durable:0.64 · board:rising · trend:up* Model Context Protocol Servers #### [asgeirtj/system_prompts_leaks](https://github.com/asgeirtj/system_prompts_leaks) *? · ★38,594 · MIT · — · score:0.00 · hot:0.70 · rising:0.72 · durable:0.61 · board:rising · trend:up* Extracted system prompts from ChatGPT (GPT-5.4, GPT-5.3, Codex), Claude (Opus 4.6, Sonnet 4.6, Claude Code), Gemini (3.1 Pro, 3 Flash, CLI), Grok (4.2, 4), Perplexity, and more. Updated regularly. #### [rerun-io/rerun](https://github.com/rerun-io/rerun) *Rust · ★10,555 · Apache-2.0 · — · score:0.00 · hot:0.70 · rising:0.71 · durable:0.53 · board:rising · trend:up* An open source SDK for logging, storing, querying, and visualizing multimodal and multi-rate data #### [oumi-ai/oumi](https://github.com/oumi-ai/oumi) *Python · ★9,185 · Apache-2.0 · — · score:0.00 · hot:0.70 · rising:0.71 · durable:0.62 · board:rising · trend:up* Easily fine-tune, evaluate and deploy gpt-oss, Qwen3, DeepSeek-R1, or any open source LLM / VLM! #### [microsoft/mimalloc](https://github.com/microsoft/mimalloc) *C · ★12,730 · MIT · — · score:0.00 · hot:0.70 · rising:0.73 · durable:0.53 · board:rising · trend:up* mimalloc is a compact general purpose allocator with excellent performance. #### [infiniflow/infinity](https://github.com/infiniflow/infinity) *C++ · ★4,483 · Apache-2.0 · — · score:0.00 · hot:0.70 · rising:0.70 · durable:0.56 · board:rising · trend:up* The AI-native database built for LLM applications, providing incredibly fast hybrid search of dense vector, sparse vector, tensor (multi-vector), and full-text. #### [langchain-ai/langgraphjs](https://github.com/langchain-ai/langgraphjs) *TypeScript · ★2,804 · MIT · — · score:0.00 · hot:0.70 · rising:0.70 · durable:0.53 · board:rising · trend:up* Framework to build resilient language agents as graphs. #### [groupultra/telegram-search](https://github.com/groupultra/telegram-search) *TypeScript · ★3,862 · AGPL-3.0 · — · score:0.00 · hot:0.70 · rising:0.70 · durable:0.57 · board:rising · trend:up* 🔍 导出并模糊搜索 Telegram 聊天记录 | Export and fuzzy search your Telegram chat history #### [manaflow-ai/cmux](https://github.com/manaflow-ai/cmux) *Swift · ★14,766 · NOASSERTION · — · score:0.00 · hot:0.70 · rising:0.70 · durable:0.52 · board:rising · trend:up* Ghostty-based macOS terminal with vertical tabs and notifications for AI coding agents #### [modelcontextprotocol/rust-sdk](https://github.com/modelcontextprotocol/rust-sdk) *Rust · ★3,319 · NOASSERTION · — · score:0.00 · hot:0.70 · rising:0.72 · durable:0.55 · board:rising · trend:up* The official Rust SDK for the Model Context Protocol #### [apache/seatunnel](https://github.com/apache/seatunnel) *Java · ★9,256 · Apache-2.0 · — · score:0.00 · hot:0.70 · rising:0.71 · durable:0.52 · board:rising · trend:up* SeaTunnel is a multimodal, high-performance, distributed, massive data integration tool. #### [nuwax-ai/nuwax](https://github.com/nuwax-ai/nuwax) *TypeScript · ★748 · Apache-2.0 · — · score:0.00 · hot:0.70 · rising:0.68 · durable:0.54 · board:hot · trend:up* Nuwax Agent OS - The world's first universal agent operating system, building your private vertical general-purpose agent. 通用智能体操作系统，打造你私有的垂类通用智能体。新一代AI应用设计、开发、实践平台，无需代码，轻松创建，适合各类人群，支持多种端发布及API，提供完善的工作流、插件以及应用开发能力，RAG知识库与数据表存储能力，MCP接入以及开放能力。 #### [pydantic/pydantic](https://github.com/pydantic/pydantic) *Python · ★27,519 · MIT · — · score:0.00 · hot:0.69 · rising:0.73 · durable:0.56 · board:rising · trend:up* Data validation using Python type hints #### [replicate/cog](https://github.com/replicate/cog) *Go · ★9,398 · Apache-2.0 · — · score:0.00 · hot:0.69 · rising:0.70 · durable:0.54 · board:rising · trend:up* Containers for machine learning #### [lucidrains/vit-pytorch](https://github.com/lucidrains/vit-pytorch) *Python · ★25,071 · MIT · — · score:0.00 · hot:0.69 · rising:0.74 · durable:0.60 · board:rising · trend:up* Implementation of Vision Transformer, a simple way to achieve SOTA in vision classification with only a single transformer encoder, in Pytorch #### [CaviraOSS/OpenMemory](https://github.com/CaviraOSS/OpenMemory) *TypeScript · ★3,975 · Apache-2.0 · — · score:0.00 · hot:0.69 · rising:0.69 · durable:0.63 · board:hot · trend:up* Local persistent memory store for LLM applications including claude desktop, github copilot, codex, antigravity, etc. #### [Nixtla/neuralforecast](https://github.com/Nixtla/neuralforecast) *Python · ★4,050 · Apache-2.0 · — · score:0.00 · hot:0.69 · rising:0.70 · durable:0.54 · board:rising · trend:up* Scalable and user friendly neural :brain: forecasting algorithms. #### [linshenkx/prompt-optimizer](https://github.com/linshenkx/prompt-optimizer) *TypeScript · ★26,533 · NOASSERTION · — · score:0.00 · hot:0.69 · rising:0.72 · durable:0.61 · board:rising · trend:up* An AI prompt optimizer for writing better prompts and getting better AI results. #### [modelcontextprotocol/typescript-sdk](https://github.com/modelcontextprotocol/typescript-sdk) *TypeScript · ★12,220 · NOASSERTION · — · score:0.00 · hot:0.69 · rising:0.72 · durable:0.55 · board:rising · trend:up* The official TypeScript SDK for Model Context Protocol servers and clients #### [Hacker-Valley-Media/Interceptor](https://github.com/Hacker-Valley-Media/Interceptor) *TypeScript · ★161 · Apache-2.0 · — · score:0.00 · hot:0.69 · rising:0.70 · durable:0.52 · board:rising · trend:up* Agent-driven Chrome extension for full browser control via CLI #### [OpenCoworkAI/open-cowork](https://github.com/OpenCoworkAI/open-cowork) *TypeScript · ★874 · MIT · — · score:0.00 · hot:0.69 · rising:0.67 · durable:0.51 · board:hot · trend:up* Open-source AI agent desktop app for Windows & macOS. One-click install Claude Code, MCP tools, and Skills — with sandbox isolation, multi-model support, and Feishu/Slack integration. #### [google/adk-java](https://github.com/google/adk-java) *Java · ★1,484 · Apache-2.0 · — · score:0.00 · hot:0.69 · rising:0.69 · durable:0.51 · board:hot · trend:up* An open-source, code-first Java toolkit for building, evaluating, and deploying sophisticated AI agents with flexibility and control. #### [run-llama/liteparse](https://github.com/run-llama/liteparse) *TypeScript · ★4,426 · Apache-2.0 · — · score:0.00 · hot:0.69 · rising:0.72 · durable:0.59 · board:rising · trend:up* A fast, helpful, and open-source document parser #### [agentscope-ai/agentscope-java](https://github.com/agentscope-ai/agentscope-java) *Java · ★2,647 · no-license · — · score:0.00 · hot:0.69 · rising:0.68 · durable:0.50 · board:hot · trend:up* AgentScope Java: Agent-Oriented Programming for Building LLM Applications #### [pydantic/monty](https://github.com/pydantic/monty) *Rust · ★6,861 · MIT · — · score:0.00 · hot:0.69 · rising:0.71 · durable:0.58 · board:rising · trend:up* A minimal, secure Python interpreter written in Rust for use by AI #### [photoprism/photoprism](https://github.com/photoprism/photoprism) *Go · ★39,557 · NOASSERTION · — · score:0.00 · hot:0.69 · rising:0.72 · durable:0.55 · board:rising · trend:up* AI-Powered Photos App for the Decentralized Web 🌈💎✨ #### [aiming-lab/MetaClaw](https://github.com/aiming-lab/MetaClaw) *Python · ★3,456 · MIT · — · score:0.00 · hot:0.69 · rising:0.69 · durable:0.60 · board:hot · trend:up* 🦞 Just talk to your agent — it learns and EVOLVES 🧬. #### [RunanywhereAI/runanywhere-sdks](https://github.com/RunanywhereAI/runanywhere-sdks) *C++ · ★10,343 · NOASSERTION · — · score:0.00 · hot:0.69 · rising:0.66 · durable:0.58 · board:hot · trend:up* Production ready toolkit to run AI locally #### [open-webui/open-terminal](https://github.com/open-webui/open-terminal) *Python · ★2,334 · MIT · — · score:0.00 · hot:0.69 · rising:0.71 · durable:0.59 · board:rising · trend:up* A computer you can curl ⚡ #### [Nagi-ovo/gemini-voyager](https://github.com/Nagi-ovo/gemini-voyager) *TypeScript · ★16,990 · GPL-3.0 · — · score:0.00 · hot:0.69 · rising:0.69 · durable:0.60 · board:hot · trend:up* An all-in-one enhancement suite for Google Gemini & AI Studio - timeline navigation, folder management, prompt library, and chat export in one powerful extension. / Google Gemini & AI Studio 全能增强插件：集成时间轴导航、文件夹管理、提示词库及聊天导出等众多功能。 #### [taylorwilsdon/google_workspace_mcp](https://github.com/taylorwilsdon/google_workspace_mcp) *Python · ★2,148 · MIT · — · score:0.00 · hot:0.69 · rising:0.69 · durable:0.52 · board:hot · trend:up* Control Gmail, Google Calendar, Docs, Sheets, Slides, Chat, Forms, Tasks, Search & Drive with AI - Comprehensive Google Workspace / G Suite MCP Server & CLI Tool #### [go-vgo/robotgo](https://github.com/go-vgo/robotgo) *Go · ★10,685 · Apache-2.0 · — · score:0.00 · hot:0.69 · rising:0.72 · durable:0.54 · board:rising · trend:up* RobotGo, Go Native cross-platform RPA, GUI automation, Auto test and Computer use @vcaesar #### [dmlc/xgboost](https://github.com/dmlc/xgboost) *C++ · ★28,290 · Apache-2.0 · — · score:0.00 · hot:0.69 · rising:0.75 · durable:0.56 · board:rising · trend:up* Scalable, Portable and Distributed Gradient Boosting (GBDT, GBRT or GBM) Library, for Python, R, Java, Scala, C++ and more. Runs on single machine, Hadoop, Spark, Dask, Flink and DataFlow #### [xonsh/xonsh](https://github.com/xonsh/xonsh) *Python · ★9,289 · NOASSERTION · — · score:0.00 · hot:0.69 · rising:0.70 · durable:0.51 · board:rising · trend:up* 🐚 Python-powered shell. Full-featured, cross-platform and AI-friendly. #### [microsoft/PyRIT](https://github.com/microsoft/PyRIT) *Python · ★3,700 · MIT · — · score:0.00 · hot:0.69 · rising:0.72 · durable:0.54 · board:rising · trend:up* The Python Risk Identification Tool for generative AI (PyRIT) is an open source framework built to empower security professionals and engineers to proactively identify risks in generative AI systems. #### [OpenCSGs/csghub](https://github.com/OpenCSGs/csghub) *Vue · ★4,674 · Apache-2.0 · — · score:0.00 · hot:0.69 · rising:0.71 · durable:0.59 · board:rising · trend:up* CSGHub is a brand-new open-source platform for managing LLMs, developed by the OpenCSG team. It offers both open-source and on-premise/SaaS solutions, with features comparable to Hugging Face. Gain full control over the lifecycle of LLMs, datasets, and agents, with Python SDK compatibility with Hugging Face. Join us! ⭐️ #### [modelcontextprotocol/ruby-sdk](https://github.com/modelcontextprotocol/ruby-sdk) *Ruby · ★782 · NOASSERTION · — · score:0.00 · hot:0.69 · rising:0.70 · durable:0.51 · board:rising · trend:up* The official Ruby SDK for the Model Context Protocol. #### [AmrDab/clawdcursor](https://github.com/AmrDab/clawdcursor) *TypeScript · ★191 · MIT · — · score:0.00 · hot:0.69 · rising:0.65 · durable:0.51 · board:hot · trend:up* OS-agnostic, model-agnostic desktop automation server. Gives any AI agent eyes, hands, and ground-truth verification on Windows, macOS, and Linux. #### [triton-inference-server/server](https://github.com/triton-inference-server/server) *Python · ★10,584 · BSD-3-Clause · — · score:0.00 · hot:0.69 · rising:0.71 · durable:0.52 · board:rising · trend:up* The Triton Inference Server provides an optimized cloud and edge inferencing solution. #### [microsoft/azure-pipelines-tasks](https://github.com/microsoft/azure-pipelines-tasks) *TypeScript · ★3,628 · MIT · — · score:0.00 · hot:0.68 · rising:0.73 · durable:0.50 · board:rising · trend:up* Tasks for Azure Pipelines #### [microsoft/msquic](https://github.com/microsoft/msquic) *C · ★4,675 · MIT · — · score:0.00 · hot:0.68 · rising:0.71 · durable:0.51 · board:rising · trend:up* Cross-platform, C implementation of the IETF QUIC protocol, exposed to C, C++, C# and Rust. #### [IAHispano/Applio](https://github.com/IAHispano/Applio) *Python · ★3,176 · MIT · — · score:0.00 · hot:0.68 · rising:0.71 · durable:0.56 · board:rising · trend:up* A simple, high-quality voice conversion tool focused on ease of use and performance. #### [roboflow/inference](https://github.com/roboflow/inference) *Python · ★2,254 · NOASSERTION · — · score:0.00 · hot:0.68 · rising:0.67 · durable:0.48 · board:hot · trend:up* Turn any computer or edge device into a command center for your computer vision projects. #### [SWE-agent/mini-swe-agent](https://github.com/SWE-agent/mini-swe-agent) *Python · ★3,904 · MIT · — · score:0.00 · hot:0.68 · rising:0.70 · durable:0.57 · board:rising · trend:up* The 100 line AI agent that solves GitHub issues or helps you in your command line. Radically simple, no huge configs, no giant monorepo—but scores >74% on SWE-bench verified! #### [modelcontextprotocol/go-sdk](https://github.com/modelcontextprotocol/go-sdk) *Go · ★4,391 · NOASSERTION · — · score:0.00 · hot:0.68 · rising:0.70 · durable:0.56 · board:rising · trend:up* The official Go SDK for Model Context Protocol servers and clients. Maintained in collaboration with Google. #### [spacedriveapp/spacebot](https://github.com/spacedriveapp/spacebot) *Rust · ★2,110 · NOASSERTION · — · score:0.00 · hot:0.68 · rising:0.67 · durable:0.49 · board:hot · trend:up* An AI agent for teams, communities, and multi-user environments. #### [coze-dev/coze-studio](https://github.com/coze-dev/coze-studio) *TypeScript · ★20,568 · Apache-2.0 · — · score:0.00 · hot:0.68 · rising:0.71 · durable:0.59 · board:rising · trend:up* An AI agent development platform with all-in-one visual tools, simplifying agent creation, debugging, and deployment like never before. Coze your way to AI Agent creation. #### [matrixorigin/matrixone](https://github.com/matrixorigin/matrixone) *Go · ★1,842 · Apache-2.0 · — · score:0.00 · hot:0.68 · rising:0.68 · durable:0.49 · board:hot · trend:up* AI-native HTAP database with Git-for-Data and built-in vector search, serving as the data and memory backbone for intelligent agents and applications. #### [oceanbase/seekdb](https://github.com/oceanbase/seekdb) *C++ · ★2,526 · Apache-2.0 · — · score:0.00 · hot:0.68 · rising:0.67 · durable:0.52 · board:hot · trend:up* The AI-Native Search Database. Unifies vector, text, structured and semi-structured data in a single engine, enabling hybrid search and in-database AI workflows. #### [simonw/llm](https://github.com/simonw/llm) *Python · ★11,644 · Apache-2.0 · — · score:0.00 · hot:0.68 · rising:0.69 · durable:0.54 · board:rising · trend:up* Access large language models from the command-line #### [qdrant/fastembed](https://github.com/qdrant/fastembed) *Python · ★2,871 · Apache-2.0 · — · score:0.00 · hot:0.68 · rising:0.66 · durable:0.54 · board:hot · trend:up* Fast, Accurate, Lightweight Python library to make State of the Art Embedding #### [UnicomAI/wanwu](https://github.com/UnicomAI/wanwu) *Go · ★3,368 · Apache-2.0 · — · score:0.00 · hot:0.68 · rising:0.64 · durable:0.59 · board:hot · trend:stable* China Unicom's Yuanjing Wanwu Agent Platform is an enterprise-grade, multi-tenant AI agent development platform. It helps users build applications such as intelligent agents, workflows, and rag, and also supports model management. The platform features a developer-friendly license, and we welcome all developers to build upon the platform. #### [Eventual-Inc/Daft](https://github.com/Eventual-Inc/Daft) *Rust · ★5,427 · Apache-2.0 · — · score:0.00 · hot:0.68 · rising:0.69 · durable:0.51 · board:rising · trend:up* High-performance data engine for AI and multimodal workloads. Process images, audio, video, and structured data at any scale #### [argmaxinc/argmax-oss-swift](https://github.com/argmaxinc/argmax-oss-swift) *Swift · ★6,011 · MIT · — · score:0.00 · hot:0.68 · rising:0.69 · durable:0.56 · board:rising · trend:up* On-device Speech AI for Apple Silicon #### [microsoft/DirectXTK12](https://github.com/microsoft/DirectXTK12) *C++ · ★1,717 · MIT · — · score:0.00 · hot:0.68 · rising:0.71 · durable:0.53 · board:rising · trend:up* The DirectX Tool Kit (aka DirectXTK12) is a collection of helper classes for writing DirectX 12 code in C++ #### [modelcontextprotocol/ext-apps](https://github.com/modelcontextprotocol/ext-apps) *TypeScript · ★2,082 · NOASSERTION · — · score:0.00 · hot:0.68 · rising:0.68 · durable:0.50 · board:rising · trend:up* Official repo for spec & SDK of MCP Apps protocol - standard for UIs embedded AI chatbots, served by MCP servers #### [vllm-project/vllm-ascend](https://github.com/vllm-project/vllm-ascend) *Python · ★1,948 · Apache-2.0 · — · score:0.00 · hot:0.68 · rising:0.68 · durable:0.55 · board:rising · trend:up* Community maintained hardware plugin for vLLM on Ascend #### [crate/crate](https://github.com/crate/crate) *Java · ★4,388 · Apache-2.0 · — · score:0.00 · hot:0.68 · rising:0.69 · durable:0.49 · board:rising · trend:up* CrateDB is a distributed and scalable SQL database for storing and analyzing massive amounts of data in near real-time, even with complex queries. It is PostgreSQL-compatible, and based on Lucene. #### [modelcontextprotocol/csharp-sdk](https://github.com/modelcontextprotocol/csharp-sdk) *C# · ★4,207 · NOASSERTION · — · score:0.00 · hot:0.68 · rising:0.69 · durable:0.52 · board:rising · trend:up* The official C# SDK for Model Context Protocol servers and clients. Maintained in collaboration with Microsoft. #### [MariaDB/server](https://github.com/MariaDB/server) *C++ · ★7,460 · GPL-2.0 · — · score:0.00 · hot:0.68 · rising:0.70 · durable:0.51 · board:rising · trend:up* MariaDB server is a community developed fork of MySQL server. Started by core members of the original MySQL team, MariaDB actively works with outside developers to deliver the most featureful, stable, and sanely licensed open SQL server in the industry. #### [CelestoAI/SmolVM](https://github.com/CelestoAI/SmolVM) *Python · ★334 · Apache-2.0 · — · score:0.00 · hot:0.67 · rising:0.67 · durable:0.52 · board:hot · trend:up* Open-source AI sandbox infrastructure for code execution, browser use, and AI agents. #### [EvolvingLMMs-Lab/lmms-eval](https://github.com/EvolvingLMMs-Lab/lmms-eval) *Python · ★4,037 · NOASSERTION · — · score:0.00 · hot:0.67 · rising:0.66 · durable:0.55 · board:hot · trend:up* One-for-All Multimodal Evaluation Toolkit Across Text, Image, Video, and Audio Tasks #### [vortex-data/vortex](https://github.com/vortex-data/vortex) *Rust · ★2,878 · Apache-2.0 · — · score:0.00 · hot:0.67 · rising:0.67 · durable:0.50 · board:hot · trend:up* An extensible, state-of-the-art framework for columnar compression, and the fastest FOSS columnar file format. Formerly at @spiraldb, now an Incubation Stage project at LFAI&Data, part of the Linux Foundation. #### [TurixAI/TuriX-CUA](https://github.com/TurixAI/TuriX-CUA) *Python · ★2,266 · MIT · — · score:0.00 · hot:0.67 · rising:0.69 · durable:0.58 · board:rising · trend:up* This is the official website for TuriX Computer-use-Agent #### [danny-avila/LibreChat](https://github.com/danny-avila/LibreChat) *TypeScript · ★35,765 · MIT · — · score:0.00 · hot:0.67 · rising:0.69 · durable:0.57 · board:rising · trend:up* Enhanced ChatGPT Clone: Features Agents, MCP, DeepSeek, Anthropic, AWS, OpenAI, Responses API, Azure, Groq, o1, GPT-5, Mistral, OpenRouter, Vertex AI, Gemini, Artifacts, AI model switching, message search, Code Interpreter, langchain, DALL-E-3, OpenAPI Actions, Functions, Secure Multi-User Auth, Presets, open-source for self-hosting. Active. #### [vllm-project/vllm-metal](https://github.com/vllm-project/vllm-metal) *Python · ★933 · Apache-2.0 · — · score:0.00 · hot:0.67 · rising:0.69 · durable:0.53 · board:rising · trend:up* Community maintained hardware plugin for vLLM on Apple Silicon #### [chatboxai/chatbox](https://github.com/chatboxai/chatbox) *TypeScript · ★39,511 · GPL-3.0 · — · score:0.00 · hot:0.67 · rising:0.69 · durable:0.58 · board:rising · trend:up* Powerful AI Client #### [langchain-ai/langchain-google](https://github.com/langchain-ai/langchain-google) *Python · ★365 · MIT · — · score:0.00 · hot:0.67 · rising:0.67 · durable:0.50 · board:rising · trend:up* 🦜🔗 LangChain interfaces to Google's suite of AI products (Gemini & Vertex AI) #### [dottxt-ai/outlines](https://github.com/dottxt-ai/outlines) *Python · ★13,685 · Apache-2.0 · — · score:0.00 · hot:0.67 · rising:0.70 · durable:0.59 · board:rising · trend:up* Structured Outputs #### [microsoft/fluentui](https://github.com/microsoft/fluentui) *TypeScript · ★19,938 · NOASSERTION · — · score:0.00 · hot:0.67 · rising:0.71 · durable:0.50 · board:rising · trend:up* Fluent UI web represents a collection of utilities, React components, and web components for building web applications. #### [sgl-project/ome](https://github.com/sgl-project/ome) *Go · ★424 · Apache-2.0 · — · score:0.00 · hot:0.67 · rising:0.66 · durable:0.50 · board:hot · trend:up* Open Model Engine (OME) — Kubernetes operator for LLM serving, GPU scheduling, and model lifecycle management. Works with SGLang, vLLM, TensorRT-LLM, and Triton #### [SeekStorm/SeekStorm](https://github.com/SeekStorm/SeekStorm) *Rust · ★1,866 · Apache-2.0 · — · score:0.00 · hot:0.67 · rising:0.65 · durable:0.54 · board:hot · trend:up* SeekStorm: vector & lexical search - in-process library & multi-tenancy server, in Rust. #### [shanraisshan/claude-code-best-practice](https://github.com/shanraisshan/claude-code-best-practice) *HTML · ★46,518 · MIT · — · score:0.00 · hot:0.67 · rising:0.72 · durable:0.59 · board:rising · trend:up* from vibe coding to agentic engineering - practice makes claude perfect #### [vercel/workflow](https://github.com/vercel/workflow) *TypeScript · ★1,924 · Apache-2.0 · — · score:0.00 · hot:0.67 · rising:0.69 · durable:0.51 · board:rising · trend:up* Workflow SDK: Build durable, reliable, and observable apps and AI Agents in TypeScript #### [elder-plinius/CL4R1T4S](https://github.com/elder-plinius/CL4R1T4S) *? · ★16,041 · AGPL-3.0 · — · score:0.00 · hot:0.67 · rising:0.69 · durable:0.58 · board:rising · trend:up* LEAKED SYSTEM PROMPTS FOR CHATGPT, GEMINI, GROK, CLAUDE, PERPLEXITY, CURSOR, DEVIN, REPLIT, AND MORE! - AI SYSTEMS TRANSPARENCY FOR ALL! 👐 #### [Tencent/ncnn](https://github.com/Tencent/ncnn) *C++ · ★23,118 · NOASSERTION · — · score:0.00 · hot:0.67 · rising:0.69 · durable:0.52 · board:rising · trend:up* ncnn is a high-performance neural network inference framework optimized for the mobile platform #### [HelixDB/helix-db](https://github.com/HelixDB/helix-db) *Rust · ★4,077 · AGPL-3.0 · — · score:0.00 · hot:0.67 · rising:0.67 · durable:0.57 · board:rising · trend:up* HelixDB is an open-source graph-vector database built from scratch in Rust. #### [microsoft/bioemu](https://github.com/microsoft/bioemu) *Python · ★789 · MIT · — · score:0.00 · hot:0.67 · rising:0.70 · durable:0.56 · board:rising · trend:up* Inference code for scalable emulation of protein equilibrium ensembles with generative deep learning #### [davila7/claude-code-templates](https://github.com/davila7/claude-code-templates) *Python · ★24,745 · MIT · — · score:0.00 · hot:0.67 · rising:0.73 · durable:0.62 · board:rising · trend:up* CLI tool for configuring and monitoring Claude Code #### [hesreallyhim/awesome-claude-code](https://github.com/hesreallyhim/awesome-claude-code) *Python · ★39,715 · NOASSERTION · — · score:0.00 · hot:0.67 · rising:0.69 · durable:0.55 · board:rising · trend:up* A curated list of awesome skills, hooks, slash-commands, agent orchestrators, applications, and plugins for Claude Code by Anthropic #### [typedb/typedb](https://github.com/typedb/typedb) *Rust · ★4,291 · MPL-2.0 · — · score:0.00 · hot:0.67 · rising:0.68 · durable:0.48 · board:rising · trend:up* TypeDB: Built for systems, not records #### [pykeio/ort](https://github.com/pykeio/ort) *Rust · ★2,186 · Apache-2.0 · — · score:0.00 · hot:0.67 · rising:0.68 · durable:0.56 · board:rising · trend:up* Fast ML inference & training for ONNX models in Rust #### [aiming-lab/SimpleMem](https://github.com/aiming-lab/SimpleMem) *Python · ★3,261 · MIT · — · score:0.00 · hot:0.67 · rising:0.64 · durable:0.62 · board:hot · trend:stable* SimpleMem: Efficient Lifelong Memory for LLM Agents — Text & Multimodal #### [huggingface/optimum-habana](https://github.com/huggingface/optimum-habana) *Python · ★209 · Apache-2.0 · — · score:0.00 · hot:0.67 · rising:0.67 · durable:0.48 · board:rising · trend:up* Easy and lightning fast training of 🤗 Transformers on Habana Gaudi processor (HPU) #### [anthropics/anthropic-sdk-csharp](https://github.com/anthropics/anthropic-sdk-csharp) *C# · ★239 · MIT · — · score:0.00 · hot:0.67 · rising:0.68 · durable:0.50 · board:rising · trend:up* Access to Anthropic's safety-first language model APIs in C# #### [langchain-ai/deepagentsjs](https://github.com/langchain-ai/deepagentsjs) *TypeScript · ★1,122 · MIT · — · score:0.00 · hot:0.67 · rising:0.67 · durable:0.51 · board:rising · trend:up* Agent harness built with LangChain and LangGraph. Equipped with a planning tool, a filesystem backend, and the ability to spawn subagents - well-equipped to handle complex agentic tasks. #### [microsoft/DirectXTex](https://github.com/microsoft/DirectXTex) *C++ · ★2,092 · MIT · — · score:0.00 · hot:0.67 · rising:0.70 · durable:0.50 · board:rising · trend:up* DirectXTex texture processing library #### [OpenRouterTeam/ai-sdk-provider](https://github.com/OpenRouterTeam/ai-sdk-provider) *TypeScript · ★626 · Apache-2.0 · — · score:0.00 · hot:0.66 · rising:0.69 · durable:0.55 · board:rising · trend:up* The OpenRouter provider for the Vercel AI SDK contains support for hundreds of models through the OpenRouter chat and completion APIs. #### [langchain-ai/langsmith-sdk](https://github.com/langchain-ai/langsmith-sdk) *Python · ★853 · MIT · — · score:0.00 · hot:0.66 · rising:0.67 · durable:0.49 · board:rising · trend:up* LangSmith Client SDK Implementations #### [nndeploy/nndeploy](https://github.com/nndeploy/nndeploy) *C++ · ★1,788 · Apache-2.0 · — · score:0.00 · hot:0.66 · rising:0.66 · durable:0.55 · board:hot · trend:up* 一款简单易用和高性能的AI部署框架 | An Easy-to-Use and High-Performance AI Deployment Framework #### [Lightning-AI/pytorch-lightning](https://github.com/Lightning-AI/pytorch-lightning) *Python · ★31,058 · Apache-2.0 · — · score:0.00 · hot:0.66 · rising:0.70 · durable:0.54 · board:rising · trend:up* Pretrain, finetune ANY AI model of ANY size on 1 or 10,000+ GPUs with zero code changes. #### [AlexAnys/awesome-openclaw-usecases-zh](https://github.com/AlexAnys/awesome-openclaw-usecases-zh) *? · ★3,835 · MIT · — · score:0.00 · hot:0.66 · rising:0.67 · durable:0.58 · board:rising · trend:up* 🇨🇳 OpenClaw中文用例大全 | 46个真实场景 | 国内特色 + 海外案例的国内适配 | 自动化办公·内容创作·运维·AI助理·知识管理 | 新手友好 | Chinese guide for OpenClaw AI agent use cases #### [RVC-Boss/GPT-SoVITS](https://github.com/RVC-Boss/GPT-SoVITS) *Python · ★56,801 · MIT · — · score:0.00 · hot:0.66 · rising:0.73 · durable:0.61 · board:rising · trend:up* 1 min voice data can also be used to train a good TTS model! (few shot voice cloning) #### [Canner/WrenAI](https://github.com/Canner/WrenAI) *TypeScript · ★14,964 · AGPL-3.0 · — · score:0.00 · hot:0.66 · rising:0.67 · durable:0.60 · board:rising · trend:up* Open-source text-to-SQL and text-to-chart GenBI agent with a semantic layer. Ask your database questions in natural language — get accurate SQL, charts, and BI insights. Supports 12+ data sources (PostgreSQL, BigQuery, Snowflake, etc.) and any LLM (OpenAI, Claude, Gemini, Ollama). #### [wshobson/agents](https://github.com/wshobson/agents) *Python · ★33,887 · MIT · — · score:0.00 · hot:0.66 · rising:0.71 · durable:0.59 · board:rising · trend:up* Intelligent automation and multi-agent orchestration for Claude Code #### [google/adk-js](https://github.com/google/adk-js) *TypeScript · ★1,014 · Apache-2.0 · — · score:0.00 · hot:0.66 · rising:0.65 · durable:0.50 · board:hot · trend:up* An open-source, code-first Typescript toolkit for building, evaluating, and deploying sophisticated AI agents with flexibility and control. #### [oceanbase/powermem](https://github.com/oceanbase/powermem) *Python · ★630 · NOASSERTION · — · score:0.00 · hot:0.66 · rising:0.63 · durable:0.47 · board:hot · trend:up* PowerMem: Your AI-Powered Long-Term Memory — Accurate, Agile, Affordable. Also friendly support for the OpenClaw Memory Plugin. #### [cheahjs/free-llm-api-resources](https://github.com/cheahjs/free-llm-api-resources) *Python · ★19,090 · no-license · — · score:0.00 · hot:0.66 · rising:0.68 · durable:0.55 · board:rising · trend:up* A list of free LLM inference resources accessible via API. #### [facebookresearch/balance](https://github.com/facebookresearch/balance) *Python · ★741 · MIT · — · score:0.00 · hot:0.66 · rising:0.68 · durable:0.52 · board:rising · trend:up* The balance python package offers a simple workflow and methods for dealing with biased data samples when looking to infer from them to some target population of interest. #### [microsoft/testfx](https://github.com/microsoft/testfx) *C# · ★1,002 · MIT · — · score:0.00 · hot:0.66 · rising:0.68 · durable:0.46 · board:rising · trend:up* This repository holds the source code of Microsoft.Testing.Platform (MTP), a lightweight alternative to VSTest, as well as MSTest adapter and framework. #### [Saganaki22/ComfyUI-OmniVoice-TTS](https://github.com/Saganaki22/ComfyUI-OmniVoice-TTS) *Python · ★334 · Apache-2.0 · — · score:0.00 · hot:0.66 · rising:0.67 · durable:0.52 · board:rising · trend:up* OmniVoice TTS nodes for ComfyUI - Zero-shot multilingual text-to-speech with voice cloning, voice design, and multi-speaker dialogue #### [microsoft/CsWin32](https://github.com/microsoft/CsWin32) *C# · ★2,479 · MIT · — · score:0.00 · hot:0.66 · rising:0.68 · durable:0.49 · board:rising · trend:up* A source generator to add a user-defined set of Win32 P/Invoke methods and supporting types to a C# project. #### [vercel/chat](https://github.com/vercel/chat) *TypeScript · ★1,756 · MIT · — · score:0.00 · hot:0.66 · rising:0.68 · durable:0.53 · board:rising · trend:up* A unified TypeScript SDK for building chat bots across Slack, Microsoft Teams, Google Chat, Discord, and more. #### [microsoft/WindowsAppSDK](https://github.com/microsoft/WindowsAppSDK) *C++ · ★4,501 · MIT · — · score:0.00 · hot:0.66 · rising:0.70 · durable:0.51 · board:rising · trend:up* The Windows App SDK empowers all Windows desktop apps with modern Windows UI, APIs, and platform features, including back-compat support, shipped via NuGet. #### [GetBindu/Bindu](https://github.com/GetBindu/Bindu) *Python · ★4,241 · NOASSERTION · — · score:0.00 · hot:0.66 · rising:0.65 · durable:0.49 · board:hot · trend:up* Bindu: Turn any AI agent into a living microservice - interoperable, observable, composable. #### [cuga-project/cuga-agent](https://github.com/cuga-project/cuga-agent) *Python · ★704 · NOASSERTION · — · score:0.00 · hot:0.66 · rising:0.63 · durable:0.46 · board:hot · trend:up* CUGA is an open-source generalist agent harness for the enterprise, supporting complex task execution on web and APIs, OpenAPI/MCP integrations, composable architecture, reasoning modes, and policy-aware features. #### [NirDiamant/GenAI_Agents](https://github.com/NirDiamant/GenAI_Agents) *Jupyter Notebook · ★21,376 · NOASSERTION · — · score:0.00 · hot:0.66 · rising:0.67 · durable:0.56 · board:rising · trend:up* 50+ tutorials and implementations for Generative AI Agent techniques, from basic conversational bots to complex multi-agent systems. #### [anthropics/anthropic-sdk-java](https://github.com/anthropics/anthropic-sdk-java) *Kotlin · ★294 · MIT · — · score:0.00 · hot:0.66 · rising:0.67 · durable:0.49 · board:rising · trend:up* #### [microsoft/DiskANN](https://github.com/microsoft/DiskANN) *Rust · ★1,768 · MIT · — · score:0.00 · hot:0.66 · rising:0.68 · durable:0.49 · board:rising · trend:up* Graph-structured Indices for Scalable, Fast, Fresh and Filtered Approximate Nearest Neighbor Search #### [modelcontextprotocol/kotlin-sdk](https://github.com/modelcontextprotocol/kotlin-sdk) *Kotlin · ★1,339 · NOASSERTION · — · score:0.00 · hot:0.66 · rising:0.66 · durable:0.49 · board:hot · trend:up* The official Kotlin SDK for Model Context Protocol servers and clients. Maintained in collaboration with JetBrains #### [milvus-io/pymilvus](https://github.com/milvus-io/pymilvus) *Python · ★1,368 · Apache-2.0 · — · score:0.00 · hot:0.66 · rising:0.67 · durable:0.49 · board:rising · trend:up* Python SDK for Milvus Vector Database #### [vercel/swr](https://github.com/vercel/swr) *TypeScript · ★32,352 · MIT · — · score:0.00 · hot:0.66 · rising:0.70 · durable:0.61 · board:rising · trend:up* React Hooks for Data Fetching #### [microsoft/WindowsProtocolTestSuites](https://github.com/microsoft/WindowsProtocolTestSuites) *C# · ★557 · NOASSERTION · — · score:0.00 · hot:0.66 · rising:0.67 · durable:0.43 · board:rising · trend:up* ⭐⭐ Join us at SambaXP for the SMB3 IO Lab (April 20-23, 2026), see upcoming Interoperability Events #### [vercel/flags](https://github.com/vercel/flags) *TypeScript · ★585 · MIT · — · score:0.00 · hot:0.66 · rising:0.67 · durable:0.49 · board:rising · trend:up* Flags SDK by Vercel #### [GoogleCloudPlatform/generative-ai](https://github.com/GoogleCloudPlatform/generative-ai) *Jupyter Notebook · ★16,688 · Apache-2.0 · — · score:0.00 · hot:0.66 · rising:0.69 · durable:0.56 · board:rising · trend:up* Sample code and notebooks for Generative AI on Google Cloud, with Gemini on Vertex AI #### [OpenCowAI/opencow](https://github.com/OpenCowAI/opencow) *TypeScript · ★374 · Apache-2.0 · — · score:0.00 · hot:0.66 · rising:0.65 · durable:0.52 · board:hot · trend:up* One task, one agent, delivered. The open-source platform for task-driven autonomous AI agents.OpenCow assigns an autonomous AI agent to every task — features, campaigns, reports, audits — and delivers them in parallel. Full context. Full control. Every department. 🐄 #### [langchain-ai/langchain-aws](https://github.com/langchain-ai/langchain-aws) *Python · ★318 · MIT · — · score:0.00 · hot:0.66 · rising:0.67 · durable:0.48 · board:rising · trend:up* Build LangChain Applications on AWS #### [weaviate/weaviate-python-client](https://github.com/weaviate/weaviate-python-client) *Python · ★221 · BSD-3-Clause · — · score:0.00 · hot:0.66 · rising:0.66 · durable:0.45 · board:rising · trend:up* A python native client for easy interaction with a Weaviate instance. #### [microsoft/vscode-pull-request-github](https://github.com/microsoft/vscode-pull-request-github) *TypeScript · ★2,572 · MIT · — · score:0.00 · hot:0.66 · rising:0.69 · durable:0.49 · board:rising · trend:up* GitHub Pull Requests for Visual Studio Code #### [langchain-ai/langchain-mcp-adapters](https://github.com/langchain-ai/langchain-mcp-adapters) *Python · ★3,490 · MIT · — · score:0.00 · hot:0.65 · rising:0.68 · durable:0.55 · board:rising · trend:up* LangChain 🔌 MCP #### [NVIDIA/NVSentinel](https://github.com/NVIDIA/NVSentinel) *Go · ★260 · Apache-2.0 · — · score:0.00 · hot:0.65 · rising:0.66 · durable:0.47 · board:rising · trend:up* NVSentinel is a cross-platform fault remediation service designed to rapidly remediate runtime node-level issues in GPU-accelerated computing environments #### [GaiZhenbiao/ChuanhuChatGPT](https://github.com/GaiZhenbiao/ChuanhuChatGPT) *Python · ★15,342 · GPL-3.0 · — · score:0.00 · hot:0.65 · rising:0.70 · durable:0.60 · board:rising · trend:up* GUI for ChatGPT API and many LLMs. Supports agents, file-based QA, GPT finetuning and query with web search. All with a neat UI. #### [alibaba/ROLL](https://github.com/alibaba/ROLL) *Python · ★3,090 · Apache-2.0 · — · score:0.00 · hot:0.65 · rising:0.66 · durable:0.51 · board:rising · trend:up* An Efficient and User-Friendly Scaling Library for Reinforcement Learning with Large Language Models #### [modelcontextprotocol/swift-sdk](https://github.com/modelcontextprotocol/swift-sdk) *Swift · ★1,352 · NOASSERTION · — · score:0.00 · hot:0.65 · rising:0.65 · durable:0.49 · board:rising · trend:up* The official Swift SDK for Model Context Protocol servers and clients. #### [langchain-ai/openevals](https://github.com/langchain-ai/openevals) *Python · ★1,029 · MIT · — · score:0.00 · hot:0.65 · rising:0.67 · durable:0.54 · board:rising · trend:up* Readymade evaluators for your LLM apps #### [wrtnlabs/agentica](https://github.com/wrtnlabs/agentica) *TypeScript · ★1,021 · MIT · — · score:0.00 · hot:0.65 · rising:0.61 · durable:0.52 · board:hot · trend:stable* TypeScript AI AI Function Calling Framework enhanced by compiler skills. #### [github/awesome-copilot](https://github.com/github/awesome-copilot) *Python · ★30,418 · MIT · — · score:0.00 · hot:0.65 · rising:0.70 · durable:0.57 · board:rising · trend:up* Community-contributed instructions, agents, skills, and configurations to help you make the most of GitHub Copilot. #### [lucidrains/vector-quantize-pytorch](https://github.com/lucidrains/vector-quantize-pytorch) *Python · ★3,903 · MIT · — · score:0.00 · hot:0.65 · rising:0.66 · durable:0.54 · board:rising · trend:up* Vector (and Scalar) Quantization, in Pytorch #### [vllm-project/llm-compressor](https://github.com/vllm-project/llm-compressor) *Python · ★3,115 · Apache-2.0 · — · score:0.00 · hot:0.65 · rising:0.68 · durable:0.52 · board:rising · trend:up* Transformers-compatible library for applying various compression algorithms to LLMs for optimized deployment with vLLM #### [Piebald-AI/tweakcc](https://github.com/Piebald-AI/tweakcc) *TypeScript · ★1,799 · MIT · — · score:0.00 · hot:0.65 · rising:0.65 · durable:0.51 · board:hot · trend:up* Customize Claude Code's system prompts, create custom toolsets, input pattern highlighters, themes/thinking verbs/spinners, customize input box & user message styling, support AGENTS.md, unlock private/unreleased features, and much more. Supports both native/npm installs on all platforms. #### [llmsresearch/paperbanana](https://github.com/llmsresearch/paperbanana) *Python · ★1,332 · MIT · — · score:0.00 · hot:0.65 · rising:0.63 · durable:0.53 · board:hot · trend:up* Open source implementation and extension of Google Research’s PaperBanana for automated academic figures, diagrams, and research visuals, expanded to new domains like slide generation. #### [vearch/vearch](https://github.com/vearch/vearch) *Go · ★2,296 · Apache-2.0 · — · score:0.00 · hot:0.65 · rising:0.65 · durable:0.50 · board:rising · trend:up* Distributed vector search for AI-native applications #### [vllm-project/guidellm](https://github.com/vllm-project/guidellm) *Python · ★1,020 · Apache-2.0 · — · score:0.00 · hot:0.65 · rising:0.67 · durable:0.49 · board:rising · trend:up* Evaluate and Enhance Your LLM Deployments for Real-World Inference Needs #### [microsoft/vscode-react-native](https://github.com/microsoft/vscode-react-native) *TypeScript · ★2,728 · NOASSERTION · — · score:0.00 · hot:0.65 · rising:0.67 · durable:0.51 · board:rising · trend:up* VSCode extension for React Native - supports debugging and editor integration #### [NVIDIA/ais-k8s](https://github.com/NVIDIA/ais-k8s) *Go · ★132 · MIT · — · score:0.00 · hot:0.65 · rising:0.66 · durable:0.50 · board:rising · trend:up* Kubernetes Operator, Helm Charts, Ansible Playbooks, and utility scripts for large-scale AIStore deployments on Kubernetes. #### [2FastLabs/agent-squad](https://github.com/2FastLabs/agent-squad) *Python · ★7,579 · Apache-2.0 · — · score:0.00 · hot:0.65 · rising:0.67 · durable:0.59 · board:rising · trend:up* Flexible and powerful framework for managing multiple AI agents and handling complex conversations #### [OpenRouterTeam/typescript-sdk](https://github.com/OpenRouterTeam/typescript-sdk) *TypeScript · ★160 · Apache-2.0 · — · score:0.00 · hot:0.65 · rising:0.65 · durable:0.46 · board:rising · trend:up* #### [microsoft/hve-core](https://github.com/microsoft/hve-core) *PowerShell · ★966 · MIT · — · score:0.00 · hot:0.65 · rising:0.67 · durable:0.50 · board:rising · trend:up* A refined collection of Hypervelocity Engineering components (instructions, prompts, agents, and skills) to start your project off right, or upgrade your existing projects to get the most out of all Copilots #### [lucidrains/x-transformers](https://github.com/lucidrains/x-transformers) *Python · ★5,830 · MIT · — · score:0.00 · hot:0.65 · rising:0.67 · durable:0.55 · board:rising · trend:up* A concise but complete full-attention transformer with a set of promising experimental features from various papers #### [langchain-ai/open-swe](https://github.com/langchain-ai/open-swe) *Python · ★9,597 · MIT · — · score:0.00 · hot:0.65 · rising:0.66 · durable:0.56 · board:rising · trend:up* An Open-Source Asynchronous Coding Agent #### [open-webui/desktop](https://github.com/open-webui/desktop) *Svelte · ★1,036 · AGPL-3.0 · — · score:0.00 · hot:0.65 · rising:0.67 · durable:0.55 · board:rising · trend:up* Open WebUI Desktop 🌐 #### [microsoft/UFO](https://github.com/microsoft/UFO) *Python · ★8,484 · MIT · — · score:0.00 · hot:0.65 · rising:0.70 · durable:0.62 · board:rising · trend:up* UFO³: Weaving the Digital Agent Galaxy #### [unbrowse-ai/unbrowse](https://github.com/unbrowse-ai/unbrowse) *TypeScript · ★633 · AGPL-3.0 · — · score:0.00 · hot:0.65 · rising:0.66 · durable:0.50 · board:rising · trend:up* Unbrowse — api native browser skill/cli for any agent. Auto-discovers APIs from browser traffic, generates skills on the fly to call APIs directly 100x faster, 80% cheaper locally. #### [modelcontextprotocol/java-sdk](https://github.com/modelcontextprotocol/java-sdk) *Java · ★3,366 · MIT · — · score:0.00 · hot:0.64 · rising:0.68 · durable:0.53 · board:rising · trend:up* The official Java SDK for Model Context Protocol servers and clients. Maintained in collaboration with Spring AI #### [allenai/vla-evaluation-harness](https://github.com/allenai/vla-evaluation-harness) *Python · ★229 · Apache-2.0 · — · score:0.00 · hot:0.64 · rising:0.66 · durable:0.52 · board:rising · trend:up* One framework to evaluate any VLA model on any robot simulation benchmark. #### [vllm-project/aibrix](https://github.com/vllm-project/aibrix) *Go · ★4,741 · Apache-2.0 · — · score:0.00 · hot:0.64 · rising:0.68 · durable:0.53 · board:rising · trend:up* Cost-efficient and pluggable Infrastructure components for GenAI inference #### [vercel/storage](https://github.com/vercel/storage) *TypeScript · ★591 · Apache-2.0 · — · score:0.00 · hot:0.64 · rising:0.66 · durable:0.48 · board:rising · trend:up* Vercel Storage - Blob and Edge Config #### [PDFMathTranslate/PDFMathTranslate](https://github.com/PDFMathTranslate/PDFMathTranslate) *Python · ★33,189 · AGPL-3.0 · — · score:0.00 · hot:0.64 · rising:0.69 · durable:0.64 · board:rising · trend:up* [EMNLP 2025 Demo] PDF scientific paper translation with preserved formats - 基于 AI 完整保留排版的 PDF 文档全文双语翻译，支持 Google/DeepL/Ollama/OpenAI 等服务，提供 CLI/GUI/MCP/Docker/Zotero #### [facebookresearch/ocean](https://github.com/facebookresearch/ocean) *C++ · ★775 · MIT · — · score:0.00 · hot:0.64 · rising:0.67 · durable:0.52 · board:rising · trend:up* Ocean is the in-house framework for Computer Vision (CV) and Augmented Reality (AR) applications at Meta. It is platform independent and is mainly implemented in C/C++. #### [microsoft/DirectXShaderCompiler](https://github.com/microsoft/DirectXShaderCompiler) *C++ · ★3,547 · NOASSERTION · — · score:0.00 · hot:0.64 · rising:0.67 · durable:0.47 · board:rising · trend:up* This repo hosts the source for the DirectX Shader Compiler which is based on LLVM/Clang. #### [NVIDIA/nvalchemi-toolkit-ops](https://github.com/NVIDIA/nvalchemi-toolkit-ops) *Python · ★169 · NOASSERTION · — · score:0.00 · hot:0.64 · rising:0.64 · durable:0.48 · board:hot · trend:up* ALCHEMI Toolkit-Ops is a collection of optimized batch kernels to accelerate computational chemistry and material science workflows. #### [NVIDIA/recsys-examples](https://github.com/NVIDIA/recsys-examples) *Python · ★250 · NOASSERTION · — · score:0.00 · hot:0.64 · rising:0.63 · durable:0.45 · board:hot · trend:up* Examples for Recommenders - easy to train and deploy on accelerated infrastructure. #### [AIPexStudio/AIPex](https://github.com/AIPexStudio/AIPex) *TypeScript · ★1,147 · MIT · — · score:0.00 · hot:0.64 · rising:0.63 · durable:0.54 · board:hot · trend:stable* AIPex: AI browser automation assistant, no migration and privacy first. Alternative to Manus Browser Operator、 Claude Chrome and Agent Browser #### [galilai-group/stable-pretraining](https://github.com/galilai-group/stable-pretraining) *Python · ★194 · MIT · — · score:0.00 · hot:0.64 · rising:0.61 · durable:0.46 · board:hot · trend:up* Reliable, minimal and scalable library for pretraining foundation and world models #### [microsoft/vscode-typescript-next](https://github.com/microsoft/vscode-typescript-next) *JavaScript · ★295 · MIT · — · score:0.00 · hot:0.64 · rising:0.65 · durable:0.47 · board:rising · trend:up* Enables typescript@next as VS Code's built-in TypeScript version #### [jnMetaCode/agency-agents-zh](https://github.com/jnMetaCode/agency-agents-zh) *Shell · ★6,946 · MIT · — · score:0.00 · hot:0.64 · rising:0.67 · durable:0.55 · board:rising · trend:up* 🎭 211 个即插即用的 AI 专家角色 — 支持 Hermes Agent/Claude Code/Cursor/Copilot 等 16 种工具，覆盖工程/设计/营销/金融等 18 个部门。含 46 个中国市场原创智能体（小红书/抖音/微信/飞书/钉钉等） #### [kyegomez/swarms](https://github.com/kyegomez/swarms) *Python · ★6,264 · Apache-2.0 · — · score:0.00 · hot:0.64 · rising:0.68 · durable:0.57 · board:rising · trend:up* The Enterprise-Grade Production-Ready Multi-Agent Orchestration Framework. Website: https://swarms.ai #### [microsoft/fabric-cicd](https://github.com/microsoft/fabric-cicd) *Python · ★264 · MIT · — · score:0.00 · hot:0.64 · rising:0.66 · durable:0.47 · board:rising · trend:up* Jumpstart CICD deployments in Microsoft Fabric #### [vercel/sdk](https://github.com/vercel/sdk) *TypeScript · ★146 · Apache-2.0 · — · score:0.00 · hot:0.64 · rising:0.64 · durable:0.46 · board:rising · trend:up* Vercel SDK is a type-safe Typescript SDK that gives you access to the Vercel REST API. #### [langchain-ai/langchain-azure](https://github.com/langchain-ai/langchain-azure) *Jupyter Notebook · ★123 · MIT · — · score:0.00 · hot:0.64 · rising:0.65 · durable:0.46 · board:rising · trend:up* Build secure LangChain applications on Azure #### [takahirom/arbigent](https://github.com/takahirom/arbigent) *Kotlin · ★572 · Apache-2.0 · — · score:0.00 · hot:0.64 · rising:0.61 · durable:0.48 · board:hot · trend:up* AI Agent for testing Android, iOS, and Web apps. Get Started in 5 Minutes. Arbigent's intuitive UI and powerful code interface make it accessible to everyone, while its scenario breakdown feature ensures scalability for even the most complex tasks. #### [vllm-project/compressed-tensors](https://github.com/vllm-project/compressed-tensors) *Python · ★274 · Apache-2.0 · — · score:0.00 · hot:0.64 · rising:0.65 · durable:0.47 · board:rising · trend:up* A safetensors extension to efficiently store sparse quantized tensors on disk #### [AccumulateMore/CV](https://github.com/AccumulateMore/CV) *Jupyter Notebook · ★20,036 · no-license · — · score:0.00 · hot:0.64 · rising:0.65 · durable:0.53 · board:rising · trend:up* ✔（已完结）超级全面的 深度学习 笔记【土堆 Pytorch】【李沐 动手学深度学习】【吴恩达 深度学习】【大飞 大模型Agent】 #### [Cinnamon/kotaemon](https://github.com/Cinnamon/kotaemon) *Python · ★25,290 · Apache-2.0 · — · score:0.00 · hot:0.64 · rising:0.66 · durable:0.62 · board:rising · trend:up* An open-source RAG-based tool for chatting with your documents. #### [microsoft/skills-for-copilot-studio](https://github.com/microsoft/skills-for-copilot-studio) *JavaScript · ★160 · MIT · — · score:0.00 · hot:0.64 · rising:0.65 · durable:0.47 · board:rising · trend:up* A skill for AI-coding tools to build and edit Microsoft Copilot Studio agents as YAML — with schema validation, templates, and AI-powered skills. Suited for Claude Code, GitHub Copilot CLI, and more. #### [anthropics/anthropic-sdk-php](https://github.com/anthropics/anthropic-sdk-php) *PHP · ★138 · MIT · — · score:0.00 · hot:0.64 · rising:0.65 · durable:0.49 · board:rising · trend:up* Access to Anthropic's safety-first language model APIs in PHP #### [microsoft/mssql-python](https://github.com/microsoft/mssql-python) *Python · ★413 · NOASSERTION · — · score:0.00 · hot:0.64 · rising:0.64 · durable:0.45 · board:hot · trend:up* Microsoft Python Driver for SQL Server #### [microsoft/secureboot_objects](https://github.com/microsoft/secureboot_objects) *Python · ★213 · NOASSERTION · — · score:0.00 · hot:0.64 · rising:0.65 · durable:0.47 · board:rising · trend:up* Secure boot objects recommended by Microsoft. #### [vllm-project/speculators](https://github.com/vllm-project/speculators) *Python · ★352 · Apache-2.0 · — · score:0.00 · hot:0.64 · rising:0.65 · durable:0.47 · board:rising · trend:up* A unified library for building, evaluating, and storing speculative decoding algorithms for LLM inference in vLLM #### [YouMind-OpenLab/awesome-nano-banana-pro-prompts](https://github.com/YouMind-OpenLab/awesome-nano-banana-pro-prompts) *TypeScript · ★11,257 · NOASSERTION · — · score:0.00 · hot:0.63 · rising:0.66 · durable:0.53 · board:rising · trend:up* 🍌 World's largest Nano Banana Pro prompt library — 10,000+ curated prompts with preview images, 16 languages. Google Gemini AI image generation. Free & open source. #### [microsoft/EventLogExpert](https://github.com/microsoft/EventLogExpert) *C# · ★265 · MIT · — · score:0.00 · hot:0.63 · rising:0.65 · durable:0.51 · board:rising · trend:up* #### [zilliztech/claude-context](https://github.com/zilliztech/claude-context) *TypeScript · ★6,012 · MIT · — · score:0.00 · hot:0.63 · rising:0.61 · durable:0.52 · board:hot · trend:up* Code search MCP for Claude Code. Make entire codebase the context for any coding agent. #### [hpcaitech/ColossalAI](https://github.com/hpcaitech/ColossalAI) *Python · ★41,375 · Apache-2.0 · — · score:0.00 · hot:0.63 · rising:0.69 · durable:0.61 · board:rising · trend:up* Making large AI models cheaper, faster and more accessible #### [dingodb/dingo](https://github.com/dingodb/dingo) *Java · ★1,700 · Apache-2.0 · — · score:0.00 · hot:0.63 · rising:0.67 · durable:0.56 · board:rising · trend:up* A multi-modal vector database that supports upserts and vector queries using unified SQL (MySQL-Compatible) on structured and unstructured data, while meeting the requirements of high concurrency and ultra-low latency. #### [coasty-ai/open-computer-use](https://github.com/coasty-ai/open-computer-use) *TypeScript · ★495 · Apache-2.0 · — · score:0.00 · hot:0.63 · rising:0.62 · durable:0.57 · board:hot · trend:stable* State of the Art 82% OSWorld Verified Computer Using Agent, fully open-source, safe, auditable, and production-ready. #### [anthropics/connect-rust](https://github.com/anthropics/connect-rust) *Rust · ★249 · Apache-2.0 · — · score:0.00 · hot:0.63 · rising:0.64 · durable:0.48 · board:rising · trend:up* An implementation of the ConnectRPC protocol for Rust #### [modelcontextprotocol/modelcontextprotocol](https://github.com/modelcontextprotocol/modelcontextprotocol) *TypeScript · ★7,864 · NOASSERTION · — · score:0.00 · hot:0.63 · rising:0.68 · durable:0.53 · board:rising · trend:up* Specification and documentation for the Model Context Protocol #### [google-deepmind/onetwo](https://github.com/google-deepmind/onetwo) *Python · ★265 · Apache-2.0 · — · score:0.00 · hot:0.63 · rising:0.65 · durable:0.51 · board:rising · trend:up* #### [vllm-project/production-stack](https://github.com/vllm-project/production-stack) *Python · ★2,282 · Apache-2.0 · — · score:0.00 · hot:0.63 · rising:0.66 · durable:0.52 · board:rising · trend:up* vLLM’s reference system for K8S-native cluster-wide deployment with community-driven performance optimization #### [jxnl/instructor-go](https://github.com/jxnl/instructor-go) *Go · ★201 · MIT · — · score:0.00 · hot:0.63 · rising:0.63 · durable:0.49 · board:rising · trend:stable* structured outputs for llms #### [vercel/terraform-provider-vercel](https://github.com/vercel/terraform-provider-vercel) *Go · ★185 · MPL-2.0 · — · score:0.00 · hot:0.63 · rising:0.65 · durable:0.45 · board:rising · trend:up* Terraform Vercel Provider #### [google-gemini/genai-processors](https://github.com/google-gemini/genai-processors) *Python · ★2,108 · Apache-2.0 · — · score:0.00 · hot:0.63 · rising:0.63 · durable:0.58 · board:rising · trend:stable* GenAI Processors is a lightweight Python library that enables efficient, parallel content processing. #### [simonw/llm-anthropic](https://github.com/simonw/llm-anthropic) *Python · ★216 · Apache-2.0 · — · score:0.00 · hot:0.63 · rising:0.63 · durable:0.44 · board:hot · trend:up* LLM access to models by Anthropic, including the Claude series #### [zebbern/claude-code-guide](https://github.com/zebbern/claude-code-guide) *? · ★3,955 · MIT · — · score:0.00 · hot:0.63 · rising:0.64 · durable:0.53 · board:rising · trend:up* Claude Code Guide - Setup, Commands, workflows, agents, skills & tips-n-tricks go from beginner to power user! #### [vercel/chatbot](https://github.com/vercel/chatbot) *TypeScript · ★20,143 · NOASSERTION · — · score:0.00 · hot:0.63 · rising:0.68 · durable:0.54 · board:rising · trend:up* A full-featured, hackable Next.js AI chatbot built by Vercel #### [microsoft/semantic-link-labs](https://github.com/microsoft/semantic-link-labs) *Python · ★500 · MIT · — · score:0.00 · hot:0.63 · rising:0.65 · durable:0.48 · board:rising · trend:up* Early access to new features for Microsoft Fabric's Semantic Link. #### [open-webui/helm-charts](https://github.com/open-webui/helm-charts) *Go Template · ★277 · no-license · — · score:0.00 · hot:0.63 · rising:0.64 · durable:0.50 · board:rising · trend:up* #### [google-gemini/cookbook](https://github.com/google-gemini/cookbook) *Jupyter Notebook · ★17,043 · Apache-2.0 · — · score:0.00 · hot:0.62 · rising:0.67 · durable:0.54 · board:rising · trend:up* Examples and guides for using the Gemini API #### [vercel/nft](https://github.com/vercel/nft) *JavaScript · ★1,612 · MIT · — · score:0.00 · hot:0.62 · rising:0.65 · durable:0.48 · board:rising · trend:up* Node.js dependency tracing utility #### [weaviate/Verba](https://github.com/weaviate/Verba) *Python · ★7,652 · BSD-3-Clause · — · score:0.00 · hot:0.62 · rising:0.69 · durable:0.59 · board:rising · trend:up* Retrieval Augmented Generation (RAG) chatbot powered by Weaviate #### [microsoft/DefinitelyTyped-tools](https://github.com/microsoft/DefinitelyTyped-tools) *TypeScript · ★412 · MIT · — · score:0.00 · hot:0.62 · rising:0.64 · durable:0.46 · board:rising · trend:up* Infrastructure for DefinitelyTyped #### [NVIDIA/numba-cuda](https://github.com/NVIDIA/numba-cuda) *Python · ★271 · BSD-2-Clause · — · score:0.00 · hot:0.62 · rising:0.64 · durable:0.47 · board:rising · trend:up* The CUDA target for Numba #### [karpathy/nanochat](https://github.com/karpathy/nanochat) *Python · ★52,148 · MIT · — · score:0.00 · hot:0.62 · rising:0.69 · durable:0.58 · board:rising · trend:up* The best ChatGPT that $100 can buy. #### [microsoft/Power-CAT-Copilot-Studio-Kit](https://github.com/microsoft/Power-CAT-Copilot-Studio-Kit) *C# · ★361 · MIT · — · score:0.00 · hot:0.62 · rising:0.64 · durable:0.47 · board:rising · trend:up* #### [qdrant/qdrant-web-ui](https://github.com/qdrant/qdrant-web-ui) *JavaScript · ★389 · Apache-2.0 · — · score:0.00 · hot:0.62 · rising:0.63 · durable:0.46 · board:rising · trend:up* Self-hosted web UI for Qdrant #### [LouisShark/chatgpt_system_prompt](https://github.com/LouisShark/chatgpt_system_prompt) *HTML · ★10,517 · MIT · — · score:0.00 · hot:0.62 · rising:0.67 · durable:0.53 · board:rising · trend:up* A collection of GPT system prompts and various prompt injection/leaking knowledge. #### [AI4Finance-Foundation/FinRobot](https://github.com/AI4Finance-Foundation/FinRobot) *Jupyter Notebook · ★6,730 · Apache-2.0 · — · score:0.00 · hot:0.62 · rising:0.64 · durable:0.59 · board:rising · trend:up* FinRobot: An Open-Source AI Agent Platform for Financial Analysis using LLMs 🚀 🚀 🚀 #### [VectifyAI/PageIndex](https://github.com/VectifyAI/PageIndex) *Python · ★25,493 · MIT · — · score:0.00 · hot:0.62 · rising:0.63 · durable:0.58 · board:rising · trend:up* 📑 PageIndex: Document Index for Vectorless, Reasoning-based RAG #### [lsdefine/GenericAgent](https://github.com/lsdefine/GenericAgent) *Python · ★4,451 · MIT · — · score:0.00 · hot:0.62 · rising:0.63 · durable:0.52 · board:rising · trend:up* Self-evolving agent: grows skill tree from 3.3K-line seed, achieving full system control with 6x less token consumption #### [pydantic/pydantic-ai-harness](https://github.com/pydantic/pydantic-ai-harness) *Python · ★134 · MIT · — · score:0.00 · hot:0.62 · rising:0.62 · durable:0.45 · board:hot · trend:up* Batteries for your Pydantic AI agent. #### [milvus-io/milvus-sdk-java](https://github.com/milvus-io/milvus-sdk-java) *Java · ★484 · Apache-2.0 · — · score:0.00 · hot:0.62 · rising:0.63 · durable:0.47 · board:rising · trend:up* Java SDK for Milvus. #### [google-gemini/gemini-fullstack-langgraph-quickstart](https://github.com/google-gemini/gemini-fullstack-langgraph-quickstart) *Jupyter Notebook · ★18,108 · Apache-2.0 · — · score:0.00 · hot:0.62 · rising:0.67 · durable:0.55 · board:rising · trend:up* Get started with building Fullstack Agents using Gemini 2.5 and LangGraph #### [dair-ai/AI-Papers-of-the-Week](https://github.com/dair-ai/AI-Papers-of-the-Week) *? · ★12,307 · no-license · — · score:0.00 · hot:0.62 · rising:0.67 · durable:0.51 · board:rising · trend:up* 🔥Highlighting the top ML papers every week. #### [superagent-ai/superagent](https://github.com/superagent-ai/superagent) *TypeScript · ★6,542 · MIT · — · score:0.00 · hot:0.62 · rising:0.65 · durable:0.61 · board:rising · trend:stable* Superagent protects your AI applications against prompt injections, data leaks, and harmful outputs. Embed safety directly into your app and prove compliance to your customers. #### [microsoft/FLAML](https://github.com/microsoft/FLAML) *Jupyter Notebook · ★4,334 · MIT · — · score:0.00 · hot:0.62 · rising:0.66 · durable:0.51 · board:rising · trend:up* A fast library for AutoML and tuning. Join our Discord: https://discord.gg/Cppx2vSPVP. #### [zilliztech/attu](https://github.com/zilliztech/attu) *Shell · ★2,812 · NOASSERTION · — · score:0.00 · hot:0.62 · rising:0.61 · durable:0.48 · board:hot · trend:up* The Best GUI for Milvus #### [xlang-ai/OSWorld](https://github.com/xlang-ai/OSWorld) *Python · ★2,791 · Apache-2.0 · — · score:0.00 · hot:0.62 · rising:0.63 · durable:0.52 · board:rising · trend:up* [NeurIPS 2024] OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments #### [InternLM/xtuner](https://github.com/InternLM/xtuner) *Python · ★5,121 · Apache-2.0 · — · score:0.00 · hot:0.62 · rising:0.63 · durable:0.53 · board:rising · trend:up* A Next-Generation Training Engine Built for Ultra-Large MoE Models #### [sgl-project/genai-bench](https://github.com/sgl-project/genai-bench) *Python · ★292 · MIT · — · score:0.00 · hot:0.62 · rising:0.63 · durable:0.47 · board:rising · trend:up* Genai-bench is a powerful benchmark tool designed for comprehensive token-level performance evaluation of large language model (LLM) serving systems. #### [alibaba/spring-ai-alibaba](https://github.com/alibaba/spring-ai-alibaba) *Java · ★9,310 · Apache-2.0 · — · score:0.00 · hot:0.62 · rising:0.65 · durable:0.55 · board:rising · trend:up* Agentic AI Framework for Java Developers #### [ashbuilds/payload-ai](https://github.com/ashbuilds/payload-ai) *TypeScript · ★482 · NOASSERTION · — · score:0.00 · hot:0.62 · rising:0.62 · durable:0.50 · board:rising · trend:stable* AI Plugin is a powerful extension for the Payload CMS, integrating advanced AI capabilities to enhance content creation and management. #### [NVIDIA/k8s-nim-operator](https://github.com/NVIDIA/k8s-nim-operator) *Go · ★154 · Apache-2.0 · — · score:0.00 · hot:0.62 · rising:0.63 · durable:0.46 · board:rising · trend:up* An Operator for deployment and maintenance of NVIDIA NIMs and NeMo microservices in a Kubernetes environment. #### [huggingface/hf-hub](https://github.com/huggingface/hf-hub) *Rust · ★287 · Apache-2.0 · — · score:0.00 · hot:0.62 · rising:0.65 · durable:0.48 · board:rising · trend:up* Rust client for the huggingface hub aiming for minimal subset of features over `huggingface-hub` python package #### [rasbt/reasoning-from-scratch](https://github.com/rasbt/reasoning-from-scratch) *Jupyter Notebook · ★4,145 · Apache-2.0 · — · score:0.00 · hot:0.62 · rising:0.64 · durable:0.53 · board:rising · trend:up* Implement a reasoning LLM in PyTorch from scratch, step by step #### [NVIDIA/OSMO](https://github.com/NVIDIA/OSMO) *TypeScript · ★143 · Apache-2.0 · — · score:0.00 · hot:0.62 · rising:0.62 · durable:0.46 · board:rising · trend:up* The developer-first platform for scaling complex Physical AI workloads across heterogeneous compute—unifying training GPUs, simulation clusters, and edge devices in a simple YAML #### [WeThinkIn/AIGC-Interview-Book](https://github.com/WeThinkIn/AIGC-Interview-Book) *? · ★3,540 · GPL-3.0 · — · score:0.00 · hot:0.61 · rising:0.63 · durable:0.52 · board:rising · trend:up* 【三年面试五年模拟】AIGC算法工程师面试秘籍。涵盖AIGC、LLM大模型、AI Agent、传统深度学习、自动驾驶、机器学习、计算机视觉、自然语言处理、强化学习、大数据挖掘、具身智能、元宇宙、AGI等AI行业面试笔试干货经验与核心知识。 #### [microsoft/promptflow](https://github.com/microsoft/promptflow) *Python · ★11,106 · MIT · — · score:0.00 · hot:0.61 · rising:0.66 · durable:0.59 · board:rising · trend:up* Build high-quality LLM apps - from prototyping, testing to production deployment and monitoring. #### [NirDiamant/Prompt_Engineering](https://github.com/NirDiamant/Prompt_Engineering) *Jupyter Notebook · ★7,412 · NOASSERTION · — · score:0.00 · hot:0.61 · rising:0.63 · durable:0.52 · board:rising · trend:up* 22 prompt engineering techniques with hands-on Jupyter Notebook tutorials, from fundamental concepts to advanced strategies for leveraging LLMs. #### [nextai-translator/nextai-translator](https://github.com/nextai-translator/nextai-translator) *TypeScript · ★24,895 · AGPL-3.0 · — · score:0.00 · hot:0.61 · rising:0.64 · durable:0.57 · board:rising · trend:up* 基于 ChatGPT API 的划词翻译浏览器插件和跨平台桌面端应用 - Browser extension and cross-platform desktop application for translation based on ChatGPT API. #### [microsoft/WPF-Samples](https://github.com/microsoft/WPF-Samples) *C# · ★5,638 · MIT · — · score:0.00 · hot:0.61 · rising:0.68 · durable:0.55 · board:rising · trend:up* Repository for WPF related samples #### [pydantic/pydantic-settings](https://github.com/pydantic/pydantic-settings) *Python · ★1,313 · MIT · — · score:0.00 · hot:0.61 · rising:0.64 · durable:0.48 · board:rising · trend:up* Settings management using pydantic #### [huggingface/optimum-neuron](https://github.com/huggingface/optimum-neuron) *Jupyter Notebook · ★265 · Apache-2.0 · — · score:0.00 · hot:0.61 · rising:0.65 · durable:0.53 · board:rising · trend:stable* Training and inference on AWS Trainium and Inferentia chips. #### [microsoft/PowerApps-Samples](https://github.com/microsoft/PowerApps-Samples) *C# · ★1,913 · MIT · — · score:0.00 · hot:0.61 · rising:0.68 · durable:0.53 · board:rising · trend:up* Sample code for Power Apps, including Dataverse, model-driven apps, canvas apps, Power Apps component framework, portals, and AI Builder. #### [microsoft/vscode-codicons](https://github.com/microsoft/vscode-codicons) *Handlebars · ★1,107 · CC-BY-4.0 · — · score:0.00 · hot:0.61 · rising:0.64 · durable:0.49 · board:rising · trend:up* The icon font for Visual Studio Code #### [pydantic/genai-prices](https://github.com/pydantic/genai-prices) *Python · ★282 · MIT · — · score:0.00 · hot:0.61 · rising:0.63 · durable:0.47 · board:rising · trend:up* Calculate prices for calling LLM inference APIs. #### [NVIDIA/mig-parted](https://github.com/NVIDIA/mig-parted) *Go · ★247 · Apache-2.0 · — · score:0.00 · hot:0.61 · rising:0.63 · durable:0.46 · board:rising · trend:up* MIG Partition Editor for NVIDIA GPUs #### [qdrant/qdrant-helm](https://github.com/qdrant/qdrant-helm) *Go · ★174 · Apache-2.0 · — · score:0.00 · hot:0.61 · rising:0.63 · durable:0.45 · board:rising · trend:up* #### [promptslab/Awesome-Prompt-Engineering](https://github.com/promptslab/Awesome-Prompt-Engineering) *TypeScript · ★5,775 · Apache-2.0 · — · score:0.00 · hot:0.61 · rising:0.65 · durable:0.50 · board:rising · trend:up* This repository contains a hand-curated resources for Prompt Engineering with a focus on Generative Pre-trained Transformer (GPT), ChatGPT, PaLM etc #### [facebookresearch/EgoBlur](https://github.com/facebookresearch/EgoBlur) *Python · ★213 · Apache-2.0 · — · score:0.00 · hot:0.61 · rising:0.63 · durable:0.48 · board:rising · trend:stable* This repository contains a command-line interface(CLI) that can detect and blur out faces and license plates(PII) from images and videos. The CLI takes an image or video file as input, runs an anonymization algorithm on it, and writes the blurred output to a specified path. #### [NVIDIA/nvImageCodec](https://github.com/NVIDIA/nvImageCodec) *Jupyter Notebook · ★145 · Apache-2.0 · — · score:0.00 · hot:0.61 · rising:0.60 · durable:0.46 · board:hot · trend:stable* A nvImageCodec library of GPU- and CPU- accelerated codecs featuring a unified interface #### [vercel/v0-sdk](https://github.com/vercel/v0-sdk) *TypeScript · ★468 · NOASSERTION · — · score:0.00 · hot:0.61 · rising:0.62 · durable:0.46 · board:rising · trend:up* SDK for the v0 Platform API #### [ykdojo/claude-code-tips](https://github.com/ykdojo/claude-code-tips) *JavaScript · ★7,811 · NOASSERTION · — · score:0.00 · hot:0.61 · rising:0.63 · durable:0.58 · board:rising · trend:stable* 45 tips for getting the most out of Claude Code, from basics to advanced - includes a custom status line script, cutting the system prompt in half, using Gemini CLI as Claude Code's minion, and Claude Code running itself in a container. Also includes the dx plugin. #### [chatanywhere/GPT_API_free](https://github.com/chatanywhere/GPT_API_free) *Python · ★37,404 · MIT · — · score:0.00 · hot:0.61 · rising:0.65 · durable:0.57 · board:rising · trend:up* Free ChatGPT&DeepSeek API Key，免费ChatGPT&DeepSeek API。免费接入DeepSeek API和GPT4 API，支持 gpt | deepseek | claude | gemini | grok 等排名靠前的常用大模型。 #### [ZeroLu/awesome-nanobanana-pro](https://github.com/ZeroLu/awesome-nanobanana-pro) *? · ★9,775 · MIT · — · score:0.00 · hot:0.61 · rising:0.65 · durable:0.55 · board:rising · trend:up* 🚀 An awesome list of curated Nano Banana pro prompts and examples. Your go-to resource for mastering prompt engineering and exploring the creative potential of the Nano banana pro(Nano banana 2) AI image model. #### [langchain-ai/langchain-nvidia](https://github.com/langchain-ai/langchain-nvidia) *Jupyter Notebook · ★198 · MIT · — · score:0.00 · hot:0.61 · rising:0.63 · durable:0.48 · board:rising · trend:up* #### [microsoft/sql-server-samples](https://github.com/microsoft/sql-server-samples) *? · ★11,012 · NOASSERTION · — · score:0.00 · hot:0.61 · rising:0.67 · durable:0.52 · board:rising · trend:up* Azure Data SQL Samples - Official Microsoft GitHub Repository containing code samples for SQL Server, Azure SQL, Azure Synapse, and Azure SQL Edge #### [run-llama/llama_deploy](https://github.com/run-llama/llama_deploy) *Python · ★2,072 · MIT · — · score:0.00 · hot:0.61 · rising:0.62 · durable:0.55 · board:rising · trend:stable* Deploy your agentic worfklows to production #### [RediSearch/RediSearch](https://github.com/RediSearch/RediSearch) *C · ★6,121 · NOASSERTION · — · score:0.00 · hot:0.61 · rising:0.63 · durable:0.46 · board:rising · trend:up* A query and indexing engine for Redis, providing secondary indexing, full-text search, vector similarity search and aggregations. #### [qdrant/qdrant-client](https://github.com/qdrant/qdrant-client) *Python · ★1,267 · Apache-2.0 · — · score:0.00 · hot:0.61 · rising:0.62 · durable:0.49 · board:rising · trend:up* Python client for Qdrant vector search engine #### [plastic-labs/honcho](https://github.com/plastic-labs/honcho) *Python · ★2,611 · AGPL-3.0 · — · score:0.00 · hot:0.61 · rising:0.57 · durable:0.47 · board:hot · trend:up* Memory library for building stateful agents #### [hao-ai-lab/FastVideo](https://github.com/hao-ai-lab/FastVideo) *Python · ★3,402 · Apache-2.0 · — · score:0.00 · hot:0.61 · rising:0.64 · durable:0.51 · board:rising · trend:up* A unified inference and post-training framework for accelerated video generation. #### [microsoft/dcp](https://github.com/microsoft/dcp) *Go · ★155 · MIT · — · score:0.00 · hot:0.61 · rising:0.61 · durable:0.45 · board:hot · trend:stable* Developer Control Plane API server and CLI. #### [kk43994/kkclaw](https://github.com/kk43994/kkclaw) *JavaScript · ★155 · NOASSERTION · — · score:0.00 · hot:0.61 · rising:0.60 · durable:0.46 · board:hot · trend:stable* 🦞 一个可爱的桌面龙虾AI助手 - Desktop lobster pet with OpenClaw AI, Edge TTS voice, and emotion animations #### [vercel/examples](https://github.com/vercel/examples) *TypeScript · ★5,032 · MIT · — · score:0.00 · hot:0.61 · rising:0.67 · durable:0.53 · board:rising · trend:up* Enjoy our curated collection of examples and solutions. Use these patterns to build your own robust and scalable applications. #### [Helicone/helicone](https://github.com/Helicone/helicone) *TypeScript · ★5,516 · Apache-2.0 · — · score:0.00 · hot:0.61 · rising:0.62 · durable:0.57 · board:rising · trend:stable* 🧊 Open source LLM observability platform. One line of code to monitor, evaluate, and experiment. YC W23 🍓 #### [NVIDIA/aicr](https://github.com/NVIDIA/aicr) *Go · ★268 · Apache-2.0 · — · score:0.00 · hot:0.61 · rising:0.61 · durable:0.46 · board:rising · trend:stable* Tooling for optimized, validated, and reproducible GPU-accelerated AI runtime in Kubernetes #### [ai-boost/awesome-prompts](https://github.com/ai-boost/awesome-prompts) *? · ★7,650 · GPL-3.0 · — · score:0.00 · hot:0.61 · rising:0.65 · durable:0.52 · board:rising · trend:up* Curated list of chatgpt prompts from the top-rated GPTs in the GPTs Store. Prompt Engineering, prompt attack & prompt protect. Advanced Prompt Engineering papers. #### [microsoft/Generative-AI-for-beginners-dotnet](https://github.com/microsoft/Generative-AI-for-beginners-dotnet) *C# · ★2,648 · MIT · — · score:0.00 · hot:0.61 · rising:0.67 · durable:0.60 · board:rising · trend:up* Five lessons, learn how to really apply AI to your .NET Applications #### [langchain-ai/chat-langchain](https://github.com/langchain-ai/chat-langchain) *Python · ★6,305 · MIT · — · score:0.00 · hot:0.61 · rising:0.66 · durable:0.53 · board:rising · trend:up* #### [Aider-AI/aider](https://github.com/Aider-AI/aider) *Python · ★43,570 · Apache-2.0 · — · score:0.00 · hot:0.61 · rising:0.64 · durable:0.59 · board:rising · trend:up* aider is AI pair programming in your terminal #### [starpig1129/DATAGEN](https://github.com/starpig1129/DATAGEN) *Python · ★1,689 · MIT · — · score:0.00 · hot:0.61 · rising:0.62 · durable:0.50 · board:rising · trend:up* DATAGEN: AI-driven multi-agent research assistant automating hypothesis generation, data analysis, and report writing. #### [mufeedvh/code2prompt](https://github.com/mufeedvh/code2prompt) *Rust · ★7,291 · MIT · — · score:0.00 · hot:0.60 · rising:0.65 · durable:0.59 · board:rising · trend:stable* A CLI tool to convert your codebase into a single LLM prompt with source tree, prompt templating, and token counting. #### [modelcontextprotocol/php-sdk](https://github.com/modelcontextprotocol/php-sdk) *PHP · ★1,455 · NOASSERTION · — · score:0.00 · hot:0.60 · rising:0.62 · durable:0.49 · board:rising · trend:up* The official PHP SDK for Model Context Protocol servers and clients. Maintained in collaboration with The PHP Foundation. #### [vercel/next-app-router-playground](https://github.com/vercel/next-app-router-playground) *TypeScript · ★2,972 · MIT · — · score:0.00 · hot:0.60 · rising:0.66 · durable:0.51 · board:rising · trend:up* A playground to explore Next.js features such as nested layouts, instant loading states, streaming, and component level data fetching. #### [sgl-project/rbg](https://github.com/sgl-project/rbg) *Go · ★206 · Apache-2.0 · — · score:0.00 · hot:0.60 · rising:0.60 · durable:0.47 · board:rising · trend:stable* A workload for deploying LLM inference services on Kubernetes #### [NVIDIA/IsaacTeleop](https://github.com/NVIDIA/IsaacTeleop) *Python · ★149 · Apache-2.0 · — · score:0.00 · hot:0.60 · rising:0.61 · durable:0.45 · board:rising · trend:stable* The unified framework for sim & real robot teleoperation #### [explosion/spaCy](https://github.com/explosion/spaCy) *Python · ★33,486 · MIT · — · score:0.00 · hot:0.60 · rising:0.66 · durable:0.58 · board:rising · trend:up* 💫 Industrial-strength Natural Language Processing (NLP) in Python #### [Hexastack/Hexabot](https://github.com/Hexastack/Hexabot) *TypeScript · ★940 · AGPL-3.0 · — · score:0.00 · hot:0.60 · rising:0.60 · durable:0.50 · board:rising · trend:stable* Hexabot is an open-source AI chatbot / agent builder. It allows you to create and manage multi-channel and multilingual chatbots / agents with ease. #### [vercel/hyper](https://github.com/vercel/hyper) *TypeScript · ★44,612 · MIT · — · score:0.00 · hot:0.60 · rising:0.67 · durable:0.56 · board:rising · trend:up* A terminal built on web technologies #### [golemcloud/golem](https://github.com/golemcloud/golem) *Rust · ★987 · NOASSERTION · — · score:0.00 · hot:0.60 · rising:0.60 · durable:0.46 · board:rising · trend:up* Golem Cloud is the agent-native platform for building AI agents and distributed applications that never lose state, never duplicate work, and never require you to build infrastructure. #### [Picovoice/picollm](https://github.com/Picovoice/picollm) *Python · ★312 · Apache-2.0 · — · score:0.00 · hot:0.60 · rising:0.61 · durable:0.53 · board:rising · trend:stable* On-device LLM Inference Powered by X-Bit Quantization #### [aimhubio/aim](https://github.com/aimhubio/aim) *Python · ★6,096 · Apache-2.0 · — · score:0.00 · hot:0.60 · rising:0.62 · durable:0.50 · board:rising · trend:up* Aim 💫 — An easy-to-use & supercharged open-source experiment tracker. #### [run-llama/llama_cloud_services](https://github.com/run-llama/llama_cloud_services) *TypeScript · ★4,246 · MIT · — · score:0.00 · hot:0.59 · rising:0.63 · durable:0.52 · board:rising · trend:up* Knowledge Agents and Management in the Cloud #### [X-PLUG/MobileAgent](https://github.com/X-PLUG/MobileAgent) *Python · ★8,514 · MIT · — · score:0.00 · hot:0.59 · rising:0.61 · durable:0.50 · board:rising · trend:up* Mobile-Agent: The Powerful GUI Agent Family #### [tmgthb/Autonomous-Agents](https://github.com/tmgthb/Autonomous-Agents) *? · ★1,229 · MIT · — · score:0.00 · hot:0.59 · rising:0.61 · durable:0.50 · board:rising · trend:stable* Autonomous Agents (LLMs) research papers. Updated Daily. #### [sgl-project/sgl-kernel-npu](https://github.com/sgl-project/sgl-kernel-npu) *C++ · ★125 · MIT · — · score:0.00 · hot:0.59 · rising:0.61 · durable:0.46 · board:rising · trend:stable* SGLang kernel library for NPU #### [Xnhyacinth/Awesome-LLM-Long-Context-Modeling](https://github.com/Xnhyacinth/Awesome-LLM-Long-Context-Modeling) *? · ★1,968 · MIT · — · score:0.00 · hot:0.59 · rising:0.57 · durable:0.52 · board:hot · trend:stable* 📰 Must-read papers and blogs on LLM based Long Context Modeling 🔥 #### [pingcap/autoflow](https://github.com/pingcap/autoflow) *TypeScript · ★2,764 · Apache-2.0 · — · score:0.00 · hot:0.59 · rising:0.61 · durable:0.52 · board:rising · trend:stable* pingcap/autoflow is a Graph RAG based and conversational knowledge base tool built with TiDB Serverless Vector Storage. Demo: https://tidb.ai #### [langchain-ai/agentevals](https://github.com/langchain-ai/agentevals) *Python · ★551 · MIT · — · score:0.00 · hot:0.59 · rising:0.61 · durable:0.48 · board:rising · trend:stable* Readymade evaluators for agent trajectories #### [simonw/datasette](https://github.com/simonw/datasette) *Python · ★10,969 · Apache-2.0 · — · score:0.00 · hot:0.59 · rising:0.64 · durable:0.50 · board:rising · trend:up* An open source multi-tool for exploring and publishing data #### [jina-ai/reader](https://github.com/jina-ai/reader) *TypeScript · ★10,603 · Apache-2.0 · — · score:0.00 · hot:0.59 · rising:0.63 · durable:0.52 · board:rising · trend:up* Convert any URL to an LLM-friendly input with a simple prefix https://r.jina.ai/ #### [strands-agents/samples](https://github.com/strands-agents/samples) *Python · ★730 · Apache-2.0 · — · score:0.00 · hot:0.59 · rising:0.57 · durable:0.45 · board:hot · trend:stable* Agent samples built using the Strands Agents SDK. #### [vllm-project/tpu-inference](https://github.com/vllm-project/tpu-inference) *Python · ★292 · Apache-2.0 · — · score:0.00 · hot:0.59 · rising:0.63 · durable:0.48 · board:rising · trend:up* TPU inference for vLLM, with unified JAX and PyTorch support. #### [roboflow/maestro](https://github.com/roboflow/maestro) *Python · ★2,668 · Apache-2.0 · — · score:0.00 · hot:0.59 · rising:0.61 · durable:0.56 · board:rising · trend:stable* streamline the fine-tuning process for multimodal models: PaliGemma 2, Florence-2, and Qwen2.5-VL #### [testdriverai/testdriverai](https://github.com/testdriverai/testdriverai) *JavaScript · ★219 · no-license · — · score:0.00 · hot:0.59 · rising:0.57 · durable:0.42 · board:hot · trend:stable* Computer-Use SDK for E2E QA Testing #### [microsoft/sample-app-aoai-chatGPT](https://github.com/microsoft/sample-app-aoai-chatGPT) *Python · ★1,918 · MIT · — · score:0.00 · hot:0.59 · rising:0.64 · durable:0.50 · board:rising · trend:up* Sample code for a simple web chat experience through Azure OpenAI, including Azure OpenAI On Your Data. #### [langchain-ai/langchain-skills](https://github.com/langchain-ai/langchain-skills) *Shell · ★598 · no-license · — · score:0.00 · hot:0.59 · rising:0.60 · durable:0.52 · board:rising · trend:stable* #### [microsoft/fast](https://github.com/microsoft/fast) *TypeScript · ★9,641 · NOASSERTION · — · score:0.00 · hot:0.59 · rising:0.62 · durable:0.47 · board:rising · trend:up* The adaptive interface system for modern web experiences. #### [facebookresearch/tensor-layouts](https://github.com/facebookresearch/tensor-layouts) *Python · ★167 · MIT · — · score:0.00 · hot:0.59 · rising:0.60 · durable:0.51 · board:rising · trend:stable* A pure-Python implementation of the Nvidia CuTe layout algebra intended to be approachable and easy to learn. #### [SamurAIGPT/Generative-Media-Skills](https://github.com/SamurAIGPT/Generative-Media-Skills) *Shell · ★3,040 · MIT · — · score:0.00 · hot:0.59 · rising:0.60 · durable:0.52 · board:rising · trend:stable* Multi-modal Generative Media Skills for AI Agents (Claude Code, Cursor, Gemini CLI). High-quality image, video, and audio generation powered by muapi.ai. #### [EleutherAI/delphi](https://github.com/EleutherAI/delphi) *Python · ★253 · Apache-2.0 · — · score:0.00 · hot:0.59 · rising:0.60 · durable:0.49 · board:rising · trend:stable* Delphi was the home of a temple to Phoebus Apollo, which famously had the inscription, 'Know Thyself.' This library lets language models know themselves through automated interpretability. #### [kaiban-ai/KaibanJS](https://github.com/kaiban-ai/KaibanJS) *TypeScript · ★1,414 · MIT · — · score:0.00 · hot:0.59 · rising:0.60 · durable:0.50 · board:rising · trend:stable* KaibanJS is a JavaScript-native framework for building and managing multi-agent systems with a Kanban-inspired approach. #### [pydantic/jiter](https://github.com/pydantic/jiter) *Rust · ★523 · MIT · — · score:0.00 · hot:0.59 · rising:0.60 · durable:0.47 · board:rising · trend:stable* Fast iterable JSON parser. #### [filipecalegario/awesome-vibe-coding](https://github.com/filipecalegario/awesome-vibe-coding) *? · ★3,929 · CC0-1.0 · — · score:0.00 · hot:0.59 · rising:0.62 · durable:0.50 · board:rising · trend:up* A curated list of vibe coding references, collaborating with AI to write code. #### [ruc-datalab/DeepAnalyze](https://github.com/ruc-datalab/DeepAnalyze) *Python · ★4,022 · MIT · — · score:0.00 · hot:0.58 · rising:0.61 · durable:0.52 · board:rising · trend:stable* DeepAnalyze is the first agentic LLM for autonomous data science. 🎈你的AI数据分析师，自动分析大量数据，一键生成专业分析报告！ #### [Integuru-AI/Integuru](https://github.com/Integuru-AI/Integuru) *Python · ★4,567 · AGPL-3.0 · — · score:0.00 · hot:0.58 · rising:0.61 · durable:0.52 · board:rising · trend:stable* The first AI agent that builds permissionless integrations through reverse engineering platforms' internal APIs. #### [openvinotoolkit/openvino_notebooks](https://github.com/openvinotoolkit/openvino_notebooks) *Jupyter Notebook · ★3,102 · Apache-2.0 · — · score:0.00 · hot:0.58 · rising:0.62 · durable:0.47 · board:rising · trend:up* 📚 Jupyter notebook tutorials for OpenVINO™ #### [langchain-ai/langgraph-swarm-py](https://github.com/langchain-ai/langgraph-swarm-py) *Python · ★1,469 · MIT · — · score:0.00 · hot:0.58 · rising:0.61 · durable:0.56 · board:rising · trend:stable* For your multi-agent needs #### [NVIDIA/TensorRT-Incubator](https://github.com/NVIDIA/TensorRT-Incubator) *MLIR · ★123 · no-license · — · score:0.00 · hot:0.58 · rising:0.58 · durable:0.42 · board:hot · trend:stable* Experimental projects related to TensorRT #### [langchain-ai/open_deep_research](https://github.com/langchain-ai/open_deep_research) *Python · ★11,153 · MIT · — · score:0.00 · hot:0.58 · rising:0.64 · durable:0.54 · board:rising · trend:up* #### [weaviate/recipes](https://github.com/weaviate/recipes) *Jupyter Notebook · ★939 · no-license · — · score:0.00 · hot:0.58 · rising:0.59 · durable:0.49 · board:rising · trend:stable* This repository shares end-to-end notebooks on how to use various Weaviate features and integrations! #### [microsoft/aitour26-WRK541-real-world-code-migration-with-github-copilot-agent-mode](https://github.com/microsoft/aitour26-WRK541-real-world-code-migration-with-github-copilot-agent-mode) *C# · ★490 · MIT · — · score:0.00 · hot:0.58 · rising:0.63 · durable:0.49 · board:rising · trend:up* #### [modelcontextprotocol/mcpb](https://github.com/modelcontextprotocol/mcpb) *TypeScript · ★1,850 · NOASSERTION · — · score:0.00 · hot:0.58 · rising:0.61 · durable:0.50 · board:rising · trend:stable* Desktop Extensions: One-click local MCP server installation in desktop apps #### [High-Logic/Genie-TTS](https://github.com/High-Logic/Genie-TTS) *Python · ★1,500 · MIT · — · score:0.00 · hot:0.58 · rising:0.61 · durable:0.52 · board:rising · trend:stable* GPT-SoVITS ONNX Inference Engine & Model Converter #### [vercel/next-forge](https://github.com/vercel/next-forge) *TypeScript · ★7,042 · MIT · — · score:0.00 · hot:0.58 · rising:0.62 · durable:0.59 · board:rising · trend:stable* Production-grade Turborepo template for Next.js apps. #### [allenai/olmoearth_pretrain](https://github.com/allenai/olmoearth_pretrain) *Python · ★172 · NOASSERTION · — · score:0.00 · hot:0.58 · rising:0.59 · durable:0.43 · board:rising · trend:stable* Earth system foundation model data, training, and eval #### [microsoft/TinyTroupe](https://github.com/microsoft/TinyTroupe) *Jupyter Notebook · ★7,398 · MIT · — · score:0.00 · hot:0.58 · rising:0.63 · durable:0.60 · board:rising · trend:stable* LLM-powered multiagent persona simulation for imagination enhancement and business insights. #### [aws/amazon-sagemaker-examples](https://github.com/aws/amazon-sagemaker-examples) *Jupyter Notebook · ★10,900 · Apache-2.0 · — · score:0.00 · hot:0.58 · rising:0.63 · durable:0.53 · board:rising · trend:up* Example 📓 Jupyter notebooks that demonstrate how to build, train, and deploy machine learning models using 🧠 Amazon SageMaker. #### [NVIDIA/vgpu-device-manager](https://github.com/NVIDIA/vgpu-device-manager) *Go · ★155 · Apache-2.0 · — · score:0.00 · hot:0.58 · rising:0.60 · durable:0.47 · board:rising · trend:stable* NVIDIA vGPU Device Manager manages NVIDIA vGPU devices on top of Kubernetes #### [google-research/kauldron](https://github.com/google-research/kauldron) *Python · ★233 · Apache-2.0 · — · score:0.00 · hot:0.58 · rising:0.59 · durable:0.45 · board:rising · trend:stable* Modular, scalable library to train ML models #### [langchain-ai/langchain-academy](https://github.com/langchain-ai/langchain-academy) *Jupyter Notebook · ★2,553 · no-license · — · score:0.00 · hot:0.58 · rising:0.62 · durable:0.46 · board:rising · trend:up* #### [langchain-ai/social-media-agent](https://github.com/langchain-ai/social-media-agent) *TypeScript · ★2,495 · MIT · — · score:0.00 · hot:0.58 · rising:0.62 · durable:0.50 · board:rising · trend:up* 📲 An agent for sourcing, curating, and scheduling social media posts with human-in-the-loop. #### [microsoft/fluentui-system-icons](https://github.com/microsoft/fluentui-system-icons) *HTML · ★10,512 · MIT · — · score:0.00 · hot:0.58 · rising:0.63 · durable:0.48 · board:rising · trend:up* Fluent System Icons are a collection of familiar, friendly and modern icons from Microsoft. #### [modal-labs/modal-examples](https://github.com/modal-labs/modal-examples) *Python · ★1,161 · MIT · — · score:0.00 · hot:0.58 · rising:0.61 · durable:0.48 · board:rising · trend:up* Examples of programs built using Modal #### [crewAIInc/crewAI-examples](https://github.com/crewAIInc/crewAI-examples) *Jupyter Notebook · ★5,882 · no-license · — · score:0.00 · hot:0.58 · rising:0.62 · durable:0.49 · board:rising · trend:up* A collection of examples that show how to use CrewAI framework to automate workflows. #### [go-kratos/kratos](https://github.com/go-kratos/kratos) *Go · ★25,620 · MIT · — · score:0.00 · hot:0.58 · rising:0.64 · durable:0.61 · board:rising · trend:stable* Your ultimate Go microservices framework for the cloud-native era. #### [langchain-ai/langserve](https://github.com/langchain-ai/langserve) *JavaScript · ★2,313 · NOASSERTION · — · score:0.00 · hot:0.57 · rising:0.60 · durable:0.49 · board:rising · trend:stable* LangServe 🦜️🏓 #### [eracle/OpenOutreach](https://github.com/eracle/OpenOutreach) *Python · ★1,459 · NOASSERTION · — · score:0.00 · hot:0.57 · rising:0.57 · durable:0.41 · board:hot · trend:up* Linkedin Automation Tool: Describe your product. Define your target market. The AI finds the leads for you. #### [AsyncFuncAI/deepwiki-open](https://github.com/AsyncFuncAI/deepwiki-open) *Python · ★15,733 · MIT · — · score:0.00 · hot:0.57 · rising:0.60 · durable:0.53 · board:rising · trend:up* Open Source DeepWiki: AI-Powered Wiki Generator for GitHub/Gitlab/Bitbucket Repositories. Join the discord: https://discord.gg/gMwThUMeme #### [vercel/vrs](https://github.com/vercel/vrs) *TypeScript · ★527 · Apache-2.0 · — · score:0.00 · hot:0.57 · rising:0.61 · durable:0.45 · board:rising · trend:up* A serverless virtual reality e-commerce experience powered by Vercel #### [pinecone-io/examples](https://github.com/pinecone-io/examples) *Jupyter Notebook · ★3,008 · MIT · — · score:0.00 · hot:0.57 · rising:0.60 · durable:0.46 · board:rising · trend:up* Jupyter Notebooks to help you get hands-on with Pinecone vector databases #### [microsoft/vscode-livepreview](https://github.com/microsoft/vscode-livepreview) *TypeScript · ★545 · MIT · — · score:0.00 · hot:0.57 · rising:0.59 · durable:0.47 · board:rising · trend:stable* Hosts a local server in your workspace for you to preview your webpages. #### [langchain-ai/docs](https://github.com/langchain-ai/docs) *MDX · ★301 · MIT · — · score:0.00 · hot:0.57 · rising:0.62 · durable:0.43 · board:rising · trend:up* 🦜🔗 Docs for LangChain projects #### [langchain-ai/langchain-community](https://github.com/langchain-ai/langchain-community) *Python · ★262 · MIT · — · score:0.00 · hot:0.57 · rising:0.62 · durable:0.48 · board:rising · trend:stable* Community-maintained LangChain integrations #### [memvid/memvid](https://github.com/memvid/memvid) *Rust · ★15,027 · Apache-2.0 · — · score:0.00 · hot:0.57 · rising:0.60 · durable:0.64 · board:durable · trend:stable* Memory layer for AI Agents. Replace complex RAG pipelines with a serverless, single-file memory layer. Give your agents instant retrieval and long-term memory. #### [NVIDIA/go-dcgm](https://github.com/NVIDIA/go-dcgm) *C · ★151 · Apache-2.0 · — · score:0.00 · hot:0.57 · rising:0.59 · durable:0.44 · board:rising · trend:stable* Golang bindings for Nvidia Datacenter GPU Manager (DCGM) #### [langchain-ai/langchain-postgres](https://github.com/langchain-ai/langchain-postgres) *Python · ★274 · MIT · — · score:0.00 · hot:0.57 · rising:0.60 · durable:0.47 · board:rising · trend:stable* LangChain abstractions backed by Postgres Backend #### [google/XNNPACK](https://github.com/google/XNNPACK) *C · ★2,315 · NOASSERTION · — · score:0.00 · hot:0.57 · rising:0.58 · durable:0.39 · board:rising · trend:up* High-efficiency floating-point neural network inference operators for mobile, server, and Web #### [BoltzmannEntropy/MimikaStudio](https://github.com/BoltzmannEntropy/MimikaStudio) *Dart · ★522 · GPL-3.0 · — · score:0.00 · hot:0.57 · rising:0.58 · durable:0.54 · board:rising · trend:stable* MimikaStudio - A local-first application for macOS (Apple Silicon) + Agentic MCP Support #### [lucidrains/rectified-flow-pytorch](https://github.com/lucidrains/rectified-flow-pytorch) *Python · ★440 · MIT · — · score:0.00 · hot:0.57 · rising:0.59 · durable:0.50 · board:rising · trend:stable* Implementation of rectified flow and some of its followup research / improvements in Pytorch #### [NVIDIA/NeMo-text-processing](https://github.com/NVIDIA/NeMo-text-processing) *Python · ★452 · Apache-2.0 · — · score:0.00 · hot:0.57 · rising:0.62 · durable:0.51 · board:rising · trend:stable* NeMo text processing for ASR and TTS #### [microsoft/FastTrack](https://github.com/microsoft/FastTrack) *PowerShell · ★205 · MIT · — · score:0.00 · hot:0.57 · rising:0.61 · durable:0.44 · board:rising · trend:up* GitHub home for Microsoft FastTrack #### [lucidrains/dreamer4](https://github.com/lucidrains/dreamer4) *Python · ★178 · MIT · — · score:0.00 · hot:0.57 · rising:0.58 · durable:0.49 · board:rising · trend:stable* Implementation of Danijar's latest iteration for his Dreamer line of work #### [hanzili/hanzi-browse](https://github.com/hanzili/hanzi-browse) *JavaScript · ★151 · NOASSERTION · — · score:0.00 · hot:0.57 · rising:0.57 · durable:0.38 · board:hot · trend:stable* let any ai agent use the local browser #### [run-llama/ParseBench](https://github.com/run-llama/ParseBench) *Python · ★206 · Apache-2.0 · — · score:0.00 · hot:0.57 · rising:0.57 · durable:0.47 · board:rising · trend:stable* ParseBench - A Document Parsing Benchmark for AI Agents #### [vercel/ai-elements](https://github.com/vercel/ai-elements) *TypeScript · ★1,934 · NOASSERTION · — · score:0.00 · hot:0.57 · rising:0.58 · durable:0.50 · board:rising · trend:stable* AI Elements is a component library and custom registry built on top of shadcn/ui to help you build AI-native applications faster. #### [microsoft/zerotrustassessment](https://github.com/microsoft/zerotrustassessment) *HTML · ★328 · MIT · — · score:0.00 · hot:0.57 · rising:0.60 · durable:0.47 · board:rising · trend:stable* Repository for the Zero Trust Assessment project #### [NVIDIA/VisRTX](https://github.com/NVIDIA/VisRTX) *C++ · ★273 · NOASSERTION · — · score:0.00 · hot:0.57 · rising:0.59 · durable:0.45 · board:rising · trend:stable* NVIDIA OptiX based implementation of ANARI #### [langchain-ai/local-deep-researcher](https://github.com/langchain-ai/local-deep-researcher) *Python · ★9,031 · MIT · — · score:0.00 · hot:0.57 · rising:0.62 · durable:0.54 · board:rising · trend:up* Fully local web research and report writing assistant #### [open-webui/docs](https://github.com/open-webui/docs) *CSS · ★712 · no-license · — · score:0.00 · hot:0.56 · rising:0.59 · durable:0.40 · board:rising · trend:up* https://docs.openwebui.com #### [mistralai/client-ts](https://github.com/mistralai/client-ts) *TypeScript · ★133 · Apache-2.0 · — · score:0.00 · hot:0.56 · rising:0.58 · durable:0.48 · board:rising · trend:stable* TS Client library for Mistral AI platform #### [microsoft/azure-skills](https://github.com/microsoft/azure-skills) *Bicep · ★647 · MIT · — · score:0.00 · hot:0.56 · rising:0.59 · durable:0.48 · board:rising · trend:stable* Official agent plugin providing skills and MCP server configurations for Azure scenarios. #### [google-deepmind/jax_privacy](https://github.com/google-deepmind/jax_privacy) *Python · ★164 · Apache-2.0 · — · score:0.00 · hot:0.56 · rising:0.60 · durable:0.48 · board:rising · trend:stable* Algorithms for Privacy-Preserving Machine Learning in JAX #### [langchain-ai/langchain-nextjs-template](https://github.com/langchain-ai/langchain-nextjs-template) *TypeScript · ★2,462 · MIT · — · score:0.00 · hot:0.56 · rising:0.61 · durable:0.49 · board:rising · trend:stable* LangChain + Next.js starter template #### [modelcontextprotocol/quickstart-resources](https://github.com/modelcontextprotocol/quickstart-resources) *Go · ★1,065 · NOASSERTION · — · score:0.00 · hot:0.56 · rising:0.60 · durable:0.46 · board:rising · trend:stable* A repository of servers and clients from the Model Context Protocol tutorials #### [huggingface/optimum-onnx](https://github.com/huggingface/optimum-onnx) *Python · ★139 · Apache-2.0 · — · score:0.00 · hot:0.56 · rising:0.61 · durable:0.47 · board:rising · trend:stable* 🤗 Optimum ONNX: Export your model to ONNX and run inference with ONNX Runtime #### [microsoft/CopilotStudioSamples](https://github.com/microsoft/CopilotStudioSamples) *TypeScript · ★729 · MIT · — · score:0.00 · hot:0.56 · rising:0.60 · durable:0.41 · board:rising · trend:up* #### [NVIDIA/nvidia-resiliency-ext](https://github.com/NVIDIA/nvidia-resiliency-ext) *Python · ★283 · NOASSERTION · — · score:0.00 · hot:0.56 · rising:0.58 · durable:0.44 · board:rising · trend:stable* NVIDIA Resiliency Extension is a python package for framework developers and users to implement fault-tolerant features. It improves the effective training time by minimizing the downtime due to failures and interruptions. #### [e2b-dev/fragments](https://github.com/e2b-dev/fragments) *TypeScript · ★6,250 · Apache-2.0 · — · score:0.00 · hot:0.56 · rising:0.59 · durable:0.53 · board:rising · trend:stable* Open-source Next.js template for building apps that are fully generated by AI. By E2B. #### [Y-Research-SBU/PosterGen](https://github.com/Y-Research-SBU/PosterGen) *Python · ★230 · MIT · — · score:0.00 · hot:0.56 · rising:0.57 · durable:0.45 · board:rising · trend:stable* Official Repository for PosterGen - CVPR Findings 2026 #### [simonw/simonw](https://github.com/simonw/simonw) *Python · ★432 · Apache-2.0 · — · score:0.00 · hot:0.56 · rising:0.59 · durable:0.43 · board:rising · trend:stable* https://simonwillison.net/2020/Jul/10/self-updating-profile-readme/ #### [microsoft/sbom-tool](https://github.com/microsoft/sbom-tool) *C# · ★2,013 · MIT · — · score:0.00 · hot:0.56 · rising:0.60 · durable:0.50 · board:rising · trend:stable* The SBOM tool is a highly scalable and enterprise ready tool to create SPDX 2.2 compatible SBOMs for any variety of artifacts. #### [zml/zml](https://github.com/zml/zml) *Zig · ★3,417 · Apache-2.0 · — · score:0.00 · hot:0.56 · rising:0.57 · durable:0.49 · board:rising · trend:stable* Any model. Any hardware. Zero compromise. Built with @ziglang / @openxla / MLIR / @bazelbuild #### [kyegomez/BitNet](https://github.com/kyegomez/BitNet) *Python · ★1,919 · MIT · — · score:0.00 · hot:0.56 · rising:0.59 · durable:0.50 · board:rising · trend:stable* Implementation of "BitNet: Scaling 1-bit Transformers for Large Language Models" in pytorch #### [vercel/streamdown](https://github.com/vercel/streamdown) *TypeScript · ★5,031 · NOASSERTION · — · score:0.00 · hot:0.56 · rising:0.58 · durable:0.55 · board:rising · trend:stable* A drop-in replacement for react-markdown, designed for AI-powered streaming. #### [microsoft/PSRule](https://github.com/microsoft/PSRule) *C# · ★464 · MIT · — · score:0.00 · hot:0.56 · rising:0.59 · durable:0.44 · board:rising · trend:stable* Validate infrastructure as code (IaC) and objects using PowerShell rules. #### [NVIDIA/SOL-ExecBench](https://github.com/NVIDIA/SOL-ExecBench) *Python · ★177 · Apache-2.0 · — · score:0.00 · hot:0.56 · rising:0.60 · durable:0.47 · board:rising · trend:stable* A benchmark of real-world DL kernel problems #### [simonw/tools](https://github.com/simonw/tools) *HTML · ★1,623 · Apache-2.0 · — · score:0.00 · hot:0.56 · rising:0.58 · durable:0.43 · board:rising · trend:up* Assorted useful tools, almost entirely generated using LLMs #### [openvinotoolkit/open_model_zoo](https://github.com/openvinotoolkit/open_model_zoo) *Python · ★4,385 · Apache-2.0 · — · score:0.00 · hot:0.56 · rising:0.58 · durable:0.48 · board:rising · trend:up* Pre-trained Deep Learning models and demos (high quality and extremely fast) #### [NVIDIA/Megatron-Energon](https://github.com/NVIDIA/Megatron-Energon) *Python · ★336 · NOASSERTION · — · score:0.00 · hot:0.56 · rising:0.57 · durable:0.44 · board:rising · trend:stable* Megatron's multi-modal data loader #### [OpenMOSS/MOSS-TTS-Nano](https://github.com/OpenMOSS/MOSS-TTS-Nano) *Python · ★1,608 · Apache-2.0 · — · score:0.00 · hot:0.56 · rising:0.58 · durable:0.45 · board:rising · trend:stable* MOSS-TTS-Nano is an open-source multilingual tiny speech generation model from MOSI.AI and the OpenMOSS team. With only 0.1B parameters, it is designed for realtime speech generation, can run directly on CPU without a GPU, and keeps the deployment stack simple enough for local demos, web serving, and lightweight product integration. #### [microsoft/pxt](https://github.com/microsoft/pxt) *TypeScript · ★2,280 · MIT · — · score:0.00 · hot:0.56 · rising:0.60 · durable:0.41 · board:rising · trend:up* Microsoft MakeCode (PXT - Programming eXperience Toolkit) #### [huggingface/doc-builder](https://github.com/huggingface/doc-builder) *Python · ★138 · Apache-2.0 · — · score:0.00 · hot:0.55 · rising:0.59 · durable:0.45 · board:rising · trend:stable* The package used to build the documentation of our Hugging Face repos #### [microsoft/debugpy](https://github.com/microsoft/debugpy) *Python · ★2,395 · NOASSERTION · — · score:0.00 · hot:0.55 · rising:0.58 · durable:0.46 · board:rising · trend:stable* An implementation of the Debug Adapter Protocol for Python #### [microsoft/vc-ue-extensions](https://github.com/microsoft/vc-ue-extensions) *C++ · ★398 · NOASSERTION · — · score:0.00 · hot:0.55 · rising:0.57 · durable:0.50 · board:rising · trend:stable* Components for integration between Visual Studio and Unreal Engine. #### [langchain-ai/langgraph-supervisor-py](https://github.com/langchain-ai/langgraph-supervisor-py) *Python · ★1,555 · MIT · — · score:0.00 · hot:0.55 · rising:0.60 · durable:0.51 · board:rising · trend:stable* #### [microsoft/dion](https://github.com/microsoft/dion) *Python · ★466 · MIT · — · score:0.00 · hot:0.55 · rising:0.58 · durable:0.46 · board:rising · trend:stable* Dion optimizer algorithm #### [langchain-ai/deep-agents-from-scratch](https://github.com/langchain-ai/deep-agents-from-scratch) *Jupyter Notebook · ★652 · MIT · — · score:0.00 · hot:0.55 · rising:0.59 · durable:0.47 · board:rising · trend:stable* #### [langchain-ai/react-agent](https://github.com/langchain-ai/react-agent) *Python · ★718 · MIT · — · score:0.00 · hot:0.55 · rising:0.59 · durable:0.44 · board:rising · trend:stable* LangGraph template for a simple ReAct agent #### [bytedance/UI-TARS-desktop](https://github.com/bytedance/UI-TARS-desktop) *TypeScript · ★29,450 · Apache-2.0 · — · score:0.00 · hot:0.55 · rising:0.60 · durable:0.63 · board:durable · trend:stable* The Open-Source Multimodal AI Agent Stack: Connecting Cutting-Edge AI Models and Agent Infra #### [microsoft/aspire-samples](https://github.com/microsoft/aspire-samples) *C# · ★1,166 · MIT · — · score:0.00 · hot:0.55 · rising:0.59 · durable:0.43 · board:rising · trend:up* #### [pydantic/pydantic-extra-types](https://github.com/pydantic/pydantic-extra-types) *Python · ★314 · MIT · — · score:0.00 · hot:0.55 · rising:0.57 · durable:0.46 · board:rising · trend:stable* Extra Pydantic types. #### [NVIDIA/optix-toolkit](https://github.com/NVIDIA/optix-toolkit) *C++ · ★133 · BSD-3-Clause · — · score:0.00 · hot:0.55 · rising:0.58 · durable:0.49 · board:rising · trend:stable* Set of utilities supporting workflows common in GPU raytracing applications #### [langchain-ai/lca-lc-foundations](https://github.com/langchain-ai/lca-lc-foundations) *TypeScript · ★407 · no-license · — · score:0.00 · hot:0.55 · rising:0.57 · durable:0.43 · board:rising · trend:stable* #### [langchain-ai/agent-chat-ui](https://github.com/langchain-ai/agent-chat-ui) *TypeScript · ★2,738 · MIT · — · score:0.00 · hot:0.55 · rising:0.57 · durable:0.47 · board:rising · trend:stable* 🦜💬 Web app for interacting with any LangGraph agent (PY & TS) via a chat interface. #### [deepset-ai/haystack-cookbook](https://github.com/deepset-ai/haystack-cookbook) *Jupyter Notebook · ★538 · no-license · — · score:0.00 · hot:0.55 · rising:0.53 · durable:0.38 · board:hot · trend:stable* 👩🏻‍🍳 A collection of example notebooks using Haystack #### [KimYx0207/Claude-Code-x-OpenClaw-Guide-Zh](https://github.com/KimYx0207/Claude-Code-x-OpenClaw-Guide-Zh) *? · ★3,377 · MIT · — · score:0.00 · hot:0.55 · rising:0.57 · durable:0.53 · board:rising · trend:stable* 从零到企业实战：Claude Code 官方编程神器 + OpenClaw 224K Stars 开源AI助手 | 双顶流中文教程 | 21篇教程 130000+字 #### [simonw/research](https://github.com/simonw/research) *Python · ★578 · no-license · — · score:0.00 · hot:0.55 · rising:0.56 · durable:0.40 · board:rising · trend:stable* Research projects #### [MiniMax-AI/MiniMax-MCP](https://github.com/MiniMax-AI/MiniMax-MCP) *Python · ★1,432 · MIT · — · score:0.00 · hot:0.55 · rising:0.57 · durable:0.45 · board:rising · trend:stable* Official MiniMax Model Context Protocol (MCP) server that enables interaction with powerful Text to Speech, image generation and video generation APIs. #### [langchain-ai/langconnect](https://github.com/langchain-ai/langconnect) *Python · ★341 · MIT · — · score:0.00 · hot:0.55 · rising:0.58 · durable:0.46 · board:rising · trend:stable* A managed RAG API server. #### [vercel/workflow-examples](https://github.com/vercel/workflow-examples) *TypeScript · ★462 · MIT · — · score:0.00 · hot:0.55 · rising:0.58 · durable:0.45 · board:rising · trend:stable* Example projects and templates build using Workflow DevKit #### [langchain-ai/langgraph-101](https://github.com/langchain-ai/langgraph-101) *Jupyter Notebook · ★379 · MIT · — · score:0.00 · hot:0.55 · rising:0.58 · durable:0.47 · board:rising · trend:stable* Learn about the fundamentals of LangGraph through a series of notebooks #### [microsoft/WindowsAgentArena](https://github.com/microsoft/WindowsAgentArena) *Python · ★855 · MIT · — · score:0.00 · hot:0.55 · rising:0.57 · durable:0.49 · board:rising · trend:stable* Windows Agent Arena (WAA) 🪟 is a scalable OS platform for testing and benchmarking of multi-modal AI agents. #### [langchain-ai/mcpdoc](https://github.com/langchain-ai/mcpdoc) *Python · ★978 · MIT · — · score:0.00 · hot:0.55 · rising:0.57 · durable:0.54 · board:rising · trend:stable* Expose llms-txt to IDEs for development #### [NVIDIA/ncx-infra-controller-core](https://github.com/NVIDIA/ncx-infra-controller-core) *Rust · ★125 · Apache-2.0 · — · score:0.00 · hot:0.55 · rising:0.57 · durable:0.39 · board:rising · trend:stable* NCX Infra Controller - Hardware Lifecycle Management and multitenant networking #### [microsoft/mcp-dotnet-samples](https://github.com/microsoft/mcp-dotnet-samples) *C# · ★175 · MIT · — · score:0.00 · hot:0.55 · rising:0.56 · durable:0.44 · board:rising · trend:stable* A comprehensive set of samples of creating and using MCP servers and clients with .NET #### [microsoft/ebpf-for-windows](https://github.com/microsoft/ebpf-for-windows) *C · ★3,474 · MIT · — · score:0.00 · hot:0.55 · rising:0.58 · durable:0.43 · board:rising · trend:up* eBPF implementation that runs on top of Windows #### [vllm-project/recipes](https://github.com/vllm-project/recipes) *Jupyter Notebook · ★683 · Apache-2.0 · — · score:0.00 · hot:0.55 · rising:0.58 · durable:0.42 · board:rising · trend:stable* Common recipes to run vLLM #### [ghostwright/ghost-os](https://github.com/ghostwright/ghost-os) *Swift · ★1,360 · MIT · — · score:0.00 · hot:0.55 · rising:0.55 · durable:0.57 · board:durable · trend:stable* Full computer-use for AI agents. Self-learning workflows. Native macOS. No screenshots required. #### [swyxio/gh-action-data-scraping](https://github.com/swyxio/gh-action-data-scraping) *JavaScript · ★241 · MIT · — · score:0.00 · hot:0.55 · rising:0.56 · durable:0.41 · board:rising · trend:stable* this shows how to use github actions to do periodic data scraping #### [elixir-nx/bumblebee](https://github.com/elixir-nx/bumblebee) *Elixir · ★1,615 · Apache-2.0 · — · score:0.00 · hot:0.54 · rising:0.57 · durable:0.51 · board:rising · trend:stable* Pre-trained Neural Network models in Axon (+ 🤗 Models integration) #### [microsoft/MIDI](https://github.com/microsoft/MIDI) *C++ · ★617 · MIT · — · score:0.00 · hot:0.54 · rising:0.57 · durable:0.40 · board:rising · trend:stable* Windows MIDI Services #### [tinyfish-io/tinyfish-cookbook](https://github.com/tinyfish-io/tinyfish-cookbook) *TypeScript · ★1,632 · MIT · — · score:0.00 · hot:0.54 · rising:0.56 · durable:0.43 · board:rising · trend:stable* A collection of sample apps and recipes built with the TinyFish web agent. Open-source examples for you to learn & build! #### [qdrant/qdrant-dotnet](https://github.com/qdrant/qdrant-dotnet) *C# · ★213 · Apache-2.0 · — · score:0.00 · hot:0.54 · rising:0.57 · durable:0.49 · board:rising · trend:stable* Qdrant .Net SDK #### [liyupi/ai-guide](https://github.com/liyupi/ai-guide) *JavaScript · ★12,139 · no-license · — · score:0.00 · hot:0.54 · rising:0.54 · durable:0.55 · board:durable · trend:stable* 程序员鱼皮的 AI 资源大全 + Vibe Coding 零基础教程，分享 OpenClaw 保姆级教程、大模型玩法（DeepSeek / GPT / Gemini / Claude）、最新 AI 资讯、Prompt 提示词大全、AI 知识百科（Agent Skills / RAG / MCP / A2A）、AI 编程教程（Harness Engineering）、AI 工具用法（Cursor / Claude Code / TRAE / Lovable / Copilot）、AI 开发框架教程（Spring AI / LangChain）、AI 产品变现指南，帮你快速掌握 AI 技术，走在时代前沿。本项目为开源文档，已升级为鱼皮 AI 导航网站 #### [vercel/release](https://github.com/vercel/release) *JavaScript · ★3,584 · MIT · — · score:0.00 · hot:0.54 · rising:0.57 · durable:0.52 · board:rising · trend:stable* Generate changelogs with a single command #### [kristerkari/react-native-svg-transformer](https://github.com/kristerkari/react-native-svg-transformer) *JavaScript · ★1,735 · MIT · — · score:0.00 · hot:0.54 · rising:0.55 · durable:0.46 · board:rising · trend:stable* Import SVG files in your React Native project the same way that you would in a Web application. #### [FoundationVision/Infinity](https://github.com/FoundationVision/Infinity) *Python · ★1,558 · MIT · — · score:0.00 · hot:0.54 · rising:0.55 · durable:0.42 · board:rising · trend:stable* [CVPR 2025 Oral]Infinity ∞ : Scaling Bitwise AutoRegressive Modeling for High-Resolution Image Synthesis #### [facebookresearch/iopath](https://github.com/facebookresearch/iopath) *Python · ★152 · MIT · — · score:0.00 · hot:0.54 · rising:0.58 · durable:0.47 · board:rising · trend:stable* A python library that provides common I/O interface across different storage backends. #### [microsoft/wslg](https://github.com/microsoft/wslg) *C++ · ★11,583 · MIT · — · score:0.00 · hot:0.54 · rising:0.58 · durable:0.53 · board:rising · trend:stable* Enabling the Windows Subsystem for Linux to include support for Wayland and X server related scenarios #### [killop/anything_about_game](https://github.com/killop/anything_about_game) *? · ★3,857 · Apache-2.0 · — · score:0.00 · hot:0.54 · rising:0.57 · durable:0.51 · board:rising · trend:stable* A wonderful list of Game Development resources. #### [pydantic/speedate](https://github.com/pydantic/speedate) *Rust · ★248 · MIT · — · score:0.00 · hot:0.54 · rising:0.58 · durable:0.49 · board:rising · trend:stable* Fast and simple datetime, date, time and duration parsing for rust. #### [GokuMohandas/Made-With-ML](https://github.com/GokuMohandas/Made-With-ML) *Jupyter Notebook · ★47,336 · MIT · — · score:0.00 · hot:0.54 · rising:0.62 · durable:0.64 · board:durable · trend:stable* Learn how to develop, deploy and iterate on production-grade ML applications. #### [facebookresearch/ResponsibleNLP](https://github.com/facebookresearch/ResponsibleNLP) *Python · ★207 · NOASSERTION · — · score:0.00 · hot:0.54 · rising:0.56 · durable:0.39 · board:rising · trend:stable* Repository for research in the field of Responsible NLP at Meta. #### [google-deepmind/ai-foundations](https://github.com/google-deepmind/ai-foundations) *Jupyter Notebook · ★198 · Apache-2.0 · — · score:0.00 · hot:0.54 · rising:0.57 · durable:0.42 · board:rising · trend:stable* #### [huggingface/optimum-executorch](https://github.com/huggingface/optimum-executorch) *Python · ★124 · Apache-2.0 · — · score:0.00 · hot:0.54 · rising:0.57 · durable:0.46 · board:rising · trend:stable* 🤗 Optimum ExecuTorch #### [facebookresearch/optimizers](https://github.com/facebookresearch/optimizers) *Python · ★564 · NOASSERTION · — · score:0.00 · hot:0.54 · rising:0.57 · durable:0.50 · board:rising · trend:stable* For optimization algorithm research and development. #### [microsoft/DirectX-Specs](https://github.com/microsoft/DirectX-Specs) *HTML · ★829 · CC-BY-4.0 · — · score:0.00 · hot:0.54 · rising:0.57 · durable:0.40 · board:rising · trend:up* Engineering specs for DirectX features. #### [microsoft/mu](https://github.com/microsoft/mu) *Python · ★653 · NOASSERTION · — · score:0.00 · hot:0.54 · rising:0.56 · durable:0.42 · board:rising · trend:stable* Project Mu Documentation #### [langchain-ai/agent-auth-payments](https://github.com/langchain-ai/agent-auth-payments) *TypeScript · ★193 · MIT · — · score:0.00 · hot:0.54 · rising:0.55 · durable:0.47 · board:rising · trend:stable* fullstack chat agent with authentication, request credits and payments built in #### [NVIDIA/JAX-Toolbox](https://github.com/NVIDIA/JAX-Toolbox) *Python · ★403 · Apache-2.0 · — · score:0.00 · hot:0.54 · rising:0.57 · durable:0.40 · board:rising · trend:stable* JAX-Toolbox #### [huggingface/ml-for-3d-course](https://github.com/huggingface/ml-for-3d-course) *MDX · ★177 · no-license · — · score:0.00 · hot:0.54 · rising:0.56 · durable:0.41 · board:rising · trend:stable* #### [FurkanGozukara/Stable-Diffusion](https://github.com/FurkanGozukara/Stable-Diffusion) *JavaScript · ★2,678 · GPL-3.0 · — · score:0.00 · hot:0.54 · rising:0.58 · durable:0.49 · board:rising · trend:stable* FLUX, Stable Diffusion, SDXL, SD3, LoRA, Fine Tuning, DreamBooth, Training, Automatic1111, Forge WebUI, SwarmUI, DeepFake, TTS, Animation, Text To Video, Tutorials, Guides, Lectures, Courses, ComfyUI, Google Colab, RunPod, Kaggle, NoteBooks, ControlNet, TTS, Voice Cloning, AI, AI News, ML, ML News, News, Tech, Tech News, Kohya, Midjourney, RunPod #### [microsoft/DirectStorage](https://github.com/microsoft/DirectStorage) *C++ · ★846 · MIT · — · score:0.00 · hot:0.54 · rising:0.56 · durable:0.44 · board:rising · trend:stable* DirectStorage for Windows is an API that allows game developers to unlock the full potential of high speed NVMe drives for loading game assets. #### [Lightricks/ComfyUI-LTXVideo](https://github.com/Lightricks/ComfyUI-LTXVideo) *Python · ★3,485 · NOASSERTION · — · score:0.00 · hot:0.54 · rising:0.56 · durable:0.42 · board:rising · trend:stable* LTX-Video Support for ComfyUI #### [modal-labs/modal-client](https://github.com/modal-labs/modal-client) *Python · ★464 · Apache-2.0 · — · score:0.00 · hot:0.54 · rising:0.56 · durable:0.42 · board:rising · trend:stable* SDK libraries for Modal #### [lucidrains/evolutionary-policy-optimization](https://github.com/lucidrains/evolutionary-policy-optimization) *Python · ★108 · MIT · — · score:0.00 · hot:0.54 · rising:0.54 · durable:0.49 · board:rising · trend:stable* Pytorch implementation of Evolutionary Policy Optimization, from Wang et al. of the Robotics Institute at Carnegie Mellon University #### [sgl-project/sgl-project.github.io](https://github.com/sgl-project/sgl-project.github.io) *HTML · ★128 · no-license · — · score:0.00 · hot:0.54 · rising:0.54 · durable:0.37 · board:rising · trend:stable* This is the documentation repository for SGLang. It is auto-generated from https://github.com/sgl-project/sglang #### [sgl-project/sglang-omni](https://github.com/sgl-project/sglang-omni) *Python · ★214 · MIT · — · score:0.00 · hot:0.53 · rising:0.56 · durable:0.40 · board:rising · trend:stable* SGLang Omni: High-Performance Multi-Stage Pipeline Framework for Omni Models #### [microsoft/security-devops-action](https://github.com/microsoft/security-devops-action) *TypeScript · ★151 · MIT · — · score:0.00 · hot:0.53 · rising:0.56 · durable:0.43 · board:rising · trend:stable* Microsoft Security DevOps for GitHub Actions. #### [NVIDIA/cuCollections](https://github.com/NVIDIA/cuCollections) *C++ · ★635 · Apache-2.0 · — · score:0.00 · hot:0.53 · rising:0.55 · durable:0.40 · board:rising · trend:stable* #### [facebookresearch/nymeria_dataset](https://github.com/facebookresearch/nymeria_dataset) *Python · ★171 · NOASSERTION · — · score:0.00 · hot:0.53 · rising:0.54 · durable:0.38 · board:rising · trend:stable* Nymeria: a massive collection of multimodal egocentric daily motion in the wild #### [google-deepmind/physics-IQ-benchmark](https://github.com/google-deepmind/physics-IQ-benchmark) *Python · ★284 · NOASSERTION · — · score:0.00 · hot:0.53 · rising:0.55 · durable:0.43 · board:rising · trend:stable* Benchmarking physical understanding in generative video models #### [datawhalechina/happy-llm](https://github.com/datawhalechina/happy-llm) *Jupyter Notebook · ★29,102 · NOASSERTION · — · score:0.00 · hot:0.53 · rising:0.57 · durable:0.62 · board:durable · trend:stable* 📚 从零开始构建大模型 #### [foxhui/WebAI2API](https://github.com/foxhui/WebAI2API) *JavaScript · ★582 · MIT · — · score:0.00 · hot:0.53 · rising:0.54 · durable:0.41 · board:rising · trend:stable* WebAI2API: 基于 Camoufox 的网页 AI 转 API 工具，支持 LMArena/Gemini等，多窗口并发与账号隔离。 | Web AI to OpenAI API via Camoufox. Supports LMArena/Gemini and more, multi-window concurrency & account isolation. #### [milvus-io/web-content](https://github.com/milvus-io/web-content) *MDX · ★136 · Apache-2.0 · — · score:0.00 · hot:0.53 · rising:0.56 · durable:0.36 · board:rising · trend:stable* Milvus web documents and contents #### [langchain-ai/langmem](https://github.com/langchain-ai/langmem) *Python · ★1,403 · MIT · — · score:0.00 · hot:0.53 · rising:0.56 · durable:0.43 · board:rising · trend:stable* #### [langchain-ai/new-langgraph-project](https://github.com/langchain-ai/new-langgraph-project) *Python · ★258 · MIT · — · score:0.00 · hot:0.53 · rising:0.57 · durable:0.45 · board:rising · trend:stable* #### [qdrant/vector-db-benchmark](https://github.com/qdrant/vector-db-benchmark) *Python · ★356 · Apache-2.0 · — · score:0.00 · hot:0.53 · rising:0.55 · durable:0.40 · board:rising · trend:stable* Framework for benchmarking vector search engines #### [huggingface/hf-mcp-server](https://github.com/huggingface/hf-mcp-server) *TypeScript · ★220 · MIT · — · score:0.00 · hot:0.53 · rising:0.56 · durable:0.51 · board:rising · trend:stable* Hugging Face MCP Server #### [langchain-ai/memory-agent](https://github.com/langchain-ai/memory-agent) *Python · ★423 · MIT · — · score:0.00 · hot:0.53 · rising:0.56 · durable:0.44 · board:rising · trend:stable* #### [huggingface/competitions](https://github.com/huggingface/competitions) *Python · ★124 · Apache-2.0 · — · score:0.00 · hot:0.53 · rising:0.56 · durable:0.43 · board:rising · trend:stable* #### [sgl-project/sglang-jax](https://github.com/sgl-project/sglang-jax) *Python · ★265 · Apache-2.0 · — · score:0.00 · hot:0.53 · rising:0.55 · durable:0.40 · board:rising · trend:stable* JAX backend for SGL #### [google-deepmind/tips](https://github.com/google-deepmind/tips) *Jupyter Notebook · ★300 · Apache-2.0 · — · score:0.00 · hot:0.53 · rising:0.56 · durable:0.46 · board:rising · trend:stable* TIPSv2 (CVPR'26) and TIPS (ICLR'25) #### [microsoft/SqlScriptDOM](https://github.com/microsoft/SqlScriptDOM) *GAP · ★244 · MIT · — · score:0.00 · hot:0.53 · rising:0.55 · durable:0.39 · board:rising · trend:stable* ScriptDOM/SqlDOM is a .NET library for parsing T-SQL statements and interacting with its abstract syntax tree #### [facebookresearch/LayerSkip](https://github.com/facebookresearch/LayerSkip) *Python · ★366 · NOASSERTION · — · score:0.00 · hot:0.53 · rising:0.52 · durable:0.41 · board:hot · trend:stable* Code for "LayerSkip: Enabling Early Exit Inference and Self-Speculative Decoding", ACL 2024 #### [lucidrains/ppo](https://github.com/lucidrains/ppo) *Python · ★112 · MIT · — · score:0.00 · hot:0.53 · rising:0.54 · durable:0.40 · board:rising · trend:stable* An implementation of PPO in Pytorch #### [microsoft/DebugMCP](https://github.com/microsoft/DebugMCP) *TypeScript · ★306 · MIT · — · score:0.00 · hot:0.53 · rising:0.51 · durable:0.41 · board:hot · trend:stable* Gift your VS Code agent a real debugger: breakpoints, stepping, inspection. #### [microsoft/copilot-camp](https://github.com/microsoft/copilot-camp) *C# · ★543 · MIT · — · score:0.00 · hot:0.53 · rising:0.56 · durable:0.41 · board:rising · trend:stable* Hands-on labs for extending Microsoft 365 Copilot and building custom engine agents #### [rasbt/mini-coding-agent](https://github.com/rasbt/mini-coding-agent) *Python · ★689 · Apache-2.0 · — · score:0.00 · hot:0.53 · rising:0.54 · durable:0.49 · board:rising · trend:stable* Minimal and readable coding agent harness implementation in Python to explain the core components of coding agents. #### [Isi-dev/Google-Colab_Notebooks](https://github.com/Isi-dev/Google-Colab_Notebooks) *Jupyter Notebook · ★443 · no-license · — · score:0.00 · hot:0.53 · rising:0.55 · durable:0.42 · board:rising · trend:stable* A Collection of Google Colab Notebooks for various projects #### [langchain-ai/openwork](https://github.com/langchain-ai/openwork) *TypeScript · ★1,384 · MIT · — · score:0.00 · hot:0.53 · rising:0.57 · durable:0.55 · board:rising · trend:stable* #### [facebookresearch/param](https://github.com/facebookresearch/param) *Python · ★154 · MIT · — · score:0.00 · hot:0.53 · rising:0.55 · durable:0.37 · board:rising · trend:stable* PArametrized Recommendation and Ai Model benchmark is a repository for development of numerous uBenchmarks as well as end to end nets for evaluation of training and inference platforms. #### [google-research/weatherbenchX](https://github.com/google-research/weatherbenchX) *Python · ★207 · Apache-2.0 · — · score:0.00 · hot:0.53 · rising:0.56 · durable:0.46 · board:rising · trend:stable* A modular framework for evaluating weather forecasts #### [langchain-ai/agent-inbox](https://github.com/langchain-ai/agent-inbox) *TypeScript · ★969 · MIT · — · score:0.00 · hot:0.52 · rising:0.55 · durable:0.43 · board:rising · trend:stable* 📥 An inbox UX for interacting with human-in-the-loop agents. #### [langchain-ai/data-enrichment](https://github.com/langchain-ai/data-enrichment) *Jupyter Notebook · ★228 · MIT · — · score:0.00 · hot:0.52 · rising:0.55 · durable:0.42 · board:rising · trend:stable* LangGraph Studio template for creating an agent that does web research to genearte or enrich structured data. #### [milvus-io/milvus-sdk-node](https://github.com/milvus-io/milvus-sdk-node) *TypeScript · ★194 · Apache-2.0 · — · score:0.00 · hot:0.52 · rising:0.55 · durable:0.50 · board:rising · trend:stable* The Official Milvus node.js sdk(client) #### [NVIDIA/nvkind](https://github.com/NVIDIA/nvkind) *Go · ★195 · Apache-2.0 · — · score:0.00 · hot:0.52 · rising:0.55 · durable:0.40 · board:rising · trend:stable* #### [langchain-ai/langsmith-docs](https://github.com/langchain-ai/langsmith-docs) *JavaScript · ★167 · MIT · — · score:0.00 · hot:0.52 · rising:0.54 · durable:0.41 · board:rising · trend:stable* This repo is deprecated. Please go to langchain-ai/docs. #### [jxnl/blog](https://github.com/jxnl/blog) *Python · ★195 · MIT · — · score:0.00 · hot:0.52 · rising:0.54 · durable:0.39 · board:rising · trend:stable* #### [NVIDIA/physicsnemo-sym](https://github.com/NVIDIA/physicsnemo-sym) *Python · ★324 · Apache-2.0 · — · score:0.00 · hot:0.52 · rising:0.54 · durable:0.48 · board:rising · trend:stable* Framework providing pythonic APIs, algorithms and utilities to be used with PhysicsNeMo core to physics inform model training as well as higher level abstraction for domain experts #### [vllm-project/router](https://github.com/vllm-project/router) *Rust · ★198 · Apache-2.0 · — · score:0.00 · hot:0.52 · rising:0.54 · durable:0.39 · board:rising · trend:stable* A high-performance and light-weight router for vLLM large scale deployment #### [langchain-ai/retrieval-agent-template](https://github.com/langchain-ai/retrieval-agent-template) *Python · ★158 · MIT · — · score:0.00 · hot:0.52 · rising:0.55 · durable:0.43 · board:rising · trend:stable* #### [langchain-ai/intro-to-langsmith](https://github.com/langchain-ai/intro-to-langsmith) *Jupyter Notebook · ★156 · MIT · — · score:0.00 · hot:0.52 · rising:0.56 · durable:0.43 · board:rising · trend:stable* Resources for the LangSmith Academy Course #### [langchain-ai/voice-sandwich-demo](https://github.com/langchain-ai/voice-sandwich-demo) *TypeScript · ★125 · no-license · — · score:0.00 · hot:0.52 · rising:0.54 · durable:0.40 · board:rising · trend:stable* #### [vercel/avatar](https://github.com/vercel/avatar) *TypeScript · ★1,377 · MIT · — · score:0.00 · hot:0.52 · rising:0.55 · durable:0.45 · board:rising · trend:stable* 💎 Beautiful avatars as a microservice #### [huggingface/Google-Cloud-Containers](https://github.com/huggingface/Google-Cloud-Containers) *Dockerfile · ★173 · Apache-2.0 · — · score:0.00 · hot:0.52 · rising:0.55 · durable:0.39 · board:rising · trend:stable* Hugging Face Deep Learning Containers (DLCs) for Google Cloud #### [karpathy/autoresearch](https://github.com/karpathy/autoresearch) *Python · ★74,448 · no-license · — · score:0.00 · hot:0.52 · rising:0.59 · durable:0.57 · board:rising · trend:stable* AI agents running research on single-GPU nanochat training automatically #### [facebookresearch/CutLER](https://github.com/facebookresearch/CutLER) *Python · ★1,062 · NOASSERTION · — · score:0.00 · hot:0.52 · rising:0.54 · durable:0.42 · board:rising · trend:stable* Code release for "Cut and Learn for Unsupervised Object Detection and Instance Segmentation" and "VideoCutLER: Surprisingly Simple Unsupervised Video Instance Segmentation" #### [WuMingDao/zenith-image-generator](https://github.com/WuMingDao/zenith-image-generator) *TypeScript · ★205 · no-license · — · score:0.00 · hot:0.52 · rising:0.53 · durable:0.41 · board:rising · trend:stable* Modern AI image generator with multi-provider support (Gitee AI, HuggingFace, ModelScope), OpenAI-compatible API, token rotation, and one-click deployment to Cloudflare Pages. #### [allenai/molmospaces](https://github.com/allenai/molmospaces) *Python · ★284 · NOASSERTION · — · score:0.00 · hot:0.52 · rising:0.53 · durable:0.39 · board:rising · trend:stable* An end-to-end open ecosystem for robot learning #### [vercel/next-evals-oss](https://github.com/vercel/next-evals-oss) *TypeScript · ★252 · MIT · — · score:0.00 · hot:0.52 · rising:0.54 · durable:0.41 · board:rising · trend:stable* Evals for Next.js up to 15.5.6 to test AI model competency at Next.js #### [NVIDIA/nsight-python](https://github.com/NVIDIA/nsight-python) *Python · ★193 · Apache-2.0 · — · score:0.00 · hot:0.52 · rising:0.54 · durable:0.44 · board:rising · trend:stable* Nsight Python is a Python kernel profiling interface based on NVIDIA Nsight Tools #### [NVIDIA/gpu-driver-container](https://github.com/NVIDIA/gpu-driver-container) *Shell · ★166 · Apache-2.0 · — · score:0.00 · hot:0.52 · rising:0.55 · durable:0.38 · board:rising · trend:stable* The NVIDIA GPU driver container allows the provisioning of the NVIDIA driver through the use of containers. #### [microsoft/fabric-samples](https://github.com/microsoft/fabric-samples) *Jupyter Notebook · ★559 · MIT · — · score:0.00 · hot:0.52 · rising:0.56 · durable:0.42 · board:rising · trend:stable* Samples and data for Microsoft Fabric Learn content #### [huggingface/transformers-research-projects](https://github.com/huggingface/transformers-research-projects) *Python · ★122 · Apache-2.0 · — · score:0.00 · hot:0.52 · rising:0.55 · durable:0.44 · board:rising · trend:stable* Research projects built on top of Transformers #### [microsoft/tslib](https://github.com/microsoft/tslib) *TypeScript · ★1,337 · 0BSD · — · score:0.00 · hot:0.52 · rising:0.56 · durable:0.46 · board:rising · trend:stable* Runtime library for TypeScript helpers. #### [google-deepmind/loft](https://github.com/google-deepmind/loft) *Python · ★230 · Apache-2.0 · — · score:0.00 · hot:0.52 · rising:0.54 · durable:0.45 · board:rising · trend:stable* LOFT: A 1 Million+ Token Long-Context Benchmark #### [replicate/cog-examples](https://github.com/replicate/cog-examples) *Python · ★176 · no-license · — · score:0.00 · hot:0.51 · rising:0.55 · durable:0.41 · board:rising · trend:stable* Some models defined with Cog to show you how it works #### [google-deepmind/tf2jax](https://github.com/google-deepmind/tf2jax) *Python · ★120 · Apache-2.0 · — · score:0.00 · hot:0.51 · rising:0.54 · durable:0.43 · board:rising · trend:stable* #### [microsoft/snmalloc](https://github.com/microsoft/snmalloc) *C++ · ★1,812 · MIT · — · score:0.00 · hot:0.51 · rising:0.54 · durable:0.48 · board:rising · trend:stable* Message passing based allocator #### [sgl-project/sgl-cookbook](https://github.com/sgl-project/sgl-cookbook) *JavaScript · ★121 · Apache-2.0 · — · score:0.00 · hot:0.51 · rising:0.54 · durable:0.41 · board:rising · trend:stable* Cookbook of SGLang - Recipe #### [simonw/llm-gemini](https://github.com/simonw/llm-gemini) *Python · ★438 · Apache-2.0 · — · score:0.00 · hot:0.51 · rising:0.52 · durable:0.45 · board:rising · trend:stable* LLM plugin to access Google's Gemini family of models #### [OpenAdaptAI/OpenAdapt](https://github.com/OpenAdaptAI/OpenAdapt) *Python · ★1,556 · MIT · — · score:0.00 · hot:0.51 · rising:0.51 · durable:0.59 · board:durable · trend:stable* Open Source Generative Process Automation (i.e. Generative RPA). AI-First Process Automation with Large ([Language (LLMs) / Action (LAMs) / Multimodal (LMMs)] / Visual Language (VLMs)) Models #### [vercel/geist-font](https://github.com/vercel/geist-font) *HTML · ★3,382 · OFL-1.1 · — · score:0.00 · hot:0.51 · rising:0.53 · durable:0.55 · board:durable · trend:stable* #### [nidhinjs/prompt-master](https://github.com/nidhinjs/prompt-master) *? · ★5,353 · MIT · — · score:0.00 · hot:0.51 · rising:0.56 · durable:0.54 · board:rising · trend:stable* A Claude skill that writes the accurate prompts for any AI tool. Zero tokens or credits wasted. Full context and memory retention #### [facebookresearch/stopes](https://github.com/facebookresearch/stopes) *Python · ★303 · MIT · — · score:0.00 · hot:0.51 · rising:0.54 · durable:0.39 · board:rising · trend:stable* A library for preparing data for machine translation research (monolingual preprocessing, bitext mining, etc.) built by the FAIR NLLB team. #### [microsoft/Semi-supervised-learning](https://github.com/microsoft/Semi-supervised-learning) *Python · ★1,568 · MIT · — · score:0.00 · hot:0.51 · rising:0.55 · durable:0.57 · board:durable · trend:stable* A Unified Semi-Supervised Learning Codebase (NeurIPS'22) #### [langchain-ai/deep-agents-ui](https://github.com/langchain-ai/deep-agents-ui) *TypeScript · ★1,545 · MIT · — · score:0.00 · hot:0.51 · rising:0.55 · durable:0.44 · board:rising · trend:stable* Custom UI for Deep Agents #### [vercel/analytics](https://github.com/vercel/analytics) *TypeScript · ★505 · MIT · — · score:0.00 · hot:0.51 · rising:0.54 · durable:0.47 · board:rising · trend:stable* Privacy-friendly, real-time traffic insights #### [2U1/Qwen-VL-Series-Finetune](https://github.com/2U1/Qwen-VL-Series-Finetune) *Python · ★1,820 · Apache-2.0 · — · score:0.00 · hot:0.51 · rising:0.52 · durable:0.43 · board:rising · trend:stable* An open-source implementaion for fine-tuning Qwen-VL series by Alibaba Cloud. #### [activeloopai/deeplake](https://github.com/activeloopai/deeplake) *C++ · ★9,091 · Apache-2.0 · — · score:0.00 · hot:0.51 · rising:0.53 · durable:0.60 · board:durable · trend:stable* Deeplake is AI Data Runtime for Agents. It provides serverless postgres with a multimodal datalake, enabling scalable retrieval and training. #### [huggingface/open_asr_leaderboard](https://github.com/huggingface/open_asr_leaderboard) *Python · ★192 · Apache-2.0 · — · score:0.00 · hot:0.51 · rising:0.54 · durable:0.39 · board:rising · trend:stable* #### [ghostsecurity/reaper](https://github.com/ghostsecurity/reaper) *Go · ★848 · Apache-2.0 · — · score:0.00 · hot:0.51 · rising:0.54 · durable:0.54 · board:durable · trend:stable* Live validation proxy tool for testing web app vulnerabilities #### [karpathy/karpathy.github.io](https://github.com/karpathy/karpathy.github.io) *CSS · ★1,293 · no-license · — · score:0.00 · hot:0.51 · rising:0.53 · durable:0.36 · board:rising · trend:stable* my blog #### [simonw/simonwillisonblog](https://github.com/simonw/simonwillisonblog) *Python · ★404 · Apache-2.0 · — · score:0.00 · hot:0.51 · rising:0.51 · durable:0.35 · board:rising · trend:stable* The source code behind my blog #### [jxnl/dots](https://github.com/jxnl/dots) *Python · ★136 · no-license · — · score:0.00 · hot:0.51 · rising:0.52 · durable:0.37 · board:rising · trend:stable* #### [yoshitomo-matsubara/torchdistill](https://github.com/yoshitomo-matsubara/torchdistill) *Python · ★1,611 · MIT · — · score:0.00 · hot:0.51 · rising:0.53 · durable:0.55 · board:durable · trend:stable* A coding-free framework built on PyTorch for reproducible deep learning studies. PyTorch Ecosystem. 🏆26 knowledge distillation methods presented at TPAMI, CVPR, ICLR, ECCV, NeurIPS, ICCV, AAAI, etc are implemented so far. 🎁 Trained models, training logs and configurations are available for ensuring the reproducibiliy and benchmark. #### [morphik-org/morphik-core](https://github.com/morphik-org/morphik-core) *Python · ★3,575 · NOASSERTION · — · score:0.00 · hot:0.51 · rising:0.50 · durable:0.47 · board:hot · trend:stable* The most accurate document search and store for building AI apps #### [microsoft/ARI](https://github.com/microsoft/ARI) *PowerShell · ★1,597 · MIT · — · score:0.00 · hot:0.51 · rising:0.57 · durable:0.55 · board:rising · trend:stable* Azure Resource Inventory - It's a Powerful tool to create EXCEL inventory from Azure Resources with low effort #### [NVIDIA/nvtrust](https://github.com/NVIDIA/nvtrust) *Python · ★311 · Apache-2.0 · — · score:0.00 · hot:0.51 · rising:0.53 · durable:0.46 · board:rising · trend:stable* Ancillary open source software to support confidential computing on NVIDIA GPUs #### [adongwanai/AgentGuide](https://github.com/adongwanai/AgentGuide) *HTML · ★3,968 · no-license · — · score:0.00 · hot:0.50 · rising:0.50 · durable:0.47 · board:hot · trend:stable* https://adongwanai.github.io/AgentGuide | AI Agent开发指南 | LangGraph实战 | 高级RAG | 转行大模型 | 大模型面试 | 算法工程师 | 面试题库 | 强化学习｜数据合成 #### [microsoft/edgeai-for-beginners](https://github.com/microsoft/edgeai-for-beginners) *Jupyter Notebook · ★1,414 · MIT · — · score:0.00 · hot:0.50 · rising:0.55 · durable:0.50 · board:rising · trend:stable* This course is designed to guide beginners through the exciting world of Edge AI, covering fundamental concepts, popular models, inference techniques, device-specific applications, model optimization, and the development of intelligent Edge AI agents. #### [swyxio/swyxio](https://github.com/swyxio/swyxio) *Python · ★136 · MIT · — · score:0.00 · hot:0.50 · rising:0.52 · durable:0.35 · board:rising · trend:stable* readmeee #### [EleutherAI/elk](https://github.com/EleutherAI/elk) *Python · ★218 · MIT · — · score:0.00 · hot:0.50 · rising:0.53 · durable:0.45 · board:rising · trend:stable* Keeping language models honest by directly eliciting knowledge encoded in their activations. #### [microsoft/GDK](https://github.com/microsoft/GDK) *? · ★1,627 · NOASSERTION · — · score:0.00 · hot:0.50 · rising:0.54 · durable:0.51 · board:rising · trend:stable* Microsoft Public GDK #### [langchain-ai/langgraph-bigtool](https://github.com/langchain-ai/langgraph-bigtool) *Python · ★529 · MIT · — · score:0.00 · hot:0.50 · rising:0.54 · durable:0.51 · board:rising · trend:stable* Build LangGraph agents with large numbers of tools #### [google-research/falken](https://github.com/google-research/falken) *Python · ★268 · Apache-2.0 · — · score:0.00 · hot:0.50 · rising:0.54 · durable:0.45 · board:rising · trend:stable* Falken provides developers with a service that allows them to train AI that can play their games #### [google-deepmind/alphaevolve_repository_of_problems](https://github.com/google-deepmind/alphaevolve_repository_of_problems) *Jupyter Notebook · ★213 · Apache-2.0 · — · score:0.00 · hot:0.50 · rising:0.52 · durable:0.45 · board:rising · trend:stable* #### [google-research/cascades](https://github.com/google-research/cascades) *Python · ★219 · Apache-2.0 · — · score:0.00 · hot:0.50 · rising:0.54 · durable:0.50 · board:rising · trend:stable* Python library which enables complex compositions of language models such as scratchpads, chain of thought, tool use, selection-inference, and more. #### [allenai/autodiscovery](https://github.com/allenai/autodiscovery) *Python · ★171 · no-license · — · score:0.00 · hot:0.50 · rising:0.52 · durable:0.43 · board:rising · trend:stable* Official code for NeurIPS 2025 paper "AutoDiscovery: Open-ended Scientific Discovery via Bayesian Surprise" #### [facebookresearch/DRTK](https://github.com/facebookresearch/DRTK) *C++ · ★143 · MIT · — · score:0.00 · hot:0.50 · rising:0.52 · durable:0.42 · board:rising · trend:stable* Differentiable Rendering Toolkit #### [qdrant/examples](https://github.com/qdrant/examples) *Jupyter Notebook · ★210 · Apache-2.0 · — · score:0.00 · hot:0.50 · rising:0.52 · durable:0.39 · board:rising · trend:stable* A collection of examples and tutorials for Qdrant vector search engine #### [facebookresearch/coconut](https://github.com/facebookresearch/coconut) *Python · ★1,571 · MIT · — · score:0.00 · hot:0.50 · rising:0.53 · durable:0.46 · board:rising · trend:stable* Training Large Language Model to Reason in a Continuous Latent Space #### [Amshaker/Mobile-O](https://github.com/Amshaker/Mobile-O) *Python · ★141 · NOASSERTION · — · score:0.00 · hot:0.50 · rising:0.50 · durable:0.40 · board:rising · trend:stable* [CVPR'26 Demo] Mobile-O: Unified Multimodal Understanding and Generation on Mobile Device #### [FoundationAgents/MetaGPT](https://github.com/FoundationAgents/MetaGPT) *Python · ★67,234 · MIT · — · score:0.00 · hot:0.50 · rising:0.59 · durable:0.66 · board:durable · trend:stable* 🌟 The Multi-Agent Framework: First AI Software Company, Towards Natural Language Programming #### [microsoft/PowerAppsCodeApps](https://github.com/microsoft/PowerAppsCodeApps) *TypeScript · ★385 · MIT · — · score:0.00 · hot:0.49 · rising:0.53 · durable:0.41 · board:rising · trend:stable* Create custom web applications to run within Power Apps! #### [facebookresearch/jepa-wms](https://github.com/facebookresearch/jepa-wms) *Python · ★216 · NOASSERTION · — · score:0.00 · hot:0.49 · rising:0.51 · durable:0.42 · board:rising · trend:stable* Code, data and weights for the paper **What drives success in physical planning with Joint-Embedding Predictive World Models?** #### [allenai/IFBench](https://github.com/allenai/IFBench) *Python · ★128 · Apache-2.0 · — · score:0.00 · hot:0.49 · rising:0.52 · durable:0.44 · board:rising · trend:stable* #### [vanna-ai/vanna](https://github.com/vanna-ai/vanna) *Python · ★23,299 · MIT · — · score:0.00 · hot:0.49 · rising:0.55 · durable:0.61 · board:durable · trend:stable* 🤖 Chat with your SQL database 📊. Accurate Text-to-SQL Generation via LLMs using Agentic Retrieval 🔄. #### [swyxio/swyxdotio](https://github.com/swyxio/swyxdotio) *Svelte · ★408 · MIT · — · score:0.00 · hot:0.49 · rising:0.51 · durable:0.35 · board:rising · trend:stable* This is the repo for swyx's blog - Blog content is created in github issues, then posted on swyx.io as blog pages! Comment/watch to follow along my blog within GitHub #### [facebookresearch/aira-dojo](https://github.com/facebookresearch/aira-dojo) *Python · ★142 · NOASSERTION · — · score:0.00 · hot:0.49 · rising:0.50 · durable:0.39 · board:rising · trend:stable* AIRA-dojo: a framework for developing and evaluating AI research agents #### [dair-ai/Prompt-Engineering-Guide](https://github.com/dair-ai/Prompt-Engineering-Guide) *MDX · ★73,513 · MIT · — · score:0.00 · hot:0.49 · rising:0.55 · durable:0.60 · board:durable · trend:stable* 🐙 Guides, papers, lessons, notebooks and resources for prompt engineering, context engineering, RAG, and AI Agents. #### [qdrant/mcp-server-qdrant](https://github.com/qdrant/mcp-server-qdrant) *Python · ★1,359 · Apache-2.0 · — · score:0.00 · hot:0.49 · rising:0.52 · durable:0.52 · board:rising · trend:stable* An official Qdrant Model Context Protocol (MCP) server implementation #### [microsoft/vscode-markdown-languageservice](https://github.com/microsoft/vscode-markdown-languageservice) *TypeScript · ★434 · MIT · — · score:0.00 · hot:0.49 · rising:0.52 · durable:0.46 · board:rising · trend:stable* The language service that powers VS Code's Markdown support, extracted so that it can be reused by other editors and tools #### [simonw/datasette.io](https://github.com/simonw/datasette.io) *HTML · ★133 · no-license · — · score:0.00 · hot:0.49 · rising:0.49 · durable:0.32 · board:hot · trend:stable* The official project website for Datasette #### [xuyang-liu16/VidCom2](https://github.com/xuyang-liu16/VidCom2) *Python · ★124 · Apache-2.0 · — · score:0.00 · hot:0.49 · rising:0.50 · durable:0.43 · board:rising · trend:stable* [EMNLP 2025 Main] Video Compression Commander: Plug-and-Play Inference Acceleration for Video Large Language Models #### [facebookresearch/hot3d](https://github.com/facebookresearch/hot3d) *Python · ★238 · Apache-2.0 · — · score:0.00 · hot:0.49 · rising:0.53 · durable:0.45 · board:rising · trend:stable* HOT3D: Hand and Object Tracking in 3D from Egocentric Multi-View Videos, CVPR 2025 #### [vercel/vercel-plugin](https://github.com/vercel/vercel-plugin) *TypeScript · ★139 · no-license · — · score:0.00 · hot:0.49 · rising:0.49 · durable:0.37 · board:rising · trend:stable* Comprehensive Vercel ecosystem plugin — relational knowledge graph, skills for every major product, specialized agents, and Vercel conventions. Turns any AI agent into a Vercel expert. #### [qdrant/rust-client](https://github.com/qdrant/rust-client) *Rust · ★399 · Apache-2.0 · — · score:0.00 · hot:0.49 · rising:0.51 · durable:0.47 · board:rising · trend:stable* Rust client for Qdrant vector search engine #### [rasbt/llm-architecture-gallery](https://github.com/rasbt/llm-architecture-gallery) *? · ★1,045 · Apache-2.0 · — · score:0.00 · hot:0.49 · rising:0.51 · durable:0.49 · board:rising · trend:stable* LLM Architecture Gallery source data #### [google-research/perceiver-ar](https://github.com/google-research/perceiver-ar) *Python · ★253 · Apache-2.0 · — · score:0.00 · hot:0.49 · rising:0.51 · durable:0.38 · board:rising · trend:stable* #### [minitap-ai/mobile-use](https://github.com/minitap-ai/mobile-use) *Python · ★2,448 · Apache-2.0 · — · score:0.00 · hot:0.49 · rising:0.53 · durable:0.57 · board:durable · trend:stable* AI agents can now use real Android and iOS apps, just like a human. #### [stas00/ml-engineering](https://github.com/stas00/ml-engineering) *Python · ★17,732 · CC-BY-SA-4.0 · — · score:0.00 · hot:0.49 · rising:0.51 · durable:0.56 · board:durable · trend:stable* Machine Learning Engineering Open Book #### [francozanardi/pictex](https://github.com/francozanardi/pictex) *Python · ★201 · MIT · — · score:0.00 · hot:0.49 · rising:0.46 · durable:0.49 · board:hot · trend:stable* A Python library for efficient image generation using CSS Flexbox #### [vercel/mcp-handler](https://github.com/vercel/mcp-handler) *TypeScript · ★585 · no-license · — · score:0.00 · hot:0.49 · rising:0.51 · durable:0.46 · board:rising · trend:stable* Easily spin up an MCP Server on Next.js, Nuxt, Svelte, and more #### [facebookresearch/mvdust3r](https://github.com/facebookresearch/mvdust3r) *Python · ★584 · NOASSERTION · — · score:0.00 · hot:0.49 · rising:0.50 · durable:0.42 · board:rising · trend:stable* Open source impl of **MV-DUSt3R+ Single-Stage Scene Reconstruction from Sparse Views In 2 Seconds** from Meta Reality Labs. Project page https://mv-dust3rp.github.io/ #### [gvergnaud/ts-pattern](https://github.com/gvergnaud/ts-pattern) *TypeScript · ★14,949 · MIT · — · score:0.00 · hot:0.49 · rising:0.48 · durable:0.57 · board:durable · trend:stable* 🎨 The exhaustive Pattern Matching library for TypeScript, with smart type inference. #### [huggingface/simulate](https://github.com/huggingface/simulate) *Python · ★193 · Apache-2.0 · — · score:0.00 · hot:0.49 · rising:0.51 · durable:0.38 · board:rising · trend:stable* 🎢 Creating and sharing simulation environments for embodied and synthetic data research #### [facebookresearch/pippo](https://github.com/facebookresearch/pippo) *Python · ★640 · NOASSERTION · — · score:0.00 · hot:0.49 · rising:0.50 · durable:0.42 · board:rising · trend:stable* Pippo: High-Resolution Multi-View Humans from a Single Image #### [NVIDIA/digital-biology-examples](https://github.com/NVIDIA/digital-biology-examples) *? · ★216 · no-license · — · score:0.00 · hot:0.48 · rising:0.50 · durable:0.40 · board:rising · trend:stable* NVIDIA Digital Biology examples for optimized inference and training at scale #### [ddz16/TSFpaper](https://github.com/ddz16/TSFpaper) *? · ★3,124 · no-license · — · score:0.00 · hot:0.48 · rising:0.51 · durable:0.48 · board:rising · trend:stable* This repository contains a reading list of papers on Time Series Forecasting/Prediction (TSF) and Spatio-Temporal Forecasting/Prediction (STF). These papers are mainly categorized according to the type of model. #### [huggingface/gym-aloha](https://github.com/huggingface/gym-aloha) *Python · ★198 · Apache-2.0 · — · score:0.00 · hot:0.48 · rising:0.51 · durable:0.43 · board:rising · trend:stable* A gym environment for ALOHA #### [jim-schwoebel/awesome_ai_agents](https://github.com/jim-schwoebel/awesome_ai_agents) *? · ★1,564 · Apache-2.0 · — · score:0.00 · hot:0.48 · rising:0.52 · durable:0.51 · board:rising · trend:stable* 🤖 A comprehensive list of 1,500+ resources and tools related to AI agents. #### [tencentmusic/cube-studio](https://github.com/tencentmusic/cube-studio) *Python · ★4,958 · NOASSERTION · — · score:0.00 · hot:0.48 · rising:0.52 · durable:0.55 · board:durable · trend:stable* cube studio开源云原生一站式机器学习/深度学习/大模型AI平台，mlops算法链路全流程，算力租赁平台，notebook在线开发，拖拉拽任务流pipeline编排，多机多卡分布式训练，超参搜索，推理服务VGPU虚拟化，边缘计算，标注平台自动化标注，deepseek等大模型sft微调/奖励模型/强化学习训练，vllm/ollama/mindie大模型多机推理，私有知识库，AI模型市场，支持国产cpu/gpu/npu 昇腾生态，支持RDMA，支持pytorch/tf/mxnet/deepspeed/paddle/colossalai/horovod/ray/volcano等分布式 #### [modal-labs/vprox](https://github.com/modal-labs/vprox) *Go · ★203 · MIT · — · score:0.00 · hot:0.48 · rising:0.52 · durable:0.51 · board:rising · trend:stable* High-availability network proxy / VPN server, powered by WireGuard #### [vercel/swr-site](https://github.com/vercel/swr-site) *MDX · ★507 · Apache-2.0 · — · score:0.00 · hot:0.48 · rising:0.52 · durable:0.40 · board:rising · trend:stable* The official website for SWR. #### [pydantic/FastUI](https://github.com/pydantic/FastUI) *Python · ★8,962 · MIT · — · score:0.00 · hot:0.48 · rising:0.53 · durable:0.57 · board:durable · trend:stable* Build better UIs faster. #### [777genius/os-ai-computer-use](https://github.com/777genius/os-ai-computer-use) *Python · ★150 · Apache-2.0 · — · score:0.00 · hot:0.48 · rising:0.47 · durable:0.52 · board:durable · trend:stable* AI controls your OS. OS AI Computer Use, OS and API agnostic. For now on OpenAI and Anthropic API. Desktop app ready. #### [langchain-ai/langgraph-cua-py](https://github.com/langchain-ai/langgraph-cua-py) *Python · ★205 · MIT · — · score:0.00 · hot:0.48 · rising:0.53 · durable:0.51 · board:rising · trend:stable* An implementation of a computer use agent (CUA) using LangGraph #### [AgentDeskAI/browser-tools-mcp](https://github.com/AgentDeskAI/browser-tools-mcp) *JavaScript · ★7,193 · MIT · — · score:0.00 · hot:0.48 · rising:0.53 · durable:0.60 · board:durable · trend:stable* Monitor browser logs directly from Cursor and other MCP compatible IDEs. #### [vercel/satori](https://github.com/vercel/satori) *TypeScript · ★13,294 · MPL-2.0 · — · score:0.00 · hot:0.48 · rising:0.51 · durable:0.58 · board:durable · trend:stable* Enlightened library to convert HTML and CSS to SVG #### [huggingface/huggingface-gemma-recipes](https://github.com/huggingface/huggingface-gemma-recipes) *Jupyter Notebook · ★294 · MIT · — · score:0.00 · hot:0.48 · rising:0.52 · durable:0.46 · board:rising · trend:stable* Inference, Fine Tuning and many more recipes with Gemma family of models #### [microsoft/powerbi-modeling-mcp](https://github.com/microsoft/powerbi-modeling-mcp) *? · ★655 · MIT · — · score:0.00 · hot:0.48 · rising:0.51 · durable:0.43 · board:rising · trend:stable* The Power BI Modeling MCP Server, brings Power BI semantic modeling capabilities to your AI agents. #### [safe-graph/graph-fraud-detection-papers](https://github.com/safe-graph/graph-fraud-detection-papers) *? · ★1,821 · no-license · — · score:0.00 · hot:0.47 · rising:0.49 · durable:0.46 · board:rising · trend:stable* A curated list of Graph/Transformer-based fraud, anomaly, and outlier detection papers & resources #### [modal-labs/synchronicity](https://github.com/modal-labs/synchronicity) *Python · ★135 · Apache-2.0 · — · score:0.00 · hot:0.47 · rising:0.48 · durable:0.38 · board:rising · trend:stable* Synchronicity lets you interoperate with asynchronous Python APIs. #### [allenai/FlexOlmo](https://github.com/allenai/FlexOlmo) *Python · ★135 · Apache-2.0 · — · score:0.00 · hot:0.47 · rising:0.50 · durable:0.39 · board:rising · trend:stable* Code and training scripts for FlexOlmo #### [huggingface/swift-jinja](https://github.com/huggingface/swift-jinja) *Swift · ★123 · Apache-2.0 · — · score:0.00 · hot:0.47 · rising:0.51 · durable:0.51 · board:durable · trend:stable* A minimalistic Swift implementation of the Jinja templating engine, specifically designed for parsing and rendering ML chat templates. #### [datawhalechina/all-in-rag](https://github.com/datawhalechina/all-in-rag) *Python · ★6,351 · no-license · — · score:0.00 · hot:0.47 · rising:0.50 · durable:0.53 · board:durable · trend:stable* 🔍大模型应用开发实战一：RAG 技术全栈指南，在线阅读地址：https://datawhalechina.github.io/all-in-rag/ #### [promptslab/Promptify](https://github.com/promptslab/Promptify) *Python · ★4,584 · Apache-2.0 · — · score:0.00 · hot:0.47 · rising:0.50 · durable:0.49 · board:rising · trend:stable* Prompt Engineering | Prompt Versioning | Use GPT or other prompt based models to get structured output. Join our discord for Prompt-Engineering, LLMs and other latest research #### [FellouAI/eko](https://github.com/FellouAI/eko) *TypeScript · ★4,907 · MIT · — · score:0.00 · hot:0.47 · rising:0.52 · durable:0.62 · board:durable · trend:stable* Eko (Eko Keeps Operating) - Build Production-ready Agentic Workflow with Natural Language - eko.fellou.ai #### [lucidrains/adam-atan2-pytorch](https://github.com/lucidrains/adam-atan2-pytorch) *Python · ★136 · MIT · — · score:0.00 · hot:0.47 · rising:0.45 · durable:0.48 · board:durable · trend:down* Implementation of the proposed Adam-atan2 from Google Deepmind in Pytorch #### [docarray/docarray](https://github.com/docarray/docarray) *Python · ★3,116 · Apache-2.0 · — · score:0.00 · hot:0.47 · rising:0.50 · durable:0.50 · board:durable · trend:stable* Represent, send, store and search multimodal data #### [facebookresearch/ReasonIR](https://github.com/facebookresearch/ReasonIR) *Python · ★227 · NOASSERTION · — · score:0.00 · hot:0.47 · rising:0.48 · durable:0.41 · board:rising · trend:stable* Official repository for paper "ReasonIR Training Retrievers for Reasoning Tasks". #### [jackmpcollins/magentic](https://github.com/jackmpcollins/magentic) *Python · ★2,405 · MIT · — · score:0.00 · hot:0.47 · rising:0.48 · durable:0.53 · board:durable · trend:stable* Seamlessly integrate LLMs as Python functions #### [asukaminato0721/telegram-summary-bot](https://github.com/asukaminato0721/telegram-summary-bot) *TypeScript · ★186 · Apache-2.0 · — · score:0.00 · hot:0.47 · rising:0.47 · durable:0.41 · board:rising · trend:stable* Summarize group chat with AI, LLM && query group chat, FREE to deploy your own, support img, link meta info, reply to, auto fold result, 支持中文检索. #### [friuns2/BlackFriday-GPTs-Prompts](https://github.com/friuns2/BlackFriday-GPTs-Prompts) *? · ★9,346 · MIT · — · score:0.00 · hot:0.47 · rising:0.50 · durable:0.51 · board:durable · trend:stable* List of free GPTs that doesn't require plus subscription #### [qdrant/go-client](https://github.com/qdrant/go-client) *Go · ★325 · Apache-2.0 · — · score:0.00 · hot:0.47 · rising:0.50 · durable:0.51 · board:durable · trend:stable* Go client for Qdrant vector search engine #### [simular-ai/Agent-S](https://github.com/simular-ai/Agent-S) *Python · ★10,874 · Apache-2.0 · — · score:0.00 · hot:0.47 · rising:0.53 · durable:0.62 · board:durable · trend:stable* Agent S: an open agentic framework that uses computers like a human #### [facebookresearch/RPG_KDD2025](https://github.com/facebookresearch/RPG_KDD2025) *Python · ★130 · NOASSERTION · — · score:0.00 · hot:0.47 · rising:0.48 · durable:0.42 · board:rising · trend:stable* This repository provides the code for implementing RPG described in our KDD'25 paper "Generating Long Semantic IDs in Parallel for Recommendation". #### [microsoft/vscode-hexeditor](https://github.com/microsoft/vscode-hexeditor) *TypeScript · ★586 · MIT · — · score:0.00 · hot:0.47 · rising:0.51 · durable:0.46 · board:rising · trend:stable* VS Code Hex Editor #### [volcengine/MineContext](https://github.com/volcengine/MineContext) *Python · ★5,257 · Apache-2.0 · — · score:0.00 · hot:0.46 · rising:0.50 · durable:0.56 · board:durable · trend:stable* MineContext is your proactive context-aware AI partner（Context-Engineering+ChatGPT Pulse） #### [rom1504/clip-retrieval](https://github.com/rom1504/clip-retrieval) *Jupyter Notebook · ★2,751 · MIT · — · score:0.00 · hot:0.46 · rising:0.50 · durable:0.49 · board:rising · trend:stable* Easily compute clip embeddings and build a clip retrieval system with them #### [open-webui/mcpo](https://github.com/open-webui/mcpo) *Python · ★4,146 · MIT · — · score:0.00 · hot:0.46 · rising:0.53 · durable:0.58 · board:durable · trend:stable* A simple, secure MCP-to-OpenAPI proxy server #### [sgl-project/SpecForge](https://github.com/sgl-project/SpecForge) *Python · ★789 · MIT · — · score:0.00 · hot:0.46 · rising:0.48 · durable:0.44 · board:rising · trend:stable* Train speculative decoding models effortlessly and port them smoothly to SGLang serving. #### [huggingface/gym-pusht](https://github.com/huggingface/gym-pusht) *Python · ★165 · Apache-2.0 · — · score:0.00 · hot:0.46 · rising:0.49 · durable:0.41 · board:rising · trend:stable* A gym environment for PushT #### [langchain-ai/agent-protocol](https://github.com/langchain-ai/agent-protocol) *Python · ★569 · MIT · — · score:0.00 · hot:0.46 · rising:0.50 · durable:0.49 · board:rising · trend:stable* #### [run-llama/semtools](https://github.com/run-llama/semtools) *Rust · ★1,773 · MIT · — · score:0.00 · hot:0.46 · rising:0.52 · durable:0.57 · board:durable · trend:stable* Semantic search and document parsing tools for the command line #### [index-tts/index-tts](https://github.com/index-tts/index-tts) *Python · ★20,110 · NOASSERTION · — · score:0.00 · hot:0.46 · rising:0.53 · durable:0.55 · board:durable · trend:stable* An Industrial-Level Controllable and Efficient Zero-Shot Text-To-Speech System #### [ComposioHQ/awesome-claude-skills](https://github.com/ComposioHQ/awesome-claude-skills) *Python · ★54,910 · no-license · — · score:0.00 · hot:0.46 · rising:0.52 · durable:0.56 · board:durable · trend:stable* A curated list of awesome Claude Skills, resources, and tools for customizing Claude AI workflows #### [jina-ai/serve](https://github.com/jina-ai/serve) *Python · ★21,876 · Apache-2.0 · — · score:0.00 · hot:0.46 · rising:0.55 · durable:0.63 · board:durable · trend:stable* ☁️ Build multimodal AI applications with cloud-native stack #### [vercel/next-learn](https://github.com/vercel/next-learn) *TypeScript · ★4,705 · MIT · — · score:0.00 · hot:0.46 · rising:0.52 · durable:0.45 · board:rising · trend:stable* Learn Next.js Starter Code #### [facebookresearch/partnr-planner](https://github.com/facebookresearch/partnr-planner) *Python · ★362 · MIT · — · score:0.00 · hot:0.46 · rising:0.49 · durable:0.40 · board:rising · trend:stable* A repository accompanying the PARTNR benchmark for using Large Planning Models (LPMs) to solve Human-Robot Collaboration or Robot Instruction Following tasks in the Habitat simulator. #### [qdrant/page-search](https://github.com/qdrant/page-search) *Python · ★151 · Apache-2.0 · — · score:0.00 · hot:0.46 · rising:0.48 · durable:0.42 · board:rising · trend:stable* Neural search for web-sites, docs, articles - online! #### [google-research/pathdreamer](https://github.com/google-research/pathdreamer) *Jupyter Notebook · ★160 · Apache-2.0 · — · score:0.00 · hot:0.46 · rising:0.49 · durable:0.39 · board:rising · trend:stable* #### [simonw/llm-openrouter](https://github.com/simonw/llm-openrouter) *Python · ★307 · Apache-2.0 · — · score:0.00 · hot:0.46 · rising:0.49 · durable:0.46 · board:rising · trend:stable* LLM plugin for models hosted by OpenRouter #### [lucidrains/deep-cross-attention](https://github.com/lucidrains/deep-cross-attention) *Python · ★102 · MIT · — · score:0.00 · hot:0.46 · rising:0.47 · durable:0.50 · board:durable · trend:down* Implementation of the proposed DeepCrossAttention by Heddes et al at Google research, in Pytorch #### [smallcloudai/refact](https://github.com/smallcloudai/refact) *Rust · ★3,533 · BSD-3-Clause · — · score:0.00 · hot:0.46 · rising:0.50 · durable:0.56 · board:durable · trend:stable* AI Agent that handles engineering tasks end-to-end: integrates with developers’ tools, plans, executes, and iterates until it achieves a successful result. #### [A9T9/RPA](https://github.com/A9T9/RPA) *JavaScript · ★1,887 · NOASSERTION · — · score:0.00 · hot:0.46 · rising:0.50 · durable:0.51 · board:durable · trend:stable* Ui.Vision Open-Source RPA Software with Computer Vision, OCR, Anthropic Computer Use/LLM. Selenium IDE import/export. #### [lucidrains/locoformer](https://github.com/lucidrains/locoformer) *Python · ★109 · MIT · — · score:0.00 · hot:0.46 · rising:0.47 · durable:0.50 · board:durable · trend:down* LocoFormer - Generalist Locomotion via Long-Context Adaptation #### [huggingface/responses.js](https://github.com/huggingface/responses.js) *TypeScript · ★228 · MIT · — · score:0.00 · hot:0.46 · rising:0.49 · durable:0.45 · board:rising · trend:stable* A lightweight express.js server implementing OpenAI’s Responses API, built on top of Chat Completions, powered by Hugging Face Inference Providers. #### [qdrant/qdrant-js](https://github.com/qdrant/qdrant-js) *TypeScript · ★440 · Apache-2.0 · — · score:0.00 · hot:0.46 · rising:0.48 · durable:0.47 · board:rising · trend:stable* JavaScript/Typescript SDK for Qdrant Vector Database #### [microsoft/PowerBI-JavaScript](https://github.com/microsoft/PowerBI-JavaScript) *TypeScript · ★1,135 · NOASSERTION · — · score:0.00 · hot:0.45 · rising:0.51 · durable:0.45 · board:rising · trend:stable* JavaScript library for embedding Power BI into your apps. Check out the docs website and wiki for more information. #### [allenai/agent-baselines](https://github.com/allenai/agent-baselines) *Python · ★130 · Apache-2.0 · — · score:0.00 · hot:0.45 · rising:0.48 · durable:0.44 · board:rising · trend:stable* #### [mlabonne/llm-course](https://github.com/mlabonne/llm-course) *? · ★78,431 · Apache-2.0 · — · score:0.00 · hot:0.45 · rising:0.55 · durable:0.59 · board:durable · trend:stable* Course to get into Large Language Models (LLMs) with roadmaps and Colab notebooks. #### [openai/openai-guardrails-python](https://github.com/openai/openai-guardrails-python) *Python · ★203 · MIT · — · score:0.00 · hot:0.45 · rising:0.50 · durable:0.52 · board:durable · trend:stable* OpenAI Guardrails - Python #### [microsoft/AirSim](https://github.com/microsoft/AirSim) *C++ · ★18,117 · NOASSERTION · — · score:0.00 · hot:0.45 · rising:0.53 · durable:0.51 · board:rising · trend:stable* Open source simulator for autonomous vehicles built on Unreal Engine / Unity, from Microsoft AI & Research #### [labmlai/annotated_deep_learning_paper_implementations](https://github.com/labmlai/annotated_deep_learning_paper_implementations) *Python · ★66,329 · MIT · — · score:0.00 · hot:0.45 · rising:0.53 · durable:0.58 · board:durable · trend:stable* 🧑‍🏫 60+ Implementations/tutorials of deep learning papers with side-by-side notes 📝; including transformers (original, xl, switch, feedback, vit, ...), optimizers (adam, adabelief, sophia, ...), gans(cyclegan, stylegan2, ...), 🎮 reinforcement learning (ppo, dqn), capsnet, distillation, ... 🧠 #### [vercel/components.build](https://github.com/vercel/components.build) *MDX · ★755 · NOASSERTION · — · score:0.00 · hot:0.45 · rising:0.46 · durable:0.39 · board:rising · trend:stable* An open-source standard for building modern, composable and accessible UI components. #### [google-research/mood-board-search](https://github.com/google-research/mood-board-search) *Jupyter Notebook · ★175 · Apache-2.0 · — · score:0.00 · hot:0.45 · rising:0.48 · durable:0.37 · board:rising · trend:stable* #### [simonw/llm-mistral](https://github.com/simonw/llm-mistral) *Python · ★213 · Apache-2.0 · — · score:0.00 · hot:0.45 · rising:0.48 · durable:0.48 · board:rising · trend:stable* LLM plugin providing access to Mistral models using the Mistral API #### [vercel/serve](https://github.com/vercel/serve) *TypeScript · ★9,845 · MIT · — · score:0.00 · hot:0.45 · rising:0.52 · durable:0.54 · board:durable · trend:stable* Static file serving and directory listing #### [huggingface/api-inference-community](https://github.com/huggingface/api-inference-community) *Python · ★172 · Apache-2.0 · — · score:0.00 · hot:0.45 · rising:0.49 · durable:0.38 · board:rising · trend:stable* #### [rasbt/python-machine-learning-book](https://github.com/rasbt/python-machine-learning-book) *Jupyter Notebook · ★12,604 · MIT · — · score:0.00 · hot:0.45 · rising:0.56 · durable:0.58 · board:durable · trend:stable* The "Python Machine Learning (1st edition)" book code repository and info resource #### [instavm/clickclickclick](https://github.com/instavm/clickclickclick) *Python · ★682 · MIT · — · score:0.00 · hot:0.45 · rising:0.48 · durable:0.55 · board:durable · trend:stable* A framework to enable autonomous android and computer use using any LLM (local or remote) #### [openai/openai-reflect](https://github.com/openai/openai-reflect) *C++ · ★191 · MIT · — · score:0.00 · hot:0.45 · rising:0.50 · durable:0.47 · board:rising · trend:stable* Physical AI Assistant that illuminates your life #### [BrowserOperator/browser-operator-core](https://github.com/BrowserOperator/browser-operator-core) *TypeScript · ★474 · BSD-3-Clause · — · score:0.00 · hot:0.45 · rising:0.46 · durable:0.51 · board:durable · trend:stable* Browser Operator - The AI browser with built in Multi-Agent platform! Open source alternative to ChatGPT Atlas, Perplexity Comet, Dia and Microsoft CoPilot Edge Browser #### [ashishps1/learn-ai-engineering](https://github.com/ashishps1/learn-ai-engineering) *? · ★5,365 · GPL-3.0 · — · score:0.00 · hot:0.45 · rising:0.49 · durable:0.56 · board:durable · trend:stable* Learn AI and LLMs from scratch using free resources #### [vercel/commerce](https://github.com/vercel/commerce) *TypeScript · ★14,003 · MIT · — · score:0.00 · hot:0.45 · rising:0.56 · durable:0.61 · board:durable · trend:stable* Next.js Commerce #### [vercel/next-devtools-mcp](https://github.com/vercel/next-devtools-mcp) *TypeScript · ★725 · no-license · — · score:0.00 · hot:0.45 · rising:0.46 · durable:0.50 · board:durable · trend:stable* Next.js Development for Coding Agent #### [Mirrowel/LLM-API-Key-Proxy](https://github.com/Mirrowel/LLM-API-Key-Proxy) *Python · ★462 · NOASSERTION · — · score:0.00 · hot:0.45 · rising:0.45 · durable:0.45 · board:durable · trend:stable* Universal LLM Gateway: One API, every LLM. OpenAI/Anthropic-compatible endpoints with multi-provider translation and intelligent load-balancing. #### [google-research/metricx](https://github.com/google-research/metricx) *Python · ★139 · Apache-2.0 · — · score:0.00 · hot:0.45 · rising:0.47 · durable:0.39 · board:rising · trend:stable* #### [simonw/datasette-graphql](https://github.com/simonw/datasette-graphql) *Python · ★109 · Apache-2.0 · — · score:0.00 · hot:0.44 · rising:0.46 · durable:0.41 · board:rising · trend:stable* Datasette plugin providing an automatic GraphQL API for your SQLite databases #### [huggingface/gym-hil](https://github.com/huggingface/gym-hil) *Python · ★216 · Apache-2.0 · — · score:0.00 · hot:0.44 · rising:0.48 · durable:0.41 · board:rising · trend:stable* Human in the loop Reinforcement Learning suite #### [jina-ai/clip-as-service](https://github.com/jina-ai/clip-as-service) *Python · ★12,834 · NOASSERTION · — · score:0.00 · hot:0.44 · rising:0.50 · durable:0.54 · board:durable · trend:stable* 🏄 Scalable embedding, reasoning, ranking for images and sentences with CLIP #### [lucidrains/denoising-diffusion-pytorch](https://github.com/lucidrains/denoising-diffusion-pytorch) *Python · ★10,509 · MIT · — · score:0.00 · hot:0.44 · rising:0.52 · durable:0.55 · board:durable · trend:stable* Implementation of Denoising Diffusion Probabilistic Model in Pytorch #### [OpenNMT/CTranslate2](https://github.com/OpenNMT/CTranslate2) *C++ · ★4,444 · MIT · — · score:0.00 · hot:0.44 · rising:0.48 · durable:0.51 · board:durable · trend:stable* Fast inference engine for Transformer models #### [NVIDIA/gds-nvidia-fs](https://github.com/NVIDIA/gds-nvidia-fs) *C · ★342 · NOASSERTION · — · score:0.00 · hot:0.44 · rising:0.47 · durable:0.43 · board:rising · trend:stable* NVIDIA GPUDirect Storage Driver #### [huggingface/inference-benchmarker](https://github.com/huggingface/inference-benchmarker) *Rust · ★151 · Apache-2.0 · — · score:0.00 · hot:0.44 · rising:0.48 · durable:0.45 · board:rising · trend:stable* Inference server benchmarking tool #### [microsoft/ai-edu](https://github.com/microsoft/ai-edu) *HTML · ★14,064 · NOASSERTION · — · score:0.00 · hot:0.44 · rising:0.54 · durable:0.57 · board:durable · trend:stable* AI education materials for Chinese students, teachers and IT professionals. #### [OpenRouterTeam/openrouter-examples](https://github.com/OpenRouterTeam/openrouter-examples) *TypeScript · ★320 · MIT · — · score:0.00 · hot:0.44 · rising:0.44 · durable:0.41 · board:hot · trend:stable* Examples of integrating the OpenRouter API #### [chroma-core/context-1-data-gen](https://github.com/chroma-core/context-1-data-gen) *Python · ★401 · Apache-2.0 · — · score:0.00 · hot:0.44 · rising:0.47 · durable:0.47 · board:rising · trend:stable* #### [huggingface/swift-huggingface](https://github.com/huggingface/swift-huggingface) *Swift · ★154 · Apache-2.0 · — · score:0.00 · hot:0.44 · rising:0.49 · durable:0.50 · board:durable · trend:stable* A Swift client for Hugging Face Hub and Inference Providers APIs #### [NVIDIA/MDL-SDK](https://github.com/NVIDIA/MDL-SDK) *C++ · ★523 · BSD-3-Clause · — · score:0.00 · hot:0.44 · rising:0.51 · durable:0.52 · board:durable · trend:stable* NVIDIA Material Definition Language SDK #### [HisMax/RedInk](https://github.com/HisMax/RedInk) *Python · ★5,173 · NOASSERTION · — · score:0.00 · hot:0.44 · rising:0.49 · durable:0.50 · board:durable · trend:stable* Red Ink - A one-stop Xiaohongshu image-and-text generator based on the 🍌Nano Banana Pro🍌, "One Sentence, One Image: Generate Xiaohongshu Text and Images." #### [qdrant/quaterion](https://github.com/qdrant/quaterion) *Python · ★661 · Apache-2.0 · — · score:0.00 · hot:0.44 · rising:0.48 · durable:0.50 · board:durable · trend:stable* Blazing fast framework for fine-tuning similarity learning models #### [smkalami/prompt-decorators](https://github.com/smkalami/prompt-decorators) *? · ★464 · MIT · — · score:0.00 · hot:0.44 · rising:0.47 · durable:0.47 · board:durable · trend:stable* Prompt Decorators are structured prefixes, such as +++Reasoning and +++StepByStep, designed to enhance AI responses. Inspired by Python decorators, they make AI outputs more logical, accurate, and well-organized without requiring lengthy instructions, simplifying interactions for users. #### [PaddlePaddle/PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg) *Python · ★9,326 · Apache-2.0 · — · score:0.00 · hot:0.44 · rising:0.52 · durable:0.58 · board:durable · trend:stable* Easy-to-use image segmentation library with awesome pre-trained model zoo, supporting wide-range of practical tasks in Semantic Segmentation, Interactive Segmentation, Panoptic Segmentation, Image Matting, 3D Segmentation, etc. #### [OpenGVLab/InternVideo](https://github.com/OpenGVLab/InternVideo) *Python · ★2,245 · Apache-2.0 · — · score:0.00 · hot:0.44 · rising:0.44 · durable:0.44 · board:rising · trend:stable* [ECCV2024] Video Foundation Models & Data for Multimodal Understanding #### [microsoft/ArchScale](https://github.com/microsoft/ArchScale) *Python · ★328 · MIT · — · score:0.00 · hot:0.44 · rising:0.47 · durable:0.46 · board:rising · trend:stable* Simple & Scalable Pretraining for Neural Architecture Research #### [lucidrains/PaLM-rlhf-pytorch](https://github.com/lucidrains/PaLM-rlhf-pytorch) *Python · ★7,871 · MIT · — · score:0.00 · hot:0.44 · rising:0.52 · durable:0.59 · board:durable · trend:stable* Implementation of RLHF (Reinforcement Learning with Human Feedback) on top of the PaLM architecture. Basically ChatGPT but with PaLM #### [langchain-ai/langgraph-fullstack-python](https://github.com/langchain-ai/langgraph-fullstack-python) *Python · ★147 · MIT · — · score:0.00 · hot:0.44 · rising:0.47 · durable:0.45 · board:rising · trend:stable* #### [lm-sys/FastChat](https://github.com/lm-sys/FastChat) *Python · ★39,453 · Apache-2.0 · — · score:0.00 · hot:0.44 · rising:0.53 · durable:0.57 · board:durable · trend:stable* An open platform for training, serving, and evaluating large language models. Release repo for Vicuna and Chatbot Arena. #### [langchain-ai/memory-template](https://github.com/langchain-ai/memory-template) *Python · ★234 · MIT · — · score:0.00 · hot:0.44 · rising:0.47 · durable:0.44 · board:rising · trend:stable* #### [NVIDIA/flownet2-pytorch](https://github.com/NVIDIA/flownet2-pytorch) *Python · ★3,282 · NOASSERTION · — · score:0.00 · hot:0.43 · rising:0.47 · durable:0.40 · board:rising · trend:stable* Pytorch implementation of FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks #### [rasbt/machine-learning-book](https://github.com/rasbt/machine-learning-book) *Jupyter Notebook · ★5,127 · MIT · — · score:0.00 · hot:0.43 · rising:0.52 · durable:0.57 · board:durable · trend:stable* Code Repository for Machine Learning with PyTorch and Scikit-Learn #### [travisvn/awesome-claude-skills](https://github.com/travisvn/awesome-claude-skills) *? · ★11,494 · no-license · — · score:0.00 · hot:0.43 · rising:0.47 · durable:0.46 · board:rising · trend:stable* A curated list of awesome Claude Skills, resources, and tools for customizing Claude AI workflows — particularly Claude Code #### [karpathy/convnetjs](https://github.com/karpathy/convnetjs) *JavaScript · ★11,150 · MIT · — · score:0.00 · hot:0.43 · rising:0.53 · durable:0.56 · board:durable · trend:stable* Deep Learning in Javascript. Train Convolutional Neural Networks (or ordinary ones) in your browser. #### [NVIDIA/Audio2Face-3D-Samples](https://github.com/NVIDIA/Audio2Face-3D-Samples) *Python · ★295 · no-license · — · score:0.00 · hot:0.43 · rising:0.47 · durable:0.48 · board:durable · trend:stable* A service to convert audio to facial blendshapes for lipsyncing and facial performances. #### [microsoft/ailab](https://github.com/microsoft/ailab) *C# · ★7,845 · MIT · — · score:0.00 · hot:0.43 · rising:0.53 · durable:0.57 · board:durable · trend:stable* Experience, Learn and Code the latest breakthrough innovations with Microsoft AI #### [lucidrains/simple-hierarchical-transformer](https://github.com/lucidrains/simple-hierarchical-transformer) *Python · ★227 · MIT · — · score:0.00 · hot:0.43 · rising:0.46 · durable:0.50 · board:durable · trend:down* Experiments around a simple idea for inducing multiple hierarchical predictive model within a GPT #### [AlibabaResearch/AdvancedLiterateMachinery](https://github.com/AlibabaResearch/AdvancedLiterateMachinery) *C++ · ★1,828 · Apache-2.0 · — · score:0.00 · hot:0.43 · rising:0.47 · durable:0.50 · board:durable · trend:stable* A collection of original, innovative ideas and algorithms towards Advanced Literate Machinery. This project is maintained by the OCR Team in the Language Technology Lab, Tongyi Lab, Alibaba Group. #### [run-llama/rags](https://github.com/run-llama/rags) *Python · ★6,533 · MIT · — · score:0.00 · hot:0.43 · rising:0.47 · durable:0.55 · board:durable · trend:stable* Build ChatGPT over your data, all with natural language #### [huggingface/workshops](https://github.com/huggingface/workshops) *Jupyter Notebook · ★153 · Apache-2.0 · — · score:0.00 · hot:0.43 · rising:0.48 · durable:0.45 · board:rising · trend:stable* Materials for workshops on the Hugging Face ecosystem #### [milvus-io/bootcamp](https://github.com/milvus-io/bootcamp) *Jupyter Notebook · ★2,403 · Apache-2.0 · — · score:0.00 · hot:0.43 · rising:0.47 · durable:0.51 · board:durable · trend:stable* Dealing with all unstructured data, such as reverse image search, audio search, molecular search, video analysis, question and answer systems, NLP, etc. #### [lucidrains/DALLE2-pytorch](https://github.com/lucidrains/DALLE2-pytorch) *Python · ★11,316 · MIT · — · score:0.00 · hot:0.43 · rising:0.52 · durable:0.58 · board:durable · trend:stable* Implementation of DALL-E 2, OpenAI's updated text-to-image synthesis neural network, in Pytorch #### [run-llama/workflows-ts](https://github.com/run-llama/workflows-ts) *TypeScript · ★259 · MIT · — · score:0.00 · hot:0.43 · rising:0.46 · durable:0.50 · board:durable · trend:down* 🌊 Simple, event-driven and stream oriented workflow for TypeScript #### [lucidrains/titans-pytorch](https://github.com/lucidrains/titans-pytorch) *Python · ★1,948 · MIT · — · score:0.00 · hot:0.43 · rising:0.49 · durable:0.52 · board:durable · trend:stable* Unofficial implementation of Titans, SOTA memory for transformers, in Pytorch #### [om-ai-lab/VLM-R1](https://github.com/om-ai-lab/VLM-R1) *Python · ★5,941 · Apache-2.0 · — · score:0.00 · hot:0.43 · rising:0.48 · durable:0.54 · board:durable · trend:stable* Solve Visual Understanding with Reinforced VLMs #### [lucidrains/transfusion-pytorch](https://github.com/lucidrains/transfusion-pytorch) *Python · ★1,344 · MIT · — · score:0.00 · hot:0.43 · rising:0.48 · durable:0.54 · board:durable · trend:stable* Pytorch implementation of Transfusion, "Predict the Next Token and Diffuse Images with One Multi-Modal Model", from MetaAI #### [iptag/jimeng-api](https://github.com/iptag/jimeng-api) *TypeScript · ★887 · GPL-3.0 · — · score:0.00 · hot:0.43 · rising:0.48 · durable:0.50 · board:durable · trend:stable* Reverse-engineered the official API for Jimeng/Dreamina’s text-to-image and image-to-image features. Drew inspiration from several experts’ projects and made some tweaks, which significantly improved stability. #### [amusi/CVPR2026-Papers-with-Code](https://github.com/amusi/CVPR2026-Papers-with-Code) *? · ★22,427 · no-license · — · score:0.00 · hot:0.43 · rising:0.49 · durable:0.51 · board:durable · trend:stable* CVPR 2026 论文和开源项目合集 #### [rasbt/deeplearning-models](https://github.com/rasbt/deeplearning-models) *Jupyter Notebook · ★17,469 · MIT · — · score:0.00 · hot:0.43 · rising:0.52 · durable:0.53 · board:durable · trend:stable* A collection of various deep learning architectures, models, and tips #### [run-llama/LlamaIndexTS](https://github.com/run-llama/LlamaIndexTS) *TypeScript · ★3,079 · MIT · — · score:0.00 · hot:0.43 · rising:0.47 · durable:0.54 · board:durable · trend:stable* Data framework for your LLM applications. Focus on server side solution #### [peremartra/Large-Language-Model-Notebooks-Course](https://github.com/peremartra/Large-Language-Model-Notebooks-Course) *Jupyter Notebook · ★1,792 · MIT · — · score:0.00 · hot:0.43 · rising:0.46 · durable:0.52 · board:durable · trend:stable* Practical course about Large Language Models. #### [google-research/skai](https://github.com/google-research/skai) *Python · ★157 · Apache-2.0 · — · score:0.00 · hot:0.42 · rising:0.47 · durable:0.44 · board:rising · trend:stable* SKAI is a machine learning based tool for performing automatic building damage assessments on aerial imagery of disaster sites. #### [karpathy/nn-zero-to-hero](https://github.com/karpathy/nn-zero-to-hero) *Jupyter Notebook · ★21,487 · MIT · — · score:0.00 · hot:0.42 · rising:0.52 · durable:0.54 · board:durable · trend:stable* Neural Networks: Zero to Hero #### [potamides/DeTikZify](https://github.com/potamides/DeTikZify) *Python · ★1,761 · Apache-2.0 · — · score:0.00 · hot:0.42 · rising:0.45 · durable:0.56 · board:durable · trend:down* Synthesizing Graphics Programs for Scientific Figures and Sketches with TikZ. #### [yoheinakajima/babyagi](https://github.com/yoheinakajima/babyagi) *Python · ★22,232 · no-license · — · score:0.00 · hot:0.42 · rising:0.51 · durable:0.52 · board:durable · trend:stable* #### [langchain-ai/context_engineering](https://github.com/langchain-ai/context_engineering) *Jupyter Notebook · ★174 · no-license · — · score:0.00 · hot:0.42 · rising:0.45 · durable:0.43 · board:rising · trend:stable* #### [karpathy/llm-council](https://github.com/karpathy/llm-council) *Python · ★17,289 · no-license · — · score:0.00 · hot:0.42 · rising:0.50 · durable:0.52 · board:durable · trend:stable* LLM Council works together to answer your hardest questions #### [allenai/satlas](https://github.com/allenai/satlas) *Python · ★273 · Apache-2.0 · — · score:0.00 · hot:0.42 · rising:0.46 · durable:0.44 · board:rising · trend:stable* #### [imbue-bit/AlphaGPT](https://github.com/imbue-bit/AlphaGPT) *Python · ★1,970 · Apache-2.0 · — · score:0.00 · hot:0.42 · rising:0.50 · durable:0.53 · board:durable · trend:stable* 使用符号回归在中国股市与加密市场上进行高效因子挖掘。 #### [simonw/showboat](https://github.com/simonw/showboat) *Go · ★1,052 · Apache-2.0 · — · score:0.00 · hot:0.42 · rising:0.45 · durable:0.51 · board:durable · trend:down* Create executable documents that demonstrate an agent's work #### [zjunlp/LightThinker](https://github.com/zjunlp/LightThinker) *Python · ★152 · MIT · — · score:0.00 · hot:0.42 · rising:0.42 · durable:0.43 · board:durable · trend:down* [EMNLP 2025] LightThinker: Thinking Step-by-Step Compression #### [microsoft/call-center-ai](https://github.com/microsoft/call-center-ai) *Python · ★6,442 · Apache-2.0 · — · score:0.00 · hot:0.42 · rising:0.51 · durable:0.59 · board:durable · trend:stable* Send a phone call from AI agent, in an API call. Or, directly call the bot from the configured phone number! #### [swyxio/ai-notes](https://github.com/swyxio/ai-notes) *HTML · ★6,203 · MIT · — · score:0.00 · hot:0.42 · rising:0.48 · durable:0.53 · board:durable · trend:stable* notes for software engineers getting up to speed on new AI developments. Serves as datastore for https://latent.space writing, and product brainstorming, but has cleaned up canonical references under the /Resources folder. #### [dair-ai/ML-YouTube-Courses](https://github.com/dair-ai/ML-YouTube-Courses) *? · ★17,153 · CC0-1.0 · — · score:0.00 · hot:0.42 · rising:0.51 · durable:0.53 · board:durable · trend:stable* 📺 Discover the latest machine learning / AI courses on YouTube. #### [CyberAlbSecOP/Awesome_GPT_Super_Prompting](https://github.com/CyberAlbSecOP/Awesome_GPT_Super_Prompting) *HTML · ★3,836 · GPL-3.0 · — · score:0.00 · hot:0.42 · rising:0.48 · durable:0.53 · board:durable · trend:stable* ChatGPT Jailbreaks, GPT Assistants Prompt Leaks, GPTs Prompt Injection, LLM Prompt Security, Super Prompts, Prompt Hack, Prompt Security, Ai Prompt Engineering, Adversarial Machine Learning. #### [lucidrains/reformer-pytorch](https://github.com/lucidrains/reformer-pytorch) *Python · ★2,188 · MIT · — · score:0.00 · hot:0.42 · rising:0.47 · durable:0.54 · board:durable · trend:stable* Reformer, the efficient Transformer, in Pytorch #### [simonw/llm-prices](https://github.com/simonw/llm-prices) *HTML · ★127 · no-license · — · score:0.00 · hot:0.42 · rising:0.41 · durable:0.34 · board:hot · trend:stable* Prices of various LLMs #### [karpathy/minGPT](https://github.com/karpathy/minGPT) *Python · ★24,186 · MIT · — · score:0.00 · hot:0.42 · rising:0.51 · durable:0.53 · board:durable · trend:stable* A minimal PyTorch re-implementation of the OpenAI GPT (Generative Pretrained Transformer) training #### [microsoft/PIKE-RAG](https://github.com/microsoft/PIKE-RAG) *Python · ★2,383 · MIT · — · score:0.00 · hot:0.42 · rising:0.49 · durable:0.58 · board:durable · trend:stable* PIKE-RAG: sPecIalized KnowledgE and Rationale Augmented Generation #### [grab/cursor-talk-to-figma-mcp](https://github.com/grab/cursor-talk-to-figma-mcp) *JavaScript · ★6,664 · MIT · — · score:0.00 · hot:0.41 · rising:0.46 · durable:0.52 · board:durable · trend:stable* TalkToFigma: MCP integration between AI Agent (Cursor, Claude Code) and Figma, allowing Agentic AI to communicate with Figma for reading designs and modifying them programmatically. #### [microsoft/vscode-ai-toolkit](https://github.com/microsoft/vscode-ai-toolkit) *? · ★1,951 · MIT · — · score:0.00 · hot:0.41 · rising:0.46 · durable:0.44 · board:rising · trend:stable* #### [lucidrains/imagen-pytorch](https://github.com/lucidrains/imagen-pytorch) *Python · ★8,404 · MIT · — · score:0.00 · hot:0.41 · rising:0.50 · durable:0.56 · board:durable · trend:stable* Implementation of Imagen, Google's Text-to-Image Neural Network, in Pytorch #### [huggingface/data-is-better-together](https://github.com/huggingface/data-is-better-together) *Jupyter Notebook · ★271 · no-license · — · score:0.00 · hot:0.41 · rising:0.44 · durable:0.42 · board:rising · trend:stable* Let's build better datasets, together! #### [BIGPPWONG/EdgeBox](https://github.com/BIGPPWONG/EdgeBox) *TypeScript · ★198 · GPL-3.0 · — · score:0.00 · hot:0.41 · rising:0.44 · durable:0.53 · board:durable · trend:down* A fully-featured, GUI-powered local LLM Agent sandbox with complete MCP protocol support. Features both CLI and full desktop environment, enabling AI agents to operate browsers, terminal, and other desktop applications just like humans. Based on E2B oss code. #### [langgptai/LangGPT](https://github.com/langgptai/LangGPT) *Jupyter Notebook · ★11,943 · Apache-2.0 · — · score:0.00 · hot:0.41 · rising:0.49 · durable:0.54 · board:durable · trend:stable* LangGPT: Empowering everyone to become a prompt expert! 🚀 📌 结构化提示词（Structured Prompt）提出者 📌 元提示词（Meta-Prompt）发起者 📌 最流行的提示词落地范式 | Language of GPT The pioneering framework for structured & meta-prompt design 10,000+ ⭐ | Battle-tested by thousands of users worldwide Created by 云中江树 #### [NVIDIA/FasterTransformer](https://github.com/NVIDIA/FasterTransformer) *C++ · ★6,412 · Apache-2.0 · — · score:0.00 · hot:0.41 · rising:0.48 · durable:0.54 · board:durable · trend:stable* Transformer related optimization, including BERT, GPT #### [OpenSPG/KAG](https://github.com/OpenSPG/KAG) *Python · ★8,682 · Apache-2.0 · — · score:0.00 · hot:0.41 · rising:0.50 · durable:0.55 · board:durable · trend:stable* KAG is a logical form-guided reasoning and retrieval framework based on OpenSPG engine and LLMs. It is used to build logical reasoning and factual Q&A solutions for professional domain knowledge bases. It can effectively overcome the shortcomings of the traditional RAG vector similarity calculation model. #### [timeseriesAI/tsai](https://github.com/timeseriesAI/tsai) *Jupyter Notebook · ★6,036 · Apache-2.0 · — · score:0.00 · hot:0.41 · rising:0.47 · durable:0.54 · board:durable · trend:stable* Time series Timeseries Deep Learning Machine Learning Python Pytorch fastai | State-of-the-art Deep Learning library for Time Series and Sequences in Pytorch / fastai #### [karpathy/llama2.c](https://github.com/karpathy/llama2.c) *C · ★19,421 · MIT · — · score:0.00 · hot:0.41 · rising:0.50 · durable:0.53 · board:durable · trend:stable* Inference Llama 2 in one file of pure C #### [qdrant/qdrant_demo](https://github.com/qdrant/qdrant_demo) *TypeScript · ★189 · no-license · — · score:0.00 · hot:0.41 · rising:0.43 · durable:0.40 · board:rising · trend:stable* Demo of the neural semantic search built with Qdrant #### [facebookresearch/nevergrad](https://github.com/facebookresearch/nevergrad) *Python · ★4,177 · MIT · — · score:0.00 · hot:0.41 · rising:0.48 · durable:0.50 · board:durable · trend:stable* A Python toolbox for performing gradient-free optimization #### [jina-ai/discoart](https://github.com/jina-ai/discoart) *Python · ★3,831 · NOASSERTION · — · score:0.00 · hot:0.41 · rising:0.47 · durable:0.54 · board:durable · trend:stable* 🪩 Create Disco Diffusion artworks in one line #### [microsoft/api-guidelines](https://github.com/microsoft/api-guidelines) *? · ★23,255 · NOASSERTION · — · score:0.00 · hot:0.41 · rising:0.49 · durable:0.49 · board:rising · trend:stable* Microsoft REST API Guidelines #### [jina-ai/node-DeepResearch](https://github.com/jina-ai/node-DeepResearch) *TypeScript · ★5,147 · Apache-2.0 · — · score:0.00 · hot:0.41 · rising:0.50 · durable:0.58 · board:durable · trend:stable* Keep searching, reading webpages, reasoning until it finds the answer (or exceeding the token budget) #### [NVIDIA/nvapi](https://github.com/NVIDIA/nvapi) *C · ★207 · NOASSERTION · — · score:0.00 · hot:0.41 · rising:0.43 · durable:0.39 · board:rising · trend:stable* NVAPI is NVIDIA's core software development kit that allows direct access to NVIDIA GPUs and drivers on supported platforms. #### [NVIDIA/MinkowskiEngine](https://github.com/NVIDIA/MinkowskiEngine) *Python · ★2,909 · NOASSERTION · — · score:0.00 · hot:0.41 · rising:0.45 · durable:0.49 · board:durable · trend:stable* Minkowski Engine is an auto-diff neural network library for high-dimensional sparse tensors #### [anthropics/healthcare](https://github.com/anthropics/healthcare) *Python · ★203 · no-license · — · score:0.00 · hot:0.41 · rising:0.46 · durable:0.48 · board:durable · trend:down* #### [microsoft/mattergen](https://github.com/microsoft/mattergen) *Python · ★1,686 · MIT · — · score:0.00 · hot:0.41 · rising:0.50 · durable:0.56 · board:durable · trend:stable* Official implementation of MatterGen -- a generative model for inorganic materials design across the periodic table that can be fine-tuned to steer the generation towards a wide range of property constraints. #### [karpathy/micrograd](https://github.com/karpathy/micrograd) *Jupyter Notebook · ★15,522 · MIT · — · score:0.00 · hot:0.41 · rising:0.49 · durable:0.52 · board:durable · trend:stable* A tiny scalar-valued autograd engine and a neural net library on top of it with PyTorch-like API #### [vercel/fun](https://github.com/vercel/fun) *TypeScript · ★982 · Apache-2.0 · — · score:0.00 · hot:0.41 · rising:0.46 · durable:0.49 · board:durable · trend:stable* ƒun - Local serverless function λ development runtime #### [huggingface/screensuite](https://github.com/huggingface/screensuite) *Python · ★143 · Apache-2.0 · — · score:0.00 · hot:0.41 · rising:0.44 · durable:0.44 · board:durable · trend:down* ScreenSuite - The most comprehensive benchmarking suite for GUI Agents! #### [microsoft/referencesource](https://github.com/microsoft/referencesource) *C# · ★3,392 · MIT · — · score:0.00 · hot:0.41 · rising:0.50 · durable:0.51 · board:durable · trend:stable* Source from the Microsoft .NET Reference Source that represent a subset of the .NET Framework #### [microsoft/Security-101](https://github.com/microsoft/Security-101) *HTML · ★6,382 · CC0-1.0 · — · score:0.00 · hot:0.41 · rising:0.50 · durable:0.54 · board:durable · trend:stable* 8 Lessons, Kick-start Your Cybersecurity Learning. #### [Shilin-LU/MACE](https://github.com/Shilin-LU/MACE) *Jupyter Notebook · ★398 · NOASSERTION · — · score:0.00 · hot:0.41 · rising:0.41 · durable:0.35 · board:rising · trend:stable* [CVPR 2024] "MACE: Mass Concept Erasure in Diffusion Models" (Official Implementation) #### [NVIDIA/soma-retargeter](https://github.com/NVIDIA/soma-retargeter) *Python · ★278 · Apache-2.0 · — · score:0.00 · hot:0.41 · rising:0.44 · durable:0.45 · board:durable · trend:down* SOMA BVH to humanoid robot motion retargeting library built with Newton and NVIDIA Warp #### [lucidrains/musiclm-pytorch](https://github.com/lucidrains/musiclm-pytorch) *Python · ★3,292 · MIT · — · score:0.00 · hot:0.40 · rising:0.48 · durable:0.56 · board:durable · trend:stable* Implementation of MusicLM, Google's new SOTA model for music generation using attention networks, in Pytorch #### [langchain-ai/rag-from-scratch](https://github.com/langchain-ai/rag-from-scratch) *Jupyter Notebook · ★8,061 · no-license · — · score:0.00 · hot:0.40 · rising:0.48 · durable:0.51 · board:durable · trend:stable* #### [vercel/serve-handler](https://github.com/vercel/serve-handler) *JavaScript · ★617 · MIT · — · score:0.00 · hot:0.40 · rising:0.47 · durable:0.46 · board:rising · trend:stable* The foundation of `serve` #### [lucidrains/alphafold3-pytorch](https://github.com/lucidrains/alphafold3-pytorch) *Python · ★1,636 · MIT · — · score:0.00 · hot:0.40 · rising:0.48 · durable:0.55 · board:durable · trend:stable* Implementation of Alphafold 3 from Google Deepmind in Pytorch #### [facebookresearch/repoprover](https://github.com/facebookresearch/repoprover) *Python · ★135 · NOASSERTION · — · score:0.00 · hot:0.40 · rising:0.41 · durable:0.42 · board:durable · trend:down* Research code base for Automatic Textbook Formalization #### [lucidrains/make-a-video-pytorch](https://github.com/lucidrains/make-a-video-pytorch) *Python · ★1,991 · MIT · — · score:0.00 · hot:0.40 · rising:0.49 · durable:0.55 · board:durable · trend:stable* Implementation of Make-A-Video, new SOTA text to video generator from Meta AI, in Pytorch #### [microsoft/dotnet](https://github.com/microsoft/dotnet) *HTML · ★15,029 · MIT · — · score:0.00 · hot:0.40 · rising:0.50 · durable:0.49 · board:rising · trend:stable* This repo is the official home of .NET on GitHub. It's a great starting point to find many .NET OSS projects from Microsoft and the community, including many that are part of the .NET Foundation. #### [karpathy/jobs](https://github.com/karpathy/jobs) *HTML · ★1,495 · no-license · — · score:0.00 · hot:0.40 · rising:0.45 · durable:0.45 · board:durable · trend:stable* A research tool for visually exploring Bureau of Labor Statistics Occupational Outlook Handbook data. This is not a report, a paper, or a serious economic publication — it is a development tool for exploring BLS data visually. #### [dair-ai/ML-Course-Notes](https://github.com/dair-ai/ML-Course-Notes) *? · ★6,446 · NOASSERTION · — · score:0.00 · hot:0.40 · rising:0.47 · durable:0.49 · board:durable · trend:stable* 🎓 Sharing machine learning course / lecture notes. #### [lucidrains/pi-zero-pytorch](https://github.com/lucidrains/pi-zero-pytorch) *Python · ★569 · MIT · — · score:0.00 · hot:0.40 · rising:0.45 · durable:0.53 · board:durable · trend:down* Implementation of π₀, the robotic foundation model architecture proposed by Physical Intelligence #### [modal-labs/gpu-glossary](https://github.com/modal-labs/gpu-glossary) *Python · ★563 · MIT · — · score:0.00 · hot:0.40 · rising:0.45 · durable:0.43 · board:rising · trend:stable* GPU documentation for humans #### [dair-ai/ML-Papers-Explained](https://github.com/dair-ai/ML-Papers-Explained) *? · ★8,548 · no-license · — · score:0.00 · hot:0.40 · rising:0.47 · durable:0.50 · board:durable · trend:stable* Explanation to key concepts in ML #### [lucidrains/CoCa-pytorch](https://github.com/lucidrains/CoCa-pytorch) *Python · ★1,198 · MIT · — · score:0.00 · hot:0.40 · rising:0.45 · durable:0.54 · board:durable · trend:down* Implementation of CoCa, Contrastive Captioners are Image-Text Foundation Models, in Pytorch #### [huggingface/jat](https://github.com/huggingface/jat) *Python · ★186 · Apache-2.0 · — · score:0.00 · hot:0.40 · rising:0.44 · durable:0.44 · board:rising · trend:stable* General multi-task deep RL Agent #### [karpathy/char-rnn](https://github.com/karpathy/char-rnn) *Lua · ★12,028 · no-license · — · score:0.00 · hot:0.40 · rising:0.48 · durable:0.46 · board:rising · trend:stable* Multi-layer Recurrent Neural Networks (LSTM, GRU, RNN) for character-level language models in Torch #### [rasbt/LLM-workshop-2024](https://github.com/rasbt/LLM-workshop-2024) *Jupyter Notebook · ★1,085 · Apache-2.0 · — · score:0.00 · hot:0.40 · rising:0.46 · durable:0.50 · board:durable · trend:stable* A 4-hour coding workshop to understand how LLMs are implemented and used #### [rasbt/pattern_classification](https://github.com/rasbt/pattern_classification) *Jupyter Notebook · ★4,213 · GPL-3.0 · — · score:0.00 · hot:0.40 · rising:0.49 · durable:0.49 · board:rising · trend:stable* A collection of tutorials and examples for solving and understanding machine learning and pattern classification tasks #### [dair-ai/ml-visuals](https://github.com/dair-ai/ml-visuals) *? · ★17,098 · MIT · — · score:0.00 · hot:0.40 · rising:0.48 · durable:0.52 · board:durable · trend:stable* 🎨 ML Visuals contains figures and templates which you can reuse and customize to improve your scientific writing. #### [adithya-s-k/AI-Engineering.academy](https://github.com/adithya-s-k/AI-Engineering.academy) *Jupyter Notebook · ★2,182 · MIT · — · score:0.00 · hot:0.40 · rising:0.44 · durable:0.51 · board:durable · trend:stable* Mastering Applied AI, One Concept at a Time #### [vercel/cosmosdb-server](https://github.com/vercel/cosmosdb-server) *TypeScript · ★178 · MIT · — · score:0.00 · hot:0.40 · rising:0.45 · durable:0.43 · board:rising · trend:stable* A Cosmos DB server implementation for testing your applications locally. #### [simonw/rodney](https://github.com/simonw/rodney) *Go · ★674 · Apache-2.0 · — · score:0.00 · hot:0.40 · rising:0.44 · durable:0.46 · board:durable · trend:down* CLI tool for interacting with the web #### [vercel/platforms](https://github.com/vercel/platforms) *TypeScript · ★6,665 · no-license · — · score:0.00 · hot:0.40 · rising:0.48 · durable:0.50 · board:durable · trend:stable* A full-stack Next.js app with multi-tenancy. #### [rasbt/python-machine-learning-book-3rd-edition](https://github.com/rasbt/python-machine-learning-book-3rd-edition) *Jupyter Notebook · ★5,012 · MIT · — · score:0.00 · hot:0.40 · rising:0.49 · durable:0.49 · board:durable · trend:stable* The "Python Machine Learning (3rd edition)" book code repository #### [Hunyuan-PromptEnhancer/PromptEnhancer](https://github.com/Hunyuan-PromptEnhancer/PromptEnhancer) *Python · ★3,667 · NOASSERTION · — · score:0.00 · hot:0.40 · rising:0.47 · durable:0.49 · board:durable · trend:stable* [CVPR 2026] PromptEnhancer is a prompt-rewriting tool, refining prompts into clearer, structured versions for better image generation. #### [dair-ai/ML-Notebooks](https://github.com/dair-ai/ML-Notebooks) *Jupyter Notebook · ★3,439 · Apache-2.0 · — · score:0.00 · hot:0.40 · rising:0.47 · durable:0.51 · board:durable · trend:stable* :fire: Machine Learning Notebooks #### [lucidrains/DALLE-pytorch](https://github.com/lucidrains/DALLE-pytorch) *Python · ★5,624 · MIT · — · score:0.00 · hot:0.40 · rising:0.47 · durable:0.53 · board:durable · trend:stable* Implementation / replication of DALL-E, OpenAI's Text to Image Transformer, in Pytorch #### [lucidrains/tab-transformer-pytorch](https://github.com/lucidrains/tab-transformer-pytorch) *Python · ★1,073 · MIT · — · score:0.00 · hot:0.40 · rising:0.45 · durable:0.51 · board:durable · trend:down* Implementation of TabTransformer, attention network for tabular data, in Pytorch #### [karpathy/minbpe](https://github.com/karpathy/minbpe) *Python · ★10,438 · MIT · — · score:0.00 · hot:0.40 · rising:0.48 · durable:0.52 · board:durable · trend:stable* Minimal, clean code for the Byte Pair Encoding (BPE) algorithm commonly used in LLM tokenization. #### [milvus-io/milvus-sdk-csharp](https://github.com/milvus-io/milvus-sdk-csharp) *C# · ★138 · Apache-2.0 · — · score:0.00 · hot:0.40 · rising:0.45 · durable:0.45 · board:rising · trend:stable* C# SDK for Milvus. #### [lucidrains/mimic-video](https://github.com/lucidrains/mimic-video) *Python · ★104 · MIT · — · score:0.00 · hot:0.40 · rising:0.43 · durable:0.50 · board:durable · trend:down* Implementation of Mimic-Video, Video-Action Models for SOTA Generalizable Robot Control Beyond VLAs #### [vercel/ncc](https://github.com/vercel/ncc) *JavaScript · ★9,806 · MIT · — · score:0.00 · hot:0.40 · rising:0.45 · durable:0.53 · board:durable · trend:stable* Compile a Node.js project into a single file. Supports TypeScript, binary addons, dynamic requires. #### [NVIDIA/deepops](https://github.com/NVIDIA/deepops) *Shell · ★1,430 · BSD-3-Clause · — · score:0.00 · hot:0.40 · rising:0.49 · durable:0.55 · board:durable · trend:stable* Tools for building GPU clusters #### [oramasearch/orama](https://github.com/oramasearch/orama) *TypeScript · ★10,301 · NOASSERTION · — · score:0.00 · hot:0.40 · rising:0.45 · durable:0.56 · board:durable · trend:down* 🌌 A complete search engine and RAG pipeline in your browser, server or edge network with support for full-text, vector, and hybrid search in less than 2kb. #### [weaviate/elysia](https://github.com/weaviate/elysia) *Python · ★1,888 · BSD-3-Clause · — · score:0.00 · hot:0.40 · rising:0.49 · durable:0.57 · board:durable · trend:down* Python package and backend for the Elysia platform app. #### [lucidrains/gigagan-pytorch](https://github.com/lucidrains/gigagan-pytorch) *Python · ★1,941 · MIT · — · score:0.00 · hot:0.40 · rising:0.47 · durable:0.53 · board:durable · trend:stable* Implementation of GigaGAN, new SOTA GAN out of Adobe. Culmination of nearly a decade of research into GANs #### [NVIDIA/BigVGAN](https://github.com/NVIDIA/BigVGAN) *Python · ★1,206 · MIT · — · score:0.00 · hot:0.40 · rising:0.49 · durable:0.56 · board:durable · trend:stable* Official PyTorch implementation of BigVGAN (ICLR 2023) #### [run-llama/notebookllama](https://github.com/run-llama/notebookllama) *Python · ★1,868 · MIT · — · score:0.00 · hot:0.40 · rising:0.48 · durable:0.57 · board:durable · trend:down* A fully open-source, LlamaCloud-backed alternative to NotebookLM #### [wladradchenko/wunjo.wladradchenko.ru](https://github.com/wladradchenko/wunjo.wladradchenko.ru) *JavaScript · ★1,137 · MIT · — · score:0.00 · hot:0.40 · rising:0.47 · durable:0.54 · board:durable · trend:down* Wunjo CE: Face Swap, Lip Sync, Control Remove Objects & Text & Background, Restyling, Audio Separator, Clone Voice, Video Generation. Open Source, Local & Free. #### [sgl-project/mini-sglang](https://github.com/sgl-project/mini-sglang) *Python · ★4,027 · MIT · — · score:0.00 · hot:0.40 · rising:0.47 · durable:0.51 · board:durable · trend:stable* A compact implementation of SGLang, designed to demystify the complexities of modern LLM serving systems. #### [vercel/ms](https://github.com/vercel/ms) *TypeScript · ★5,516 · MIT · — · score:0.00 · hot:0.40 · rising:0.48 · durable:0.54 · board:durable · trend:stable* Tiny millisecond conversion utility #### [lucidrains/deep-daze](https://github.com/lucidrains/deep-daze) *Python · ★4,324 · MIT · — · score:0.00 · hot:0.40 · rising:0.46 · durable:0.52 · board:durable · trend:stable* Simple command line tool for text to image generation using OpenAI's CLIP and Siren (Implicit neural representation network). Technique was originally created by https://twitter.com/advadnoun #### [vercel/micro](https://github.com/vercel/micro) *TypeScript · ★10,611 · MIT · — · score:0.00 · hot:0.40 · rising:0.48 · durable:0.57 · board:durable · trend:stable* Asynchronous HTTP microservices #### [RichardHruby/login-machine](https://github.com/RichardHruby/login-machine) *TypeScript · ★288 · MIT · — · score:0.00 · hot:0.39 · rising:0.44 · durable:0.47 · board:durable · trend:down* AI-powered login automation. Uses Claude to classify login pages and Playwright to interact with them. #### [dair-ai/Mathematics-for-ML](https://github.com/dair-ai/Mathematics-for-ML) *? · ★5,929 · no-license · — · score:0.00 · hot:0.39 · rising:0.46 · durable:0.48 · board:durable · trend:stable* 🧮 A collection of resources to learn mathematics for machine learning #### [simonw/claude-code-transcripts](https://github.com/simonw/claude-code-transcripts) *Python · ★1,451 · Apache-2.0 · — · score:0.00 · hot:0.39 · rising:0.45 · durable:0.49 · board:durable · trend:stable* Tools for publishing transcripts for Claude Code sessions #### [simonw/files-to-prompt](https://github.com/simonw/files-to-prompt) *Python · ★2,646 · Apache-2.0 · — · score:0.00 · hot:0.39 · rising:0.47 · durable:0.54 · board:durable · trend:down* Concatenate a directory full of files into a single prompt for use with LLMs #### [simonw/til](https://github.com/simonw/til) *HTML · ★1,408 · Apache-2.0 · — · score:0.00 · hot:0.39 · rising:0.44 · durable:0.41 · board:rising · trend:stable* Today I Learned #### [rasbt/python_reference](https://github.com/rasbt/python_reference) *Jupyter Notebook · ★3,887 · no-license · — · score:0.00 · hot:0.39 · rising:0.46 · durable:0.45 · board:rising · trend:stable* Useful functions, tutorials, and other Python-related things #### [lucidrains/muse-maskgit-pytorch](https://github.com/lucidrains/muse-maskgit-pytorch) *Python · ★920 · MIT · — · score:0.00 · hot:0.39 · rising:0.46 · durable:0.52 · board:durable · trend:down* Implementation of Muse: Text-to-Image Generation via Masked Generative Transformers, in Pytorch #### [lucidrains/toolformer-pytorch](https://github.com/lucidrains/toolformer-pytorch) *Python · ★2,057 · MIT · — · score:0.00 · hot:0.39 · rising:0.46 · durable:0.55 · board:durable · trend:down* Implementation of Toolformer, Language Models That Can Use Tools, by MetaAI #### [karpathy/ng-video-lecture](https://github.com/karpathy/ng-video-lecture) *Python · ★4,652 · no-license · — · score:0.00 · hot:0.39 · rising:0.46 · durable:0.47 · board:durable · trend:stable* #### [facebookresearch/pytorchvideo](https://github.com/facebookresearch/pytorchvideo) *Python · ★3,554 · Apache-2.0 · — · score:0.00 · hot:0.39 · rising:0.48 · durable:0.51 · board:durable · trend:stable* A deep learning library for video understanding research. #### [lucidrains/naturalspeech2-pytorch](https://github.com/lucidrains/naturalspeech2-pytorch) *Python · ★1,333 · MIT · — · score:0.00 · hot:0.39 · rising:0.46 · durable:0.53 · board:durable · trend:down* Implementation of Natural Speech 2, Zero-shot Speech and Singing Synthesizer, in Pytorch #### [allenai/ai2-scholarqa-lib](https://github.com/allenai/ai2-scholarqa-lib) *Python · ★274 · Apache-2.0 · — · score:0.00 · hot:0.39 · rising:0.44 · durable:0.45 · board:durable · trend:stable* Repo housing the open sourced code for the ai2 scholar qa app and also the corresponding library #### [lucidrains/lambda-networks](https://github.com/lucidrains/lambda-networks) *Python · ★1,528 · MIT · — · score:0.00 · hot:0.39 · rising:0.47 · durable:0.52 · board:durable · trend:stable* Implementation of LambdaNetworks, a new approach to image recognition that reaches SOTA with less compute #### [lucidrains/alphafold2](https://github.com/lucidrains/alphafold2) *Python · ★1,632 · MIT · — · score:0.00 · hot:0.39 · rising:0.47 · durable:0.52 · board:durable · trend:stable* To eventually become an unofficial Pytorch implementation / replication of Alphafold2, as details of the architecture get released #### [karpathy/neuraltalk](https://github.com/karpathy/neuraltalk) *Python · ★5,489 · no-license · — · score:0.00 · hot:0.39 · rising:0.46 · durable:0.45 · board:rising · trend:stable* NeuralTalk is a Python+numpy project for learning Multimodal Recurrent Neural Networks that describe images with sentences. #### [microsoft/pict](https://github.com/microsoft/pict) *C++ · ★1,425 · NOASSERTION · — · score:0.00 · hot:0.39 · rising:0.46 · durable:0.48 · board:durable · trend:stable* Pairwise Independent Combinatorial Tool #### [karpathy/build-nanogpt](https://github.com/karpathy/build-nanogpt) *Python · ★4,937 · no-license · — · score:0.00 · hot:0.39 · rising:0.46 · durable:0.48 · board:durable · trend:stable* Video+code lecture on building nanoGPT from scratch #### [openai/gym3](https://github.com/openai/gym3) *Python · ★146 · MIT · — · score:0.00 · hot:0.39 · rising:0.45 · durable:0.43 · board:rising · trend:stable* Vectorized interface for reinforcement learning environments #### [microsoft/SEAL](https://github.com/microsoft/SEAL) *C++ · ★3,966 · MIT · — · score:0.00 · hot:0.39 · rising:0.48 · durable:0.50 · board:durable · trend:stable* Microsoft SEAL is an easy-to-use and powerful homomorphic encryption library. #### [SamurAIGPT/AI-Youtube-Shorts-Generator](https://github.com/SamurAIGPT/AI-Youtube-Shorts-Generator) *Python · ★3,225 · MIT · — · score:0.00 · hot:0.39 · rising:0.47 · durable:0.50 · board:durable · trend:stable* A python tool that uses GPT-4, FFmpeg, and OpenCV to automatically analyze videos, extract the most interesting sections, and crop them for an improved viewing experience. #### [vibesurf-ai/VibeSurf](https://github.com/vibesurf-ai/VibeSurf) *Python · ★481 · NOASSERTION · — · score:0.00 · hot:0.39 · rising:0.43 · durable:0.50 · board:durable · trend:down* A powerful browser assistant for vibe surfing 一个开源的AI浏览器智能助手 #### [lucidrains/audiolm-pytorch](https://github.com/lucidrains/audiolm-pytorch) *Python · ★2,619 · MIT · — · score:0.00 · hot:0.39 · rising:0.46 · durable:0.52 · board:durable · trend:stable* Implementation of AudioLM, a SOTA Language Modeling Approach to Audio Generation out of Google Research, in Pytorch #### [Yutong-Zhou-cv/Awesome-Text-to-Image](https://github.com/Yutong-Zhou-cv/Awesome-Text-to-Image) *? · ★2,436 · MIT · — · score:0.00 · hot:0.39 · rising:0.45 · durable:0.49 · board:durable · trend:stable* (ෆ`꒳´ෆ) A Survey on Text-to-Image Generation/Synthesis. #### [simonw/shot-scraper](https://github.com/simonw/shot-scraper) *Python · ★2,328 · Apache-2.0 · — · score:0.00 · hot:0.39 · rising:0.44 · durable:0.49 · board:durable · trend:down* A command-line utility for taking automated screenshots of websites #### [rasbt/mlxtend](https://github.com/rasbt/mlxtend) *Python · ★5,132 · NOASSERTION · — · score:0.00 · hot:0.39 · rising:0.45 · durable:0.46 · board:durable · trend:stable* A library of extension and helper modules for Python's data analysis and machine learning libraries. #### [microsoft/BatteryML](https://github.com/microsoft/BatteryML) *Jupyter Notebook · ★732 · MIT · — · score:0.00 · hot:0.39 · rising:0.47 · durable:0.53 · board:durable · trend:down* #### [microsoft/evodiff](https://github.com/microsoft/evodiff) *Python · ★666 · MIT · — · score:0.00 · hot:0.39 · rising:0.46 · durable:0.50 · board:durable · trend:stable* Generation of protein sequences and evolutionary alignments via discrete diffusion models #### [lucidrains/BS-RoFormer](https://github.com/lucidrains/BS-RoFormer) *Python · ★777 · MIT · — · score:0.00 · hot:0.39 · rising:0.43 · durable:0.50 · board:durable · trend:down* Implementation of Band Split Roformer, SOTA Attention network for music source separation out of ByteDance AI Labs #### [karpathy/makemore](https://github.com/karpathy/makemore) *Python · ★3,841 · MIT · — · score:0.00 · hot:0.39 · rising:0.46 · durable:0.49 · board:durable · trend:stable* An autoregressive character-level language model for making more things #### [weaviate/elysia-frontend](https://github.com/weaviate/elysia-frontend) *TypeScript · ★175 · MIT · — · score:0.00 · hot:0.39 · rising:0.46 · durable:0.54 · board:durable · trend:down* Frontend Repository for Elysia #### [AIStream-Peelout/flow-forecast](https://github.com/AIStream-Peelout/flow-forecast) *Python · ★2,277 · GPL-3.0 · — · score:0.00 · hot:0.39 · rising:0.44 · durable:0.49 · board:durable · trend:stable* Deep learning PyTorch library for time series forecasting, classification, and anomaly detection (originally for flood forecasting). #### [sanbuphy/nanoAgent](https://github.com/sanbuphy/nanoAgent) *Python · ★553 · MIT · — · score:0.00 · hot:0.39 · rising:0.42 · durable:0.47 · board:durable · trend:down* If you can read ~100 lines of Python, you understand agents. #### [lucidrains/stylegan2-pytorch](https://github.com/lucidrains/stylegan2-pytorch) *Python · ★3,786 · MIT · — · score:0.00 · hot:0.39 · rising:0.46 · durable:0.50 · board:durable · trend:stable* Simplest working implementation of Stylegan2, state of the art generative adversarial network, in Pytorch. Enabling everyone to experience disentanglement #### [facebookresearch/SustainableConcrete](https://github.com/facebookresearch/SustainableConcrete) *Jupyter Notebook · ★178 · MIT · — · score:0.00 · hot:0.39 · rising:0.44 · durable:0.45 · board:durable · trend:down* Repository to track versions of concrete strength data, models, and active learning proposals. #### [microsoft/promptbase](https://github.com/microsoft/promptbase) *Python · ★5,738 · MIT · — · score:0.00 · hot:0.39 · rising:0.47 · durable:0.52 · board:durable · trend:stable* All things prompt engineering #### [swyxio/spark-joy](https://github.com/swyxio/spark-joy) *? · ★9,745 · MIT · — · score:0.00 · hot:0.38 · rising:0.44 · durable:0.49 · board:durable · trend:stable* ✨😂 2000+ ways to add design flair, user delight, and whimsy to your product. #### [lucidrains/MEGABYTE-pytorch](https://github.com/lucidrains/MEGABYTE-pytorch) *Python · ★655 · MIT · — · score:0.00 · hot:0.38 · rising:0.44 · durable:0.52 · board:durable · trend:down* Implementation of MEGABYTE, Predicting Million-byte Sequences with Multiscale Transformers, in Pytorch #### [lucidrains/soundstorm-pytorch](https://github.com/lucidrains/soundstorm-pytorch) *Python · ★1,546 · MIT · — · score:0.00 · hot:0.38 · rising:0.46 · durable:0.55 · board:durable · trend:down* Implementation of SoundStorm, Efficient Parallel Audio Generation from Google Deepmind, in Pytorch #### [langchain-ai/agents-from-scratch](https://github.com/langchain-ai/agents-from-scratch) *Jupyter Notebook · ★1,733 · no-license · — · score:0.00 · hot:0.38 · rising:0.44 · durable:0.48 · board:durable · trend:stable* Build an email assistant with human-in-the-loop and memory #### [lucidrains/hyper-connections](https://github.com/lucidrains/hyper-connections) *Python · ★180 · MIT · — · score:0.00 · hot:0.38 · rising:0.43 · durable:0.49 · board:durable · trend:down* Attempt to make multiple residual streams from Bytedance's Hyper-Connections paper accessible to the public #### [tairov/llama2.mojo](https://github.com/tairov/llama2.mojo) *Mojo · ★2,119 · MIT · — · score:0.00 · hot:0.38 · rising:0.43 · durable:0.51 · board:durable · trend:down* Inference Llama 2 in one file of pure 🔥 #### [microsoft/sammo](https://github.com/microsoft/sammo) *Python · ★751 · MIT · — · score:0.00 · hot:0.38 · rising:0.45 · durable:0.54 · board:durable · trend:down* A library for prompt engineering and optimization (SAMMO = Structure-aware Multi-Objective Metaprompt Optimization) #### [karpathy/reader3](https://github.com/karpathy/reader3) *Python · ★3,534 · no-license · — · score:0.00 · hot:0.38 · rising:0.44 · durable:0.48 · board:durable · trend:stable* Quick illustration of how one can easily read books together with LLMs. It's great and I highly recommend it. #### [lucidrains/big-sleep](https://github.com/lucidrains/big-sleep) *Python · ★2,567 · MIT · — · score:0.00 · hot:0.38 · rising:0.45 · durable:0.50 · board:durable · trend:stable* A simple command line tool for text to image generation, using OpenAI's CLIP and a BigGAN. Technique was originally created by https://twitter.com/advadnoun #### [jsksxs360/How-to-use-Transformers](https://github.com/jsksxs360/How-to-use-Transformers) *Python · ★1,866 · Apache-2.0 · — · score:0.00 · hot:0.38 · rising:0.43 · durable:0.48 · board:durable · trend:down* Transformers 库快速入门教程 #### [rasbt/python-machine-learning-book-2nd-edition](https://github.com/rasbt/python-machine-learning-book-2nd-edition) *Jupyter Notebook · ★7,206 · MIT · — · score:0.00 · hot:0.38 · rising:0.48 · durable:0.49 · board:durable · trend:stable* The "Python Machine Learning (2nd edition)" book code repository and info resource #### [lucidrains/mlp-mixer-pytorch](https://github.com/lucidrains/mlp-mixer-pytorch) *Python · ★1,059 · MIT · — · score:0.00 · hot:0.38 · rising:0.45 · durable:0.52 · board:durable · trend:down* An All-MLP solution for Vision, from Google AI #### [lucidrains/flamingo-pytorch](https://github.com/lucidrains/flamingo-pytorch) *Python · ★1,268 · MIT · — · score:0.00 · hot:0.38 · rising:0.44 · durable:0.53 · board:durable · trend:down* Implementation of 🦩 Flamingo, state-of-the-art few-shot visual question answering attention net out of Deepmind, in Pytorch #### [LetterLiGo/SafeGen_CCS2024](https://github.com/LetterLiGo/SafeGen_CCS2024) *Python · ★138 · Apache-2.0 · — · score:0.00 · hot:0.38 · rising:0.41 · durable:0.42 · board:durable · trend:down* [CCS'24] SafeGen: Mitigating Unsafe Content Generation in Text-to-Image Models #### [karpathy/rustbpe](https://github.com/karpathy/rustbpe) *Rust · ★435 · MIT · — · score:0.00 · hot:0.38 · rising:0.44 · durable:0.51 · board:durable · trend:down* The missing tiktoken training code #### [zchoi/Awesome-Embodied-Robotics-and-Agent](https://github.com/zchoi/Awesome-Embodied-Robotics-and-Agent) *? · ★1,772 · Apache-2.0 · — · score:0.00 · hot:0.38 · rising:0.44 · durable:0.49 · board:durable · trend:down* This is a curated list of "Embodied AI or robot with Large Language Models" research. Watch this repository for the latest updates! 🔥 #### [SeekingDream/Static-to-Dynamic-LLMEval](https://github.com/SeekingDream/Static-to-Dynamic-LLMEval) *? · ★505 · no-license · — · score:0.00 · hot:0.38 · rising:0.41 · durable:0.46 · board:durable · trend:down* The official GitHub repository of the paper "Recent advances in large language model benchmarks against data contamination: From static to dynamic evaluation" #### [open-webui/openapi-servers](https://github.com/open-webui/openapi-servers) *Python · ★927 · MIT · — · score:0.00 · hot:0.38 · rising:0.44 · durable:0.50 · board:durable · trend:down* OpenAPI Tool Servers #### [microsoft/PromptWizard](https://github.com/microsoft/PromptWizard) *Python · ★3,835 · MIT · — · score:0.00 · hot:0.38 · rising:0.46 · durable:0.51 · board:durable · trend:stable* Task-Aware Agent-driven Prompt Optimization Framework #### [microsoft/fluentui-emoji](https://github.com/microsoft/fluentui-emoji) *Python · ★9,920 · MIT · — · score:0.00 · hot:0.38 · rising:0.46 · durable:0.50 · board:durable · trend:stable* A collection of familiar, friendly, and modern emoji from Microsoft #### [lm-sys/RouteLLM](https://github.com/lm-sys/RouteLLM) *Python · ★4,795 · Apache-2.0 · — · score:0.00 · hot:0.38 · rising:0.45 · durable:0.49 · board:durable · trend:stable* A framework for serving and evaluating LLM routers - save LLM costs without compromising quality #### [lucidrains/video-diffusion-pytorch](https://github.com/lucidrains/video-diffusion-pytorch) *Python · ★1,382 · MIT · — · score:0.00 · hot:0.38 · rising:0.45 · durable:0.50 · board:durable · trend:down* Implementation of Video Diffusion Models, Jonathan Ho's new paper extending DDPMs to Video Generation - in Pytorch #### [lucidrains/voicebox-pytorch](https://github.com/lucidrains/voicebox-pytorch) *Python · ★683 · MIT · — · score:0.00 · hot:0.38 · rising:0.45 · durable:0.50 · board:durable · trend:down* Implementation of Voicebox, new SOTA Text-to-speech network from MetaAI, in Pytorch #### [replicate/scribble-diffusion](https://github.com/replicate/scribble-diffusion) *JavaScript · ★2,982 · MIT · — · score:0.00 · hot:0.38 · rising:0.46 · durable:0.49 · board:durable · trend:stable* Turn your rough sketch into a refined image using AI #### [karpathy/cryptos](https://github.com/karpathy/cryptos) *Jupyter Notebook · ★1,882 · no-license · — · score:0.00 · hot:0.38 · rising:0.43 · durable:0.45 · board:durable · trend:stable* Pure Python from-scratch zero-dependency implementation of Bitcoin for educational purposes #### [jina-ai/auto-gpt-web](https://github.com/jina-ai/auto-gpt-web) *TypeScript · ★761 · MIT · — · score:0.00 · hot:0.38 · rising:0.45 · durable:0.52 · board:durable · trend:down* Set Your Goals, AI Achieves Them. #### [facebookresearch/moco](https://github.com/facebookresearch/moco) *? · ★5,126 · no-license · — · score:0.00 · hot:0.38 · rising:0.44 · durable:0.46 · board:durable · trend:stable* PyTorch implementation of MoCo: https://arxiv.org/abs/1911.05722 #### [microsoft/scalar](https://github.com/microsoft/scalar) *C# · ★1,502 · MIT · — · score:0.00 · hot:0.38 · rising:0.45 · durable:0.54 · board:durable · trend:down* Scalar: A set of tools and extensions for Git to allow very large monorepos to run on Git without a virtualization layer #### [lucidrains/mixture-of-experts](https://github.com/lucidrains/mixture-of-experts) *Python · ★855 · MIT · — · score:0.00 · hot:0.38 · rising:0.43 · durable:0.50 · board:durable · trend:down* A Pytorch implementation of Sparsely-Gated Mixture of Experts, for massively increasing the parameter count of language models #### [microsoft/CDM](https://github.com/microsoft/CDM) *C# · ★1,811 · CC-BY-4.0 · — · score:0.00 · hot:0.38 · rising:0.46 · durable:0.45 · board:rising · trend:stable* The Common Data Model (CDM) is a standard and extensible collection of schemas (entities, attributes, relationships) that represents business concepts and activities with well-defined semantics, to facilitate data interoperability. Examples of entities include: Account, Contact, Lead, Opportunity, Product, etc. #### [lucidrains/self-rewarding-lm-pytorch](https://github.com/lucidrains/self-rewarding-lm-pytorch) *Python · ★1,407 · MIT · — · score:0.00 · hot:0.37 · rising:0.44 · durable:0.54 · board:durable · trend:down* Implementation of the training framework proposed in Self-Rewarding Language Model, from MetaAI #### [showlab/ShowUI](https://github.com/showlab/ShowUI) *Python · ★1,803 · Apache-2.0 · — · score:0.00 · hot:0.37 · rising:0.43 · durable:0.48 · board:durable · trend:down* [CVPR 2025] Open-source, End-to-end, Vision-Language-Action model for GUI Agent & Computer Use. #### [lucidrains/phenaki-pytorch](https://github.com/lucidrains/phenaki-pytorch) *Python · ★793 · MIT · — · score:0.00 · hot:0.37 · rising:0.44 · durable:0.50 · board:durable · trend:down* Implementation of Phenaki Video, which uses Mask GIT to produce text guided videos of up to 2 minutes in length, in Pytorch #### [FireRedTeam/FireRedASR](https://github.com/FireRedTeam/FireRedASR) *Python · ★1,847 · Apache-2.0 · — · score:0.00 · hot:0.37 · rising:0.40 · durable:0.45 · board:durable · trend:down* Open-source industrial-grade ASR models supporting Mandarin, Chinese dialects and English, achieving a new SOTA on public Mandarin ASR benchmarks, while also offering outstanding singing lyrics recognition capability. #### [lucidrains/linear-attention-transformer](https://github.com/lucidrains/linear-attention-transformer) *Python · ★826 · MIT · — · score:0.00 · hot:0.37 · rising:0.42 · durable:0.50 · board:durable · trend:down* Transformer based on a variant of attention that is linear complexity in respect to sequence length #### [lucidrains/enformer-pytorch](https://github.com/lucidrains/enformer-pytorch) *Python · ★562 · MIT · — · score:0.00 · hot:0.37 · rising:0.43 · durable:0.47 · board:durable · trend:down* Implementation of Enformer, Deepmind's attention network for predicting gene expression, in Pytorch #### [Capsize-Games/airunner](https://github.com/Capsize-Games/airunner) *Python · ★1,316 · GPL-3.0 · — · score:0.00 · hot:0.37 · rising:0.43 · durable:0.50 · board:durable · trend:down* Offline inference engine for art, real-time voice conversations, LLM powered chatbots and automated workflows #### [jina-ai/dalle-flow](https://github.com/jina-ai/dalle-flow) *Python · ★2,831 · no-license · — · score:0.00 · hot:0.37 · rising:0.43 · durable:0.47 · board:durable · trend:stable* 🌊 A Human-in-the-Loop workflow for creating HD images from text #### [run-llama/sec-insights](https://github.com/run-llama/sec-insights) *TypeScript · ★2,596 · MIT · — · score:0.00 · hot:0.37 · rising:0.45 · durable:0.48 · board:durable · trend:stable* A real world full-stack application using LlamaIndex #### [allenai/asta-paper-finder](https://github.com/allenai/asta-paper-finder) *Python · ★236 · Apache-2.0 · — · score:0.00 · hot:0.37 · rising:0.42 · durable:0.46 · board:durable · trend:down* frozen-in-time version of our Paper Finder agent for reproducing evaluation results #### [karpathy/rendergit](https://github.com/karpathy/rendergit) *Python · ★2,219 · no-license · — · score:0.00 · hot:0.37 · rising:0.43 · durable:0.46 · board:durable · trend:down* Render any git repo into a single static HTML page for humans or LLMs #### [karpathy/arxiv-sanity-preserver](https://github.com/karpathy/arxiv-sanity-preserver) *Python · ★5,647 · MIT · — · score:0.00 · hot:0.37 · rising:0.45 · durable:0.45 · board:rising · trend:stable* Web interface for browsing, search and filtering recent arxiv submissions #### [karpathy/neuraltalk2](https://github.com/karpathy/neuraltalk2) *Jupyter Notebook · ★5,581 · no-license · — · score:0.00 · hot:0.37 · rising:0.43 · durable:0.40 · board:rising · trend:stable* Efficient Image Captioning code in Torch, runs on GPU #### [lucidrains/siren-pytorch](https://github.com/lucidrains/siren-pytorch) *Python · ★506 · MIT · — · score:0.00 · hot:0.37 · rising:0.43 · durable:0.49 · board:durable · trend:down* Pytorch implementation of SIREN - Implicit Neural Representations with Periodic Activation Function #### [rasbt/stat479-machine-learning-fs19](https://github.com/rasbt/stat479-machine-learning-fs19) *Jupyter Notebook · ★779 · no-license · — · score:0.00 · hot:0.37 · rising:0.44 · durable:0.44 · board:durable · trend:stable* Course material for STAT 479: Machine Learning (FS 2019) taught by Sebastian Raschka at University Wisconsin-Madison #### [lucidrains/e2-tts-pytorch](https://github.com/lucidrains/e2-tts-pytorch) *Python · ★516 · MIT · — · score:0.00 · hot:0.37 · rising:0.45 · durable:0.52 · board:durable · trend:down* Implementation of E2-TTS, "Embarrassingly Easy Fully Non-Autoregressive Zero-Shot TTS", in Pytorch #### [facebookresearch/SONAR](https://github.com/facebookresearch/SONAR) *Python · ★883 · NOASSERTION · — · score:0.00 · hot:0.37 · rising:0.44 · durable:0.49 · board:durable · trend:down* SONAR, a new multilingual and multimodal fixed-size sentence embedding space, with a full suite of speech and text encoders and decoders. #### [lucidrains/ring-attention-pytorch](https://github.com/lucidrains/ring-attention-pytorch) *Python · ★547 · MIT · — · score:0.00 · hot:0.37 · rising:0.43 · durable:0.51 · board:durable · trend:down* Implementation of 💍 Ring Attention, from Liu et al. at Berkeley AI, in Pytorch #### [dair-ai/Transformers-Recipe](https://github.com/dair-ai/Transformers-Recipe) *? · ★1,634 · CC0-1.0 · — · score:0.00 · hot:0.37 · rising:0.44 · durable:0.47 · board:durable · trend:down* 🧠 A study guide to learn about Transformers #### [lucidrains/memorizing-transformers-pytorch](https://github.com/lucidrains/memorizing-transformers-pytorch) *Python · ★643 · MIT · — · score:0.00 · hot:0.37 · rising:0.41 · durable:0.50 · board:durable · trend:down* Implementation of Memorizing Transformers (ICLR 2022), attention net augmented with indexing and retrieval of memories using approximate nearest neighbors, in Pytorch #### [lucidrains/metnet3-pytorch](https://github.com/lucidrains/metnet3-pytorch) *Python · ★239 · MIT · — · score:0.00 · hot:0.37 · rising:0.42 · durable:0.50 · board:durable · trend:down* Implementation of MetNet-3, SOTA neural weather model out of Google Deepmind, in Pytorch #### [lucidrains/conformer](https://github.com/lucidrains/conformer) *Python · ★433 · MIT · — · score:0.00 · hot:0.37 · rising:0.42 · durable:0.49 · board:durable · trend:down* Implementation of the convolutional module from the Conformer paper, for use in Transformers #### [AMAP-ML/Tree-GRPO](https://github.com/AMAP-ML/Tree-GRPO) *Python · ★341 · Apache-2.0 · — · score:0.00 · hot:0.37 · rising:0.40 · durable:0.46 · board:durable · trend:down* [ICLR 2026] Tree Search for LLM Agent Reinforcement Learning #### [vercel/virtual-event-starter-kit](https://github.com/vercel/virtual-event-starter-kit) *TypeScript · ★2,186 · MIT · — · score:0.00 · hot:0.37 · rising:0.45 · durable:0.48 · board:durable · trend:stable* Open source demo that Next.js developers can clone, deploy, and fully customize for events. #### [jina-ai/MCP](https://github.com/jina-ai/MCP) *TypeScript · ★646 · Apache-2.0 · — · score:0.00 · hot:0.37 · rising:0.41 · durable:0.44 · board:durable · trend:down* Official Jina AI Remote MCP Server #### [lucidrains/lightweight-gan](https://github.com/lucidrains/lightweight-gan) *Python · ★1,676 · MIT · — · score:0.00 · hot:0.37 · rising:0.44 · durable:0.47 · board:durable · trend:stable* Implementation of 'lightweight' GAN, proposed in ICLR 2021, in Pytorch. High resolution image generations that can be trained within a day or two #### [vercel/hazel](https://github.com/vercel/hazel) *JavaScript · ★3,037 · MIT · — · score:0.00 · hot:0.37 · rising:0.45 · durable:0.50 · board:durable · trend:stable* Lightweight update server for Electron apps #### [SamurAIGPT/Text-To-Video-AI](https://github.com/SamurAIGPT/Text-To-Video-AI) *Jupyter Notebook · ★718 · MIT · — · score:0.00 · hot:0.37 · rising:0.43 · durable:0.47 · board:durable · trend:down* Generate video from text using AI #### [facebookresearch/dietgpu](https://github.com/facebookresearch/dietgpu) *Cuda · ★382 · MIT · — · score:0.00 · hot:0.37 · rising:0.42 · durable:0.42 · board:durable · trend:down* GPU implementation of a fast generalized ANS (asymmetric numeral system) entropy encoder and decoder, with extensions for lossless compression of numerical and other data types in HPC/ML applications. #### [jina-ai/correlations](https://github.com/jina-ai/correlations) *HTML · ★312 · Apache-2.0 · — · score:0.00 · hot:0.37 · rising:0.45 · durable:0.48 · board:durable · trend:down* Simple UI for debugging correlations of text embeddings #### [vercel/nextjs-postgres-nextauth-tailwindcss-template](https://github.com/vercel/nextjs-postgres-nextauth-tailwindcss-template) *TypeScript · ★1,598 · MIT · — · score:0.00 · hot:0.37 · rising:0.44 · durable:0.48 · board:durable · trend:stable* Admin dashboard template. #### [jina-ai/vectordb](https://github.com/jina-ai/vectordb) *Python · ★648 · Apache-2.0 · — · score:0.00 · hot:0.37 · rising:0.43 · durable:0.52 · board:durable · trend:down* A Python vector database you just need - no more, no less. #### [lucidrains/minGRU-pytorch](https://github.com/lucidrains/minGRU-pytorch) *Python · ★323 · MIT · — · score:0.00 · hot:0.37 · rising:0.43 · durable:0.50 · board:durable · trend:down* Implementation of the proposed minGRU in Pytorch #### [microsoft/Codex-CLI](https://github.com/microsoft/Codex-CLI) *Python · ★2,364 · MIT · — · score:0.00 · hot:0.37 · rising:0.44 · durable:0.47 · board:durable · trend:stable* CLI tool that uses Codex to turn natural language commands into their Bash/ZShell/PowerShell equivalents #### [replicate/replicate-javascript](https://github.com/replicate/replicate-javascript) *TypeScript · ★594 · Apache-2.0 · — · score:0.00 · hot:0.37 · rising:0.45 · durable:0.50 · board:durable · trend:down* Node.js client for Replicate #### [lucidrains/native-sparse-attention-pytorch](https://github.com/lucidrains/native-sparse-attention-pytorch) *Python · ★799 · MIT · — · score:0.00 · hot:0.37 · rising:0.43 · durable:0.50 · board:durable · trend:down* Implementation of the sparse attention pattern proposed by the Deepseek team in their "Native Sparse Attention" paper #### [Lakonik/piFlow](https://github.com/Lakonik/piFlow) *Python · ★282 · Apache-2.0 · — · score:0.00 · hot:0.37 · rising:0.40 · durable:0.50 · board:durable · trend:down* [ICLR 2026] pi-Flow: Policy-Based Few-Step Generation via Imitation Distillation #### [lucidrains/tiny-recursive-model](https://github.com/lucidrains/tiny-recursive-model) *Python · ★175 · MIT · — · score:0.00 · hot:0.37 · rising:0.43 · durable:0.51 · board:durable · trend:down* Unofficial implementation of Tiny Recursive Model (TRM), improvement to HRM from Sapient AI, by Alexia Jolicoeur-Martineau #### [VinAIResearch/Anti-DreamBooth](https://github.com/VinAIResearch/Anti-DreamBooth) *Python · ★268 · AGPL-3.0 · — · score:0.00 · hot:0.37 · rising:0.41 · durable:0.42 · board:durable · trend:down* Anti-DreamBooth: Protecting users from personalized text-to-image synthesis (ICCV 2023) #### [yoheinakajima/instagraph](https://github.com/yoheinakajima/instagraph) *Python · ★3,547 · MIT · — · score:0.00 · hot:0.37 · rising:0.44 · durable:0.48 · board:durable · trend:stable* Converts text input or URL into knowledge graph and displays #### [microsoft/verona](https://github.com/microsoft/verona) *C++ · ★3,709 · MIT · — · score:0.00 · hot:0.37 · rising:0.44 · durable:0.49 · board:durable · trend:stable* Research programming language for concurrent ownership #### [lucidrains/robotic-transformer-pytorch](https://github.com/lucidrains/robotic-transformer-pytorch) *Python · ★448 · MIT · — · score:0.00 · hot:0.36 · rising:0.42 · durable:0.51 · board:durable · trend:down* Implementation of RT1 (Robotic Transformer) in Pytorch #### [NVIDIA/Stable-Diffusion-WebUI-TensorRT](https://github.com/NVIDIA/Stable-Diffusion-WebUI-TensorRT) *Python · ★1,993 · MIT · — · score:0.00 · hot:0.36 · rising:0.44 · durable:0.50 · board:durable · trend:down* TensorRT Extension for Stable Diffusion Web UI #### [vercel/modelfusion](https://github.com/vercel/modelfusion) *TypeScript · ★1,319 · MIT · — · score:0.00 · hot:0.36 · rising:0.42 · durable:0.50 · board:durable · trend:down* The TypeScript library for building AI applications. #### [sgl-project/sgl-learning-materials](https://github.com/sgl-project/sgl-learning-materials) *? · ★801 · MIT · — · score:0.00 · hot:0.36 · rising:0.44 · durable:0.49 · board:durable · trend:down* Materials for learning SGLang #### [lucidrains/memory-efficient-attention-pytorch](https://github.com/lucidrains/memory-efficient-attention-pytorch) *Python · ★390 · MIT · — · score:0.00 · hot:0.36 · rising:0.43 · durable:0.49 · board:durable · trend:down* Implementation of a memory efficient multi-head attention as proposed in the paper, "Self-attention Does Not Need O(n²) Memory" #### [rasbt/MachineLearning-QandAI-book](https://github.com/rasbt/MachineLearning-QandAI-book) *Jupyter Notebook · ★718 · BSD-3-Clause · — · score:0.00 · hot:0.36 · rising:0.42 · durable:0.47 · board:durable · trend:down* Machine Learning Q and AI book #### [vercel/styled-jsx](https://github.com/vercel/styled-jsx) *JavaScript · ★7,788 · MIT · — · score:0.00 · hot:0.36 · rising:0.44 · durable:0.54 · board:durable · trend:down* Full CSS support for JSX without compromises #### [yoheinakajima/babyagi3](https://github.com/yoheinakajima/babyagi3) *Python · ★121 · MIT · — · score:0.00 · hot:0.36 · rising:0.43 · durable:0.44 · board:durable · trend:down* #### [Kenza-AI/sagify](https://github.com/Kenza-AI/sagify) *Python · ★443 · MIT · — · score:0.00 · hot:0.36 · rising:0.39 · durable:0.47 · board:durable · trend:down* LLMs and Machine Learning done easily #### [run-llama/create-llama](https://github.com/run-llama/create-llama) *Python · ★1,484 · MIT · — · score:0.00 · hot:0.36 · rising:0.44 · durable:0.51 · board:durable · trend:down* The easiest way to get started with LlamaIndex #### [lucidrains/point-transformer-pytorch](https://github.com/lucidrains/point-transformer-pytorch) *Python · ★599 · MIT · — · score:0.00 · hot:0.36 · rising:0.43 · durable:0.48 · board:durable · trend:down* Implementation of the Point Transformer layer, in Pytorch #### [run-llama/chat-ui](https://github.com/run-llama/chat-ui) *TypeScript · ★578 · MIT · — · score:0.00 · hot:0.36 · rising:0.42 · durable:0.51 · board:durable · trend:down* Chat UI components for LLM apps #### [microsoft/T-MAC](https://github.com/microsoft/T-MAC) *C++ · ★952 · MIT · — · score:0.00 · hot:0.36 · rising:0.43 · durable:0.49 · board:durable · trend:down* Low-bit LLM inference on CPU/NPU with lookup table #### [openai/human-eval-infilling](https://github.com/openai/human-eval-infilling) *Python · ★203 · MIT · — · score:0.00 · hot:0.36 · rising:0.44 · durable:0.45 · board:durable · trend:down* Code for the paper "Efficient Training of Language Models to Fill in the Middle" #### [swyxio/brain](https://github.com/swyxio/brain) *? · ★1,619 · no-license · — · score:0.00 · hot:0.36 · rising:0.42 · durable:0.43 · board:durable · trend:stable* Swyx's second brain! #### [NVIDIA/AMGX](https://github.com/NVIDIA/AMGX) *Cuda · ★661 · no-license · — · score:0.00 · hot:0.36 · rising:0.42 · durable:0.43 · board:durable · trend:stable* Distributed multigrid linear solver library on GPU #### [facebookresearch/FlowDec](https://github.com/facebookresearch/FlowDec) *Python · ★202 · NOASSERTION · — · score:0.00 · hot:0.36 · rising:0.42 · durable:0.48 · board:durable · trend:down* An neural full-band audio codec for general audio sampled at 48 kHz with 7.5 kps or 4.5 kbps. #### [NVIDIA/jetson-gpio](https://github.com/NVIDIA/jetson-gpio) *Python · ★1,055 · MIT · — · score:0.00 · hot:0.36 · rising:0.44 · durable:0.46 · board:durable · trend:stable* A Python library that enables the use of Jetson's GPIOs #### [lucidrains/TRI-LBM](https://github.com/lucidrains/TRI-LBM) *Python · ★105 · MIT · — · score:0.00 · hot:0.36 · rising:0.40 · durable:0.50 · board:durable · trend:down* Implementation of the Large Behavioral Model architecture for dexterous manipulation from Toyota Research Institute #### [open-webui/pipelines](https://github.com/open-webui/pipelines) *Python · ★2,348 · MIT · — · score:0.00 · hot:0.36 · rising:0.44 · durable:0.45 · board:durable · trend:stable* Pipelines: Versatile, UI-Agnostic OpenAI-Compatible Plugin Framework #### [rasbt/stat453-deep-learning-ss20](https://github.com/rasbt/stat453-deep-learning-ss20) *Jupyter Notebook · ★575 · MIT · — · score:0.00 · hot:0.36 · rising:0.43 · durable:0.44 · board:durable · trend:stable* STAT 453: Intro to Deep Learning @ UW-Madison (Spring 2020) #### [lucidrains/egnn-pytorch](https://github.com/lucidrains/egnn-pytorch) *Python · ★523 · MIT · — · score:0.00 · hot:0.36 · rising:0.42 · durable:0.47 · board:durable · trend:down* Implementation of E(n)-Equivariant Graph Neural Networks, in Pytorch #### [yoheinakajima/ditto](https://github.com/yoheinakajima/ditto) *Python · ★1,054 · MIT · — · score:0.00 · hot:0.36 · rising:0.42 · durable:0.46 · board:durable · trend:down* the simplest self-building coding agent #### [microsoft/ms-tpm-20-ref](https://github.com/microsoft/ms-tpm-20-ref) *C · ★392 · NOASSERTION · — · score:0.00 · hot:0.36 · rising:0.42 · durable:0.42 · board:durable · trend:stable* Reference implementation of the TCG Trusted Platform Module 2.0 specification. #### [PaddlePaddle/PaddleMIX](https://github.com/PaddlePaddle/PaddleMIX) *Python · ★721 · Apache-2.0 · — · score:0.00 · hot:0.36 · rising:0.43 · durable:0.48 · board:durable · trend:down* Paddle Multimodal Integration and eXploration, supporting mainstream multi-modal tasks, including end-to-end large-scale multi-modal pretrain models and diffusion model toolbox. Equipped with high performance and flexibility. #### [lucidrains/pi-GAN-pytorch](https://github.com/lucidrains/pi-GAN-pytorch) *Python · ★125 · MIT · — · score:0.00 · hot:0.36 · rising:0.42 · durable:0.47 · board:durable · trend:down* Implementation of π-GAN, for 3d-aware image synthesis, in Pytorch #### [taylorwilsdon/open-webui-postgres-migration](https://github.com/taylorwilsdon/open-webui-postgres-migration) *Python · ★218 · MIT · — · score:0.00 · hot:0.36 · rising:0.42 · durable:0.49 · board:durable · trend:down* Interactive, locally hosted tool to migrate Open-WebUI SQLite databases to PostgreSQL #### [NVIDIA/tacotron2](https://github.com/NVIDIA/tacotron2) *Jupyter Notebook · ★5,304 · BSD-3-Clause · — · score:0.00 · hot:0.36 · rising:0.44 · durable:0.44 · board:rising · trend:stable* Tacotron 2 - PyTorch implementation with faster-than-realtime inference #### [huggingface/neuralcoref](https://github.com/huggingface/neuralcoref) *C · ★2,891 · MIT · — · score:0.00 · hot:0.36 · rising:0.45 · durable:0.51 · board:durable · trend:down* ✨Fast Coreference Resolution in spaCy with Neural Networks #### [rasbt/stat479-deep-learning-ss19](https://github.com/rasbt/stat479-deep-learning-ss19) *Jupyter Notebook · ★547 · no-license · — · score:0.00 · hot:0.36 · rising:0.42 · durable:0.42 · board:durable · trend:down* Course material for STAT 479: Deep Learning (SS 2019) at University Wisconsin-Madison #### [microsoft/accessibility-insights-windows](https://github.com/microsoft/accessibility-insights-windows) *C# · ★521 · NOASSERTION · — · score:0.00 · hot:0.36 · rising:0.42 · durable:0.43 · board:durable · trend:down* Accessibility Insights for Windows #### [lucidrains/grokfast-pytorch](https://github.com/lucidrains/grokfast-pytorch) *Python · ★105 · MIT · — · score:0.00 · hot:0.36 · rising:0.40 · durable:0.48 · board:durable · trend:down* Explorations into the proposal from the paper "Grokfast, Accelerated Grokking by Amplifying Slow Gradients" #### [vercel/kirimase](https://github.com/vercel/kirimase) *TypeScript · ★2,802 · MIT · — · score:0.00 · hot:0.36 · rising:0.43 · durable:0.51 · board:durable · trend:down* Build full-stack Next.js apps, incredibly fast #### [ELS-RD/kernl](https://github.com/ELS-RD/kernl) *Jupyter Notebook · ★1,585 · Apache-2.0 · — · score:0.00 · hot:0.36 · rising:0.41 · durable:0.52 · board:durable · trend:down* Kernl lets you run PyTorch transformer models several times faster on GPU with a single line of code, and is designed to be easily hackable. #### [NVIDIA/runx](https://github.com/NVIDIA/runx) *Python · ★640 · BSD-3-Clause · — · score:0.00 · hot:0.36 · rising:0.43 · durable:0.49 · board:durable · trend:down* Deep Learning Experiment Management #### [jina-ai/dev-gpt](https://github.com/jina-ai/dev-gpt) *Python · ★1,862 · Apache-2.0 · — · score:0.00 · hot:0.36 · rising:0.44 · durable:0.48 · board:durable · trend:down* Your Virtual Development Team #### [vercel/next-react-server-components](https://github.com/vercel/next-react-server-components) *JavaScript · ★997 · MIT · — · score:0.00 · hot:0.36 · rising:0.43 · durable:0.47 · board:durable · trend:down* Demo repository for Next.js + React Server Components #### [OpenRouterTeam/awesome-openrouter](https://github.com/OpenRouterTeam/awesome-openrouter) *JavaScript · ★256 · no-license · — · score:0.00 · hot:0.36 · rising:0.39 · durable:0.39 · board:rising · trend:down* Awesome list of apps that work with OpenRouter. OpenRouter provides access to 300+ AI Models through a single API. #### [facebookresearch/EasyComDataset](https://github.com/facebookresearch/EasyComDataset) *? · ★139 · NOASSERTION · — · score:0.00 · hot:0.36 · rising:0.41 · durable:0.47 · board:durable · trend:down* The Easy Communications (EasyCom) dataset is a world-first dataset designed to help mitigate the *cocktail party effect* from an augmented-reality (AR) -motivated multi-sensor egocentric world view. #### [dair-ai/GNNs-Recipe](https://github.com/dair-ai/GNNs-Recipe) *? · ★1,288 · CC0-1.0 · — · score:0.00 · hot:0.36 · rising:0.42 · durable:0.47 · board:durable · trend:down* 🟠 A study guide to learn about Graph Neural Networks (GNNs) #### [huggingface/pytorch-pretrained-BigGAN](https://github.com/huggingface/pytorch-pretrained-BigGAN) *Python · ★1,043 · MIT · — · score:0.00 · hot:0.36 · rising:0.44 · durable:0.52 · board:durable · trend:down* 🦋A PyTorch implementation of BigGAN with pretrained weights and conversion scripts. #### [microsoft/NeuralSpeech](https://github.com/microsoft/NeuralSpeech) *Python · ★1,459 · MIT · — · score:0.00 · hot:0.36 · rising:0.43 · durable:0.46 · board:durable · trend:stable* #### [karpathy/arxiv-sanity-lite](https://github.com/karpathy/arxiv-sanity-lite) *Python · ★1,561 · MIT · — · score:0.00 · hot:0.36 · rising:0.42 · durable:0.45 · board:durable · trend:down* arxiv-sanity lite: tag arxiv papers of interest get recommendations of similar papers in a nice UI using SVMs over tfidf feature vectors based on paper abstracts. #### [lucidrains/iTransformer](https://github.com/lucidrains/iTransformer) *Python · ★535 · MIT · — · score:0.00 · hot:0.36 · rising:0.41 · durable:0.49 · board:durable · trend:down* Unofficial implementation of iTransformer - SOTA Time Series Forecasting using Attention networks, out of Tsinghua / Ant group #### [lucidrains/nuwa-pytorch](https://github.com/lucidrains/nuwa-pytorch) *Python · ★549 · MIT · — · score:0.00 · hot:0.36 · rising:0.42 · durable:0.50 · board:durable · trend:down* Implementation of NÜWA, state of the art attention network for text to video synthesis, in Pytorch #### [dair-ai/nlp_paper_summaries](https://github.com/dair-ai/nlp_paper_summaries) *? · ★1,477 · MIT · — · score:0.00 · hot:0.36 · rising:0.43 · durable:0.45 · board:durable · trend:stable* ✍️ A carefully curated list of NLP paper summaries #### [lucidrains/halonet-pytorch](https://github.com/lucidrains/halonet-pytorch) *Python · ★199 · MIT · — · score:0.00 · hot:0.36 · rising:0.41 · durable:0.48 · board:durable · trend:down* Implementation of the 😇 Attention layer from the paper, Scaling Local Self-Attention For Parameter Efficient Visual Backbones #### [microsoft/vscode-icons](https://github.com/microsoft/vscode-icons) *? · ★939 · CC-BY-4.0 · — · score:0.00 · hot:0.36 · rising:0.43 · durable:0.46 · board:durable · trend:down* Icons for Visual Studio Code #### [yoheinakajima/mindgraph](https://github.com/yoheinakajima/mindgraph) *Python · ★930 · no-license · — · score:0.00 · hot:0.36 · rising:0.41 · durable:0.44 · board:durable · trend:down* proof of concept prototype for generating and querying against an ever-expanding knowledge graph with ai #### [replicate/paint-by-text](https://github.com/replicate/paint-by-text) *JavaScript · ★470 · MIT · — · score:0.00 · hot:0.36 · rising:0.43 · durable:0.46 · board:durable · trend:down* A microsite for InstructPix2Pix #### [vercel/hyper-site](https://github.com/vercel/hyper-site) *JavaScript · ★461 · MIT · — · score:0.00 · hot:0.36 · rising:0.44 · durable:0.45 · board:durable · trend:stable* The official website for the Hyper terminal #### [rasbt/stat451-machine-learning-fs20](https://github.com/rasbt/stat451-machine-learning-fs20) *Jupyter Notebook · ★456 · no-license · — · score:0.00 · hot:0.36 · rising:0.41 · durable:0.43 · board:durable · trend:down* STAT 451: Intro to Machine Learning @ UW-Madison (Fall 2020) #### [NVIDIA/multi-gpu-programming-models](https://github.com/NVIDIA/multi-gpu-programming-models) *Cuda · ★883 · BSD-3-Clause · — · score:0.00 · hot:0.36 · rising:0.43 · durable:0.46 · board:durable · trend:down* Examples demonstrating available options to program multiple GPUs in a single node or a cluster #### [karpathy/reinforcejs](https://github.com/karpathy/reinforcejs) *HTML · ★1,454 · no-license · — · score:0.00 · hot:0.36 · rising:0.41 · durable:0.39 · board:rising · trend:stable* Reinforcement Learning Agents in Javascript (Dynamic Programming, Temporal Difference, Deep Q-Learning, Stochastic/Deterministic Policy Gradients) #### [wooyeolbaek/attention-map-diffusers](https://github.com/wooyeolbaek/attention-map-diffusers) *Python · ★402 · MIT · — · score:0.00 · hot:0.36 · rising:0.40 · durable:0.44 · board:durable · trend:down* 🚀 Cross attention map tools for huggingface/diffusers #### [allenai/discoveryworld](https://github.com/allenai/discoveryworld) *Python · ★210 · Apache-2.0 · — · score:0.00 · hot:0.36 · rising:0.39 · durable:0.45 · board:durable · trend:down* A virtual environment for developing and evaluating automated scientific discovery agents. #### [microsoft/multi-agent-reference-architecture](https://github.com/microsoft/multi-agent-reference-architecture) *Python · ★190 · MIT · — · score:0.00 · hot:0.36 · rising:0.41 · durable:0.45 · board:durable · trend:down* Guide for designing adaptive, scalable, and secure enterprise multi-agent systems #### [vercel/nextjs-postgres-auth-starter](https://github.com/vercel/nextjs-postgres-auth-starter) *TypeScript · ★1,053 · no-license · — · score:0.00 · hot:0.36 · rising:0.42 · durable:0.45 · board:durable · trend:down* Next.js + Tailwind + Typescript + Drizzle + NextAuth + PostgreSQL starter template. #### [vercel/nextgram](https://github.com/vercel/nextgram) *TypeScript · ★1,014 · no-license · — · score:0.00 · hot:0.36 · rising:0.41 · durable:0.44 · board:durable · trend:down* A sample Next.js app showing dynamic routing with modals as a route. #### [allenai/wimbd](https://github.com/allenai/wimbd) *Python · ★227 · Apache-2.0 · — · score:0.00 · hot:0.35 · rising:0.43 · durable:0.50 · board:durable · trend:down* What's In My Big Data (WIMBD) - a toolkit for analyzing large text datasets #### [NVIDIA/MegaMolBART](https://github.com/NVIDIA/MegaMolBART) *Python · ★182 · no-license · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.46 · board:durable · trend:down* A deep learning model for small molecule drug discovery and cheminformatics based on SMILES #### [jxnl/systematically-improving-rag](https://github.com/jxnl/systematically-improving-rag) *HTML · ★229 · no-license · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.43 · board:durable · trend:down* #### [lucidrains/linformer](https://github.com/lucidrains/linformer) *Python · ★306 · MIT · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.48 · board:durable · trend:down* Implementation of Linformer for Pytorch #### [lucidrains/classifier-free-guidance-pytorch](https://github.com/lucidrains/classifier-free-guidance-pytorch) *Python · ★541 · MIT · — · score:0.00 · hot:0.35 · rising:0.42 · durable:0.51 · board:durable · trend:down* Implementation of Classifier Free Guidance in Pytorch, with emphasis on text conditioning, and flexibility to include multiple text embedding models #### [simonw/sqlite-utils](https://github.com/simonw/sqlite-utils) *Python · ★2,037 · Apache-2.0 · — · score:0.00 · hot:0.35 · rising:0.42 · durable:0.46 · board:durable · trend:down* Python CLI utility and library for manipulating SQLite databases #### [vercel/react-tweet](https://github.com/vercel/react-tweet) *TypeScript · ★1,860 · MIT · — · score:0.00 · hot:0.35 · rising:0.43 · durable:0.50 · board:durable · trend:down* Embed tweets in your React application. #### [simonw/claude-skills](https://github.com/simonw/claude-skills) *? · ★921 · no-license · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.45 · board:durable · trend:down* The contents of /mnt/skills in Claude's code interpreter environment #### [replicate/replicate-python](https://github.com/replicate/replicate-python) *Python · ★900 · Apache-2.0 · — · score:0.00 · hot:0.35 · rising:0.44 · durable:0.49 · board:durable · trend:down* Python client for Replicate #### [microsoft/DeBERTa](https://github.com/microsoft/DeBERTa) *Python · ★2,211 · MIT · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.43 · board:durable · trend:stable* The implementation of DeBERTa #### [facebookresearch/blt](https://github.com/facebookresearch/blt) *Python · ★2,033 · NOASSERTION · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.43 · board:durable · trend:stable* Code for BLT research paper #### [chroma-core/chroma-mcp](https://github.com/chroma-core/chroma-mcp) *Python · ★535 · Apache-2.0 · — · score:0.00 · hot:0.35 · rising:0.42 · durable:0.48 · board:durable · trend:down* A Model Context Protocol (MCP) server implementation that provides database capabilities for Chroma #### [jina-ai/mlx-retrieval](https://github.com/jina-ai/mlx-retrieval) *Python · ★181 · Apache-2.0 · — · score:0.00 · hot:0.35 · rising:0.43 · durable:0.47 · board:durable · trend:down* Train embedding and reranker models for retrieval tasks on Apple Silicon with MLX #### [rasbt/scipy2023-deeplearning](https://github.com/rasbt/scipy2023-deeplearning) *Jupyter Notebook · ★599 · Apache-2.0 · — · score:0.00 · hot:0.35 · rising:0.42 · durable:0.46 · board:durable · trend:down* #### [lucidrains/bit-diffusion](https://github.com/lucidrains/bit-diffusion) *Python · ★356 · MIT · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.49 · board:durable · trend:down* Implementation of Bit Diffusion, Hinton's group's attempt at discrete denoising diffusion, in Pytorch #### [facebookresearch/emg2pose](https://github.com/facebookresearch/emg2pose) *Python · ★214 · NOASSERTION · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.43 · board:durable · trend:down* Code repository for emg2pose dataset and model benchmarks #### [jina-ai/late-chunking](https://github.com/jina-ai/late-chunking) *Python · ★500 · Apache-2.0 · — · score:0.00 · hot:0.35 · rising:0.42 · durable:0.47 · board:durable · trend:down* Code for explaining and evaluating late chunking (chunked pooling) #### [run-llama/agentfs-claude](https://github.com/run-llama/agentfs-claude) *TypeScript · ★322 · MIT · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.48 · board:durable · trend:down* Run Claude Code/Codex within AgentFS, orchestrated by LlamaIndex Workflows #### [allenai/genesys](https://github.com/allenai/genesys) *Python · ★166 · Apache-2.0 · — · score:0.00 · hot:0.35 · rising:0.43 · durable:0.52 · board:durable · trend:down* Source code and utilities for the Genesys distributed language model architecture discovery system. #### [chatsci/Aeiva](https://github.com/chatsci/Aeiva) *Python · ★159 · Apache-2.0 · — · score:0.00 · hot:0.35 · rising:0.38 · durable:0.45 · board:durable · trend:down* A general AI agent framework that can be adapted to various tasks and environments. #### [rasbt/watermark](https://github.com/rasbt/watermark) *Python · ★943 · NOASSERTION · — · score:0.00 · hot:0.35 · rising:0.40 · durable:0.43 · board:durable · trend:down* An IPython magic extension for printing date and time stamps, version numbers, and hardware information #### [microsoft/DNS-Challenge](https://github.com/microsoft/DNS-Challenge) *Python · ★1,407 · CC-BY-4.0 · — · score:0.00 · hot:0.35 · rising:0.42 · durable:0.42 · board:rising · trend:stable* This repo contains the scripts, models, and required files for the Deep Noise Suppression (DNS) Challenge. #### [facebookresearch/locate-3d](https://github.com/facebookresearch/locate-3d) *Python · ★431 · NOASSERTION · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.45 · board:durable · trend:down* Open source repo for Locate 3D Model, 3D-JEPA and Locate 3D Dataset #### [lucidrains/meshgpt-pytorch](https://github.com/lucidrains/meshgpt-pytorch) *Python · ★861 · MIT · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.48 · board:durable · trend:down* Implementation of MeshGPT, SOTA Mesh generation using Attention, in Pytorch #### [lucidrains/RQ-Transformer](https://github.com/lucidrains/RQ-Transformer) *Python · ★125 · MIT · — · score:0.00 · hot:0.35 · rising:0.39 · durable:0.48 · board:durable · trend:down* Implementation of RQ Transformer, proposed in the paper "Autoregressive Image Generation using Residual Quantization" #### [NVIDIA/Dataset_Synthesizer](https://github.com/NVIDIA/Dataset_Synthesizer) *C++ · ★600 · NOASSERTION · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.43 · board:durable · trend:down* NVIDIA Deep learning Dataset Synthesizer (NDDS) #### [microsoft/beginners-series-rust](https://github.com/microsoft/beginners-series-rust) *Rust · ★484 · MIT · — · score:0.00 · hot:0.35 · rising:0.42 · durable:0.45 · board:durable · trend:down* Beginner's Series to Rust #### [NVIDIA/G-Assist](https://github.com/NVIDIA/G-Assist) *C · ★224 · Apache-2.0 · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.46 · board:durable · trend:down* Help shape the future of Project G-Assist #### [lucidrains/se3-transformer-pytorch](https://github.com/lucidrains/se3-transformer-pytorch) *Python · ★327 · MIT · — · score:0.00 · hot:0.35 · rising:0.40 · durable:0.48 · board:durable · trend:down* Implementation of SE3-Transformers for Equivariant Self-Attention, in Pytorch. This specific repository is geared towards integration with eventual Alphafold2 replication. #### [huggingface/llm-swarm](https://github.com/huggingface/llm-swarm) *Python · ★287 · MIT · — · score:0.00 · hot:0.35 · rising:0.42 · durable:0.45 · board:durable · trend:down* Manage scalable open LLM inference endpoints in Slurm clusters #### [microsoft/mggraph-intune-samples](https://github.com/microsoft/mggraph-intune-samples) *PowerShell · ★247 · MIT · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.44 · board:durable · trend:down* Repository to hold Microsoft Intune script samples for the Microsoft Graph PowerShell SDK #### [lucidrains/speculative-decoding](https://github.com/lucidrains/speculative-decoding) *Python · ★300 · MIT · — · score:0.00 · hot:0.35 · rising:0.40 · durable:0.52 · board:durable · trend:down* Explorations into some recent techniques surrounding speculative decoding #### [openai/openai-voice-agent-sdk-sample](https://github.com/openai/openai-voice-agent-sdk-sample) *TypeScript · ★252 · MIT · — · score:0.00 · hot:0.35 · rising:0.43 · durable:0.47 · board:durable · trend:down* Sample application to add voice capabilities to the Agents SDK #### [rasbt/machine-learning-notes](https://github.com/rasbt/machine-learning-notes) *Jupyter Notebook · ★838 · BSD-3-Clause · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.43 · board:durable · trend:down* Collection of useful machine learning codes and snippets (originally intended for my personal use) #### [facebookresearch/algonauts-2025](https://github.com/facebookresearch/algonauts-2025) *Python · ★301 · NOASSERTION · — · score:0.00 · hot:0.35 · rising:0.40 · durable:0.44 · board:durable · trend:down* Training and evaluating encoding models to predict fMRI brain responses to naturalistic video stimuli #### [karpathy/randomfun](https://github.com/karpathy/randomfun) *Jupyter Notebook · ★1,164 · no-license · — · score:0.00 · hot:0.35 · rising:0.40 · durable:0.41 · board:durable · trend:down* Notebooks and various random fun #### [lucidrains/perceiver-pytorch](https://github.com/lucidrains/perceiver-pytorch) *Python · ★1,200 · MIT · — · score:0.00 · hot:0.35 · rising:0.42 · durable:0.48 · board:durable · trend:down* Implementation of Perceiver, General Perception with Iterative Attention, in Pytorch #### [lucidrains/performer-pytorch](https://github.com/lucidrains/performer-pytorch) *Python · ★1,177 · MIT · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.47 · board:durable · trend:down* An implementation of Performer, a linear attention-based transformer, in Pytorch #### [SamurAIGPT/AI-Influencer-Generator](https://github.com/SamurAIGPT/AI-Influencer-Generator) *Jupyter Notebook · ★208 · MIT · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.45 · board:durable · trend:down* Create and customize your AI influencer open-source #### [karpathy/lecun1989-repro](https://github.com/karpathy/lecun1989-repro) *Jupyter Notebook · ★748 · MIT · — · score:0.00 · hot:0.35 · rising:0.40 · durable:0.45 · board:durable · trend:down* Reproducing Yann LeCun 1989 paper "Backpropagation Applied to Handwritten Zip Code Recognition", to my knowledge the earliest real-world application of a neural net trained with backpropagation. #### [jina-ai/jina-grep-cli](https://github.com/jina-ai/jina-grep-cli) *Python · ★209 · Apache-2.0 · — · score:0.00 · hot:0.35 · rising:0.38 · durable:0.47 · board:durable · trend:down* Semantic grep powered by Jina embeddings v5 (MLX on Apple Silicon) #### [rasbt/stat453-deep-learning-ss21](https://github.com/rasbt/stat453-deep-learning-ss21) *Jupyter Notebook · ★544 · MIT · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.44 · board:durable · trend:down* STAT 453: Intro to Deep Learning @ UW-Madison (Spring 2021) #### [microsoft/tsyringe](https://github.com/microsoft/tsyringe) *TypeScript · ★5,935 · MIT · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.53 · board:durable · trend:down* Lightweight dependency injection container for JavaScript/TypeScript #### [rasbt/deep-learning-book](https://github.com/rasbt/deep-learning-book) *Jupyter Notebook · ★2,825 · NOASSERTION · — · score:0.00 · hot:0.35 · rising:0.42 · durable:0.45 · board:durable · trend:stable* Repository for "Introduction to Artificial Neural Networks and Deep Learning: A Practical Guide with Applications in Python" #### [microsoft/langchain-for-beginners](https://github.com/microsoft/langchain-for-beginners) *Python · ★365 · MIT · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.46 · board:durable · trend:down* LangChain for Beginners Course #### [google-research/papervizagent](https://github.com/google-research/papervizagent) *Python · ★293 · Apache-2.0 · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.46 · board:durable · trend:down* #### [microsoft/LLaVA-Med](https://github.com/microsoft/LLaVA-Med) *Python · ★2,169 · NOASSERTION · — · score:0.00 · hot:0.35 · rising:0.40 · durable:0.41 · board:durable · trend:stable* Large Language-and-Vision Assistant for Biomedicine, built towards multimodal GPT-4 level capabilities. #### [microsoft/DAViD](https://github.com/microsoft/DAViD) *Python · ★382 · MIT · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.47 · board:durable · trend:down* #### [zjysteven/lmms-finetune](https://github.com/zjysteven/lmms-finetune) *Python · ★372 · Apache-2.0 · — · score:0.00 · hot:0.35 · rising:0.39 · durable:0.45 · board:durable · trend:down* A minimal codebase for finetuning large multimodal models, supporting llava-1.5/1.6, llava-interleave, llava-next-video, llava-onevision, llama-3.2-vision, qwen-vl, qwen2-vl, phi3-v etc. #### [lucidrains/ema-pytorch](https://github.com/lucidrains/ema-pytorch) *Python · ★645 · MIT · — · score:0.00 · hot:0.35 · rising:0.40 · durable:0.50 · board:durable · trend:down* A simple way to keep track of an Exponential Moving Average (EMA) version of your Pytorch model #### [vercel/edge-runtime](https://github.com/vercel/edge-runtime) *TypeScript · ★903 · MIT · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.47 · board:durable · trend:down* Developing, testing, and defining the runtime Web APIs for Edge infrastructure. #### [vercel/ai-sdk-rag-starter](https://github.com/vercel/ai-sdk-rag-starter) *TypeScript · ★261 · no-license · — · score:0.00 · hot:0.35 · rising:0.40 · durable:0.42 · board:durable · trend:down* #### [vercel/arg](https://github.com/vercel/arg) *JavaScript · ★1,293 · MIT · — · score:0.00 · hot:0.35 · rising:0.42 · durable:0.50 · board:durable · trend:down* Simple argument parsing #### [XiaomingX/awesome-qwen-prompt-insight](https://github.com/XiaomingX/awesome-qwen-prompt-insight) *? · ★386 · MIT · — · score:0.00 · hot:0.35 · rising:0.39 · durable:0.46 · board:durable · trend:down* 🧠 世界上覆盖最全的优秀Qwen提示语大全，欢迎贡献你的提示词。🧠 The most comprehensive collection of excellent Qwen prompts in the world. Feel free to contribute your own prompts! #### [replicate/getting-started-nextjs](https://github.com/replicate/getting-started-nextjs) *JavaScript · ★249 · Apache-2.0 · — · score:0.00 · hot:0.35 · rising:0.42 · durable:0.43 · board:durable · trend:down* A Next.js starter app using Replicate #### [microsoft/farmvibes-ai](https://github.com/microsoft/farmvibes-ai) *Jupyter Notebook · ★855 · MIT · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.42 · board:durable · trend:stable* FarmVibes.AI: Multi-Modal GeoSpatial ML Models for Agriculture and Sustainability #### [microsoft/EVA](https://github.com/microsoft/EVA) *C++ · ★255 · MIT · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.44 · board:durable · trend:down* Compiler for the SEAL homomorphic encryption library #### [langchain-ai/deep_research_from_scratch](https://github.com/langchain-ai/deep_research_from_scratch) *Jupyter Notebook · ★704 · no-license · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.45 · board:durable · trend:down* #### [rasbt/stat479-machine-learning-fs18](https://github.com/rasbt/stat479-machine-learning-fs18) *Jupyter Notebook · ★504 · no-license · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.42 · board:durable · trend:down* Course material for STAT 479: Machine Learning (FS 2018) at University Wisconsin-Madison #### [SamurAIGPT/AI-Faceless-Video-Generator](https://github.com/SamurAIGPT/AI-Faceless-Video-Generator) *Jupyter Notebook · ★411 · MIT · — · score:0.00 · hot:0.35 · rising:0.40 · durable:0.45 · board:durable · trend:down* Generate a video script, voice and a talking face completely with AI #### [modal-labs/quillman](https://github.com/modal-labs/quillman) *Python · ★1,203 · MIT · — · score:0.00 · hot:0.35 · rising:0.42 · durable:0.48 · board:durable · trend:down* A voice chat app #### [NVIDIA/semantic-segmentation](https://github.com/NVIDIA/semantic-segmentation) *Python · ★1,818 · BSD-3-Clause · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.42 · board:durable · trend:stable* Nvidia Semantic Segmentation monorepo #### [facebookresearch/fast3r](https://github.com/facebookresearch/fast3r) *Python · ★1,535 · NOASSERTION · — · score:0.00 · hot:0.35 · rising:0.40 · durable:0.44 · board:durable · trend:down* [CVPR 2025] Fast3R: Towards 3D Reconstruction of 1000+ Images in One Forward Pass #### [lucidrains/recurrent-interface-network-pytorch](https://github.com/lucidrains/recurrent-interface-network-pytorch) *Python · ★209 · MIT · — · score:0.00 · hot:0.35 · rising:0.41 · durable:0.48 · board:durable · trend:down* Implementation of Recurrent Interface Network (RIN), for highly efficient generation of images and video without cascading networks, in Pytorch #### [dair-ai/AI-Product-Index](https://github.com/dair-ai/AI-Product-Index) *? · ★779 · MIT · — · score:0.00 · hot:0.34 · rising:0.41 · durable:0.45 · board:durable · trend:down* A curated index to track AI-powered products. #### [rasbt/dora-from-scratch](https://github.com/rasbt/dora-from-scratch) *Jupyter Notebook · ★220 · MIT · — · score:0.00 · hot:0.34 · rising:0.37 · durable:0.44 · board:durable · trend:down* LoRA and DoRA from Scratch Implementations #### [lucidrains/coconut-pytorch](https://github.com/lucidrains/coconut-pytorch) *Python · ★183 · MIT · — · score:0.00 · hot:0.34 · rising:0.39 · durable:0.49 · board:durable · trend:down* Implementation of 🥥 Coconut, Chain of Continuous Thought, in Pytorch #### [replicate/keepsake](https://github.com/replicate/keepsake) *Python · ★1,674 · Apache-2.0 · — · score:0.00 · hot:0.34 · rising:0.41 · durable:0.48 · board:durable · trend:down* Version control for machine learning #### [facebookresearch/DocAgent](https://github.com/facebookresearch/DocAgent) *Python · ★430 · MIT · — · score:0.00 · hot:0.34 · rising:0.41 · durable:0.46 · board:durable · trend:down* DocAgent is a system designed to generate high-quality, context-aware code documentation for Python codebases using a multi-agent approach and hierarchical processing. #### [NVIDIA/partialconv](https://github.com/NVIDIA/partialconv) *Python · ★1,279 · NOASSERTION · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.40 · board:rising · trend:stable* A New Padding Scheme: Partial Convolution based Padding #### [microsoft/vscode-edge-devtools](https://github.com/microsoft/vscode-edge-devtools) *TypeScript · ★815 · MIT · — · score:0.00 · hot:0.34 · rising:0.41 · durable:0.40 · board:rising · trend:stable* A VSCode extension that allows you to use browser devtools from within the editor. The devtools will connect to an instance of Microsoft Edge giving you the ability to alter CSS styling, perform diagnostics, and debugging. Get it now at http://aka.ms/devtools-for-code #### [facebookresearch/PhysicsLM4](https://github.com/facebookresearch/PhysicsLM4) *HTML · ★339 · Apache-2.0 · — · score:0.00 · hot:0.34 · rising:0.41 · durable:0.47 · board:durable · trend:down* Physics of Language Models: Part 4.2, Canon Layers at Scale where Synthetic Pretraining Resonates in Reality #### [huggingface/smol2operator](https://github.com/huggingface/smol2operator) *Python · ★131 · no-license · — · score:0.00 · hot:0.34 · rising:0.41 · durable:0.43 · board:durable · trend:down* #### [karpathy/deep-vector-quantization](https://github.com/karpathy/deep-vector-quantization) *Jupyter Notebook · ★641 · MIT · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.43 · board:durable · trend:down* VQVAEs, GumbelSoftmaxes and friends #### [jxnl/kura](https://github.com/jxnl/kura) *Python · ★447 · no-license · — · score:0.00 · hot:0.34 · rising:0.39 · durable:0.43 · board:durable · trend:down* Kura is a simple reproduction of the CLIO paper which uses language models to label user behaviour before clustering them based on embeddings recursively. This helps us understand user behaviour on a higher level without sacrificing PII. #### [lucidrains/pixel-level-contrastive-learning](https://github.com/lucidrains/pixel-level-contrastive-learning) *Python · ★265 · MIT · — · score:0.00 · hot:0.34 · rising:0.41 · durable:0.47 · board:durable · trend:down* Implementation of Pixel-level Contrastive Learning, proposed in the paper "Propagate Yourself", in Pytorch #### [rasbt/LLM-finetuning-scripts](https://github.com/rasbt/LLM-finetuning-scripts) *Jupyter Notebook · ★219 · Apache-2.0 · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.44 · board:durable · trend:down* #### [NVIDIA/metropolis-nim-workflows](https://github.com/NVIDIA/metropolis-nim-workflows) *Jupyter Notebook · ★196 · Apache-2.0 · — · score:0.00 · hot:0.34 · rising:0.41 · durable:0.46 · board:durable · trend:down* Collection of reference workflows for building intelligent agents with NIMs #### [run-llama/create_llama_projects](https://github.com/run-llama/create_llama_projects) *HTML · ★460 · no-license · — · score:0.00 · hot:0.34 · rising:0.39 · durable:0.42 · board:durable · trend:down* #### [jina-ai/rungpt](https://github.com/jina-ai/rungpt) *Python · ★166 · Apache-2.0 · — · score:0.00 · hot:0.34 · rising:0.41 · durable:0.52 · board:durable · trend:down* An open-source cloud-native of large multi-modal models (LMMs) serving framework. #### [vercel/resumable-stream](https://github.com/vercel/resumable-stream) *TypeScript · ★537 · MIT · — · score:0.00 · hot:0.34 · rising:0.39 · durable:0.41 · board:durable · trend:down* Stream resumption for web streams #### [dair-ai/MLOPs-Primer](https://github.com/dair-ai/MLOPs-Primer) *? · ★972 · no-license · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.44 · board:durable · trend:down* A collection of resources to learn about MLOPs. #### [run-llama/llama-lab](https://github.com/run-llama/llama-lab) *Python · ★1,517 · no-license · — · score:0.00 · hot:0.34 · rising:0.41 · durable:0.45 · board:durable · trend:down* #### [lucidrains/slot-attention](https://github.com/lucidrains/slot-attention) *Python · ★482 · MIT · — · score:0.00 · hot:0.34 · rising:0.41 · durable:0.49 · board:durable · trend:down* Implementation of Slot Attention from GoogleAI #### [anthropics/devcontainer-features](https://github.com/anthropics/devcontainer-features) *Shell · ★244 · MIT · — · score:0.00 · hot:0.34 · rising:0.41 · durable:0.41 · board:durable · trend:down* Anthropic Dev Container Features, including Claude Code CLI #### [jina-ai/meta-prompt](https://github.com/jina-ai/meta-prompt) *HTML · ★181 · Apache-2.0 · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.45 · board:durable · trend:down* For LLMs to better code with Jina API #### [simonw/llm-hacker-news](https://github.com/simonw/llm-hacker-news) *Python · ★127 · Apache-2.0 · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.48 · board:durable · trend:down* LLM plugin for pulling content from Hacker News #### [run-llama/vibe-llama](https://github.com/run-llama/vibe-llama) *Python · ★175 · MIT · — · score:0.00 · hot:0.34 · rising:0.45 · durable:0.54 · board:durable · trend:down* Vibe-coding tools for the LlamaIndex ecosystem #### [jina-ai/agentchain](https://github.com/jina-ai/agentchain) *Python · ★609 · MIT · — · score:0.00 · hot:0.34 · rising:0.39 · durable:0.47 · board:durable · trend:down* Chain together LLMs for reasoning & orchestrate multiple large models for accomplishing complex tasks #### [lucidrains/electra-pytorch](https://github.com/lucidrains/electra-pytorch) *Python · ★236 · MIT · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.47 · board:durable · trend:down* A simple and working implementation of Electra, the fastest way to pretrain language models from scratch, in Pytorch #### [facebookresearch/co3d](https://github.com/facebookresearch/co3d) *Python · ★1,142 · NOASSERTION · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.39 · board:rising · trend:stable* Tooling for the Common Objects In 3D dataset. #### [google-deepmind/jeo](https://github.com/google-deepmind/jeo) *Python · ★159 · Apache-2.0 · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.45 · board:durable · trend:down* Jeo: Jax model training lib for Earth Observation #### [karpathy/recurrentjs](https://github.com/karpathy/recurrentjs) *HTML · ★982 · no-license · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.41 · board:durable · trend:down* Deep Recurrent Neural Networks and LSTMs in Javascript. More generally also arbitrary expression graphs with automatic differentiation. #### [lucidrains/st-moe-pytorch](https://github.com/lucidrains/st-moe-pytorch) *Python · ★381 · MIT · — · score:0.00 · hot:0.34 · rising:0.41 · durable:0.50 · board:durable · trend:down* Implementation of ST-Moe, the latest incarnation of MoE after years of research at Brain, in Pytorch #### [vercel/ai-chatbot-svelte](https://github.com/vercel/ai-chatbot-svelte) *Svelte · ★426 · NOASSERTION · — · score:0.00 · hot:0.34 · rising:0.39 · durable:0.43 · board:durable · trend:down* A full-featured, hackable SvelteKit AI chatbot built by Vercel #### [allenai/OpenBookQA](https://github.com/allenai/OpenBookQA) *Python · ★132 · Apache-2.0 · — · score:0.00 · hot:0.34 · rising:0.41 · durable:0.48 · board:durable · trend:down* Code for experiments on OpenBookQA from the EMNLP 2018 paper "Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answering" #### [microsoft/azure-genai-design-patterns](https://github.com/microsoft/azure-genai-design-patterns) *Python · ★687 · MIT · — · score:0.00 · hot:0.34 · rising:0.41 · durable:0.41 · board:durable · trend:down* #### [karpathy/paper-notes](https://github.com/karpathy/paper-notes) *? · ★710 · no-license · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.40 · board:durable · trend:down* Random notes on papers, likely a short-term repo. #### [jina-ai/jina-video-chat](https://github.com/jina-ai/jina-video-chat) *Python · ★327 · Apache-2.0 · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.44 · board:durable · trend:down* #### [replicate/cog-stable-diffusion](https://github.com/replicate/cog-stable-diffusion) *Python · ★653 · Apache-2.0 · — · score:0.00 · hot:0.34 · rising:0.41 · durable:0.41 · board:durable · trend:stable* Diffusers Stable Diffusion as a Cog model #### [NVIDIA/cuda-checkpoint](https://github.com/NVIDIA/cuda-checkpoint) *C · ★440 · NOASSERTION · — · score:0.00 · hot:0.34 · rising:0.38 · durable:0.39 · board:durable · trend:down* CUDA checkpoint and restore utility #### [chroma-core/context-rot](https://github.com/chroma-core/context-rot) *Python · ★247 · MIT · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.46 · board:durable · trend:down* This repository contains the toolkit for replicating results from our technical report. #### [lucidrains/x-unet](https://github.com/lucidrains/x-unet) *Python · ★291 · MIT · — · score:0.00 · hot:0.34 · rising:0.41 · durable:0.49 · board:durable · trend:down* Implementation of a U-net complete with efficient attention as well as the latest research findings #### [NVIDIA/Audio2Face-3D](https://github.com/NVIDIA/Audio2Face-3D) *? · ★216 · no-license · — · score:0.00 · hot:0.34 · rising:0.39 · durable:0.43 · board:durable · trend:down* repo collection for NVIDIA Audio2Face-3D models and tools #### [facebookresearch/egocentric_splats](https://github.com/facebookresearch/egocentric_splats) *Python · ★137 · NOASSERTION · — · score:0.00 · hot:0.34 · rising:0.39 · durable:0.42 · board:durable · trend:down* The implementation for "Photoreal Scene Reconstruction from an Egocentric Camera, SIGGRAPH 2025" #### [vercel/spr-landing](https://github.com/vercel/spr-landing) *CSS · ★450 · MIT · — · score:0.00 · hot:0.34 · rising:0.41 · durable:0.43 · board:durable · trend:down* Serverless Pre-Rendering Landing Page #### [lucidrains/equiformer-pytorch](https://github.com/lucidrains/equiformer-pytorch) *Python · ★289 · MIT · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.50 · board:durable · trend:down* Implementation of the Equiformer, SE3/E3 equivariant attention network that reaches new SOTA, and adopted for use by EquiFold for protein folding #### [lucidrains/rotary-embedding-torch](https://github.com/lucidrains/rotary-embedding-torch) *Python · ★806 · MIT · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.46 · board:durable · trend:down* Implementation of Rotary Embeddings, from the Roformer paper, in Pytorch #### [milvus-io/milvus-lite](https://github.com/milvus-io/milvus-lite) *Python · ★435 · Apache-2.0 · — · score:0.00 · hot:0.34 · rising:0.41 · durable:0.46 · board:durable · trend:down* A lightweight version of Milvus #### [NVIDIA/CleanUNet](https://github.com/NVIDIA/CleanUNet) *Python · ★351 · MIT · — · score:0.00 · hot:0.34 · rising:0.41 · durable:0.45 · board:durable · trend:down* Official PyTorch Implementation of CleanUNet (ICASSP 2022) #### [langchain-ai/multi-modal-researcher](https://github.com/langchain-ai/multi-modal-researcher) *Python · ★590 · no-license · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.45 · board:durable · trend:down* #### [lucidrains/res-mlp-pytorch](https://github.com/lucidrains/res-mlp-pytorch) *Python · ★200 · MIT · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.49 · board:durable · trend:down* Implementation of ResMLP, an all MLP solution to image classification, in Pytorch #### [lucidrains/byol-pytorch](https://github.com/lucidrains/byol-pytorch) *Python · ★1,878 · MIT · — · score:0.00 · hot:0.34 · rising:0.42 · durable:0.49 · board:durable · trend:down* Usable Implementation of "Bootstrap Your Own Latent" self-supervised learning, from Deepmind, in Pytorch #### [karpathy/researchpooler](https://github.com/karpathy/researchpooler) *Python · ★399 · no-license · — · score:0.00 · hot:0.34 · rising:0.38 · durable:0.38 · board:durable · trend:down* Automating research publications discovery and analysis. For example, ever wish your computer could automatically open papers that are most similar to a paper at an arbitrary url? How about finding all papers that report results on some dataset? Let's re-imagine literature review. #### [huggingface/picotron_tutorial](https://github.com/huggingface/picotron_tutorial) *Python · ★246 · no-license · — · score:0.00 · hot:0.34 · rising:0.39 · durable:0.44 · board:durable · trend:down* #### [microsoft/protein-frame-flow](https://github.com/microsoft/protein-frame-flow) *Python · ★323 · MIT · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.44 · board:durable · trend:down* Fast protein backbone generation with SE(3) flow matching. #### [NVIDIA/nim-anywhere](https://github.com/NVIDIA/nim-anywhere) *Python · ★219 · Apache-2.0 · — · score:0.00 · hot:0.34 · rising:0.37 · durable:0.40 · board:durable · trend:down* Accelerate your Gen AI with NVIDIA NIM and NVIDIA AI Workbench #### [crewAIInc/awesome-crewai](https://github.com/crewAIInc/awesome-crewai) *? · ★483 · MIT · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.41 · board:durable · trend:down* A curated list of open-source projects built by the CrewAI community. Discover, contribute, and extend the possibilities of AI agents with CrewAI. #### [facebookresearch/3detr](https://github.com/facebookresearch/3detr) *Python · ★700 · Apache-2.0 · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.41 · board:durable · trend:down* Code & Models for 3DETR - an End-to-end transformer model for 3D object detection #### [vercel/react-keyframes](https://github.com/vercel/react-keyframes) *TypeScript · ★628 · MIT · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.46 · board:durable · trend:down* Create frame-based animations in React #### [karpathy/hn-time-capsule](https://github.com/karpathy/hn-time-capsule) *Python · ★607 · no-license · — · score:0.00 · hot:0.34 · rising:0.38 · durable:0.40 · board:durable · trend:down* Analyzing Hacker News discussions from a decade ago in hindsight with LLMs #### [huggingface/large_language_model_training_playbook](https://github.com/huggingface/large_language_model_training_playbook) *Python · ★496 · Apache-2.0 · — · score:0.00 · hot:0.34 · rising:0.38 · durable:0.49 · board:durable · trend:down* An open collection of implementation tips, tricks and resources for training large language models #### [simonw/cia-world-factbook-2020](https://github.com/simonw/cia-world-factbook-2020) *HTML · ★205 · no-license · — · score:0.00 · hot:0.34 · rising:0.38 · durable:0.41 · board:durable · trend:down* Recovered cia.gov/the-world-factbook/about/archives/download/factbook-2020.zip from Internet Archive #### [anthropics/claude-code-monitoring-guide](https://github.com/anthropics/claude-code-monitoring-guide) *? · ★265 · no-license · — · score:0.00 · hot:0.34 · rising:0.38 · durable:0.41 · board:durable · trend:down* #### [NVIDIA/trt-samples-for-hackathon-cn](https://github.com/NVIDIA/trt-samples-for-hackathon-cn) *Python · ★1,653 · Apache-2.0 · — · score:0.00 · hot:0.34 · rising:0.41 · durable:0.42 · board:durable · trend:stable* Simple samples for TensorRT programming #### [lucidrains/flash-attention-jax](https://github.com/lucidrains/flash-attention-jax) *Python · ★228 · MIT · — · score:0.00 · hot:0.34 · rising:0.39 · durable:0.47 · board:durable · trend:down* Implementation of Flash Attention in Jax #### [stanford-crfm/ecosystem-graphs](https://github.com/stanford-crfm/ecosystem-graphs) *JavaScript · ★272 · no-license · — · score:0.00 · hot:0.34 · rising:0.39 · durable:0.43 · board:durable · trend:down* #### [facebookresearch/MLGym](https://github.com/facebookresearch/MLGym) *Python · ★594 · NOASSERTION · — · score:0.00 · hot:0.34 · rising:0.39 · durable:0.42 · board:durable · trend:down* MLGym A New Framework and Benchmark for Advancing AI Research Agents #### [archersama/awesome-recommend-system-pretraining-papers](https://github.com/archersama/awesome-recommend-system-pretraining-papers) *? · ★349 · no-license · — · score:0.00 · hot:0.34 · rising:0.37 · durable:0.42 · board:durable · trend:down* Paper List for Recommend-system PreTrained Models #### [lucidrains/magvit2-pytorch](https://github.com/lucidrains/magvit2-pytorch) *Python · ★660 · MIT · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.48 · board:durable · trend:down* Implementation of MagViT2 Tokenizer in Pytorch #### [google-deepmind/md4](https://github.com/google-deepmind/md4) *Python · ★159 · Apache-2.0 · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.45 · board:durable · trend:down* Official Jax Implementation of MD4 Masked Diffusion Models #### [anthropics/political-neutrality-eval](https://github.com/anthropics/political-neutrality-eval) *Python · ★127 · CC-BY-4.0 · — · score:0.00 · hot:0.34 · rising:0.39 · durable:0.44 · board:durable · trend:down* This repo contains detailed implementation information about Anthropic's paired prompts approach for evaluating political neutrality. #### [replicate/inpainter](https://github.com/replicate/inpainter) *JavaScript · ★431 · MIT · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.42 · board:durable · trend:down* A web GUI built with Next.js for inpainting with AI models using Replicate's API #### [google-deepmind/xquad](https://github.com/google-deepmind/xquad) *? · ★209 · no-license · — · score:0.00 · hot:0.34 · rising:0.38 · durable:0.40 · board:durable · trend:down* #### [pydantic/fasta2a](https://github.com/pydantic/fasta2a) *Python · ★183 · MIT · — · score:0.00 · hot:0.34 · rising:0.40 · durable:0.46 · board:durable · trend:down* Convert an AI Agent into a A2A server! ✨ #### [facebookresearch/memory](https://github.com/facebookresearch/memory) *Python · ★376 · NOASSERTION · — · score:0.00 · hot:0.34 · rising:0.39 · durable:0.43 · board:durable · trend:down* Memory layers use a trainable key-value lookup mechanism to add extra parameters to a model without increasing FLOPs. Conceptually, sparsely activated memory layers complement compute-heavy dense feed-forward layers, providing dedicated capacity to store and retrieve information cheaply. #### [NVIDIA/gvdb-voxels](https://github.com/NVIDIA/gvdb-voxels) *C · ★716 · NOASSERTION · — · score:0.00 · hot:0.34 · rising:0.39 · durable:0.36 · board:rising · trend:stable* Sparse volume compute and rendering on NVIDIA GPUs #### [microsoft/DigiFace1M](https://github.com/microsoft/DigiFace1M) *? · ★315 · NOASSERTION · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.42 · board:durable · trend:down* #### [microsoft/aisdkforsapabap](https://github.com/microsoft/aisdkforsapabap) *ABAP · ★143 · MIT · — · score:0.00 · hot:0.33 · rising:0.39 · durable:0.41 · board:durable · trend:down* AI SDK for SAP ABAP #### [simonw/nicar-2025-scraping](https://github.com/simonw/nicar-2025-scraping) *? · ★375 · no-license · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.42 · board:durable · trend:down* Cutting-edge web scraping techniques workshop at NICAR 2025 #### [modal-labs/awesome-modal](https://github.com/modal-labs/awesome-modal) *? · ★187 · no-license · — · score:0.00 · hot:0.33 · rising:0.39 · durable:0.42 · board:durable · trend:down* A curated list of amazingly awesome Modal applications, demos, and shiny things. Inspired by awesome-php. #### [lucidrains/RETRO-pytorch](https://github.com/lucidrains/RETRO-pytorch) *Python · ★879 · Apache-2.0 · — · score:0.00 · hot:0.33 · rising:0.40 · durable:0.50 · board:durable · trend:down* Implementation of RETRO, Deepmind's Retrieval based Attention net, in Pytorch #### [mem0ai/grok3-api](https://github.com/mem0ai/grok3-api) *Python · ★540 · MIT · — · score:0.00 · hot:0.33 · rising:0.40 · durable:0.45 · board:durable · trend:down* Unofficial Grok 3 API #### [huggingface/coreml-examples](https://github.com/huggingface/coreml-examples) *Jupyter Notebook · ★260 · Apache-2.0 · — · score:0.00 · hot:0.33 · rising:0.39 · durable:0.43 · board:durable · trend:down* Swift Core ML Examples #### [google-deepmind/predictingthepast](https://github.com/google-deepmind/predictingthepast) *Python · ★190 · Apache-2.0 · — · score:0.00 · hot:0.33 · rising:0.40 · durable:0.45 · board:durable · trend:down* #### [qdrant/awesome-metric-learning](https://github.com/qdrant/awesome-metric-learning) *? · ★519 · CC0-1.0 · — · score:0.00 · hot:0.33 · rising:0.40 · durable:0.45 · board:durable · trend:down* 😎 A curated list of awesome practical Metric Learning and its applications #### [microsoft/villa-x](https://github.com/microsoft/villa-x) *Python · ★193 · MIT · — · score:0.00 · hot:0.33 · rising:0.39 · durable:0.43 · board:durable · trend:down* #### [NVIDIA/Audio2Face-3D-SDK](https://github.com/NVIDIA/Audio2Face-3D-SDK) *C++ · ★169 · MIT · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.44 · board:durable · trend:down* High-performance C++/CUDA SDK for running Audio2Emotion and Audio2Face inference with integrated post-processing. #### [lucidrains/lumiere-pytorch](https://github.com/lucidrains/lumiere-pytorch) *Python · ★282 · MIT · — · score:0.00 · hot:0.33 · rising:0.39 · durable:0.49 · board:durable · trend:down* Implementation of Lumiere, SOTA text-to-video generation from Google Deepmind, in Pytorch #### [microsoft/sudo](https://github.com/microsoft/sudo) *Rust · ★5,574 · MIT · — · score:0.00 · hot:0.33 · rising:0.39 · durable:0.50 · board:durable · trend:down* It's sudo, for Windows #### [microsoft/demikernel](https://github.com/microsoft/demikernel) *Rust · ★1,217 · MIT · — · score:0.00 · hot:0.33 · rising:0.40 · durable:0.40 · board:durable · trend:down* Kernel-Bypass LibOS Architecture #### [facebookresearch/GeoRT](https://github.com/facebookresearch/GeoRT) *C · ★174 · NOASSERTION · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.40 · board:durable · trend:down* Geometric Retargeting A Principled, Ultrafast Neural Hand Retargeting Algorithm #### [yoheinakajima/babyagi_archive](https://github.com/yoheinakajima/babyagi_archive) *Python · ★100 · MIT · — · score:0.00 · hot:0.33 · rising:0.39 · durable:0.43 · board:durable · trend:down* Snapshot of the original BabyAGI repo as of September 2024 #### [karpathy/ulogme](https://github.com/karpathy/ulogme) *Python · ★1,145 · no-license · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.35 · board:rising · trend:stable* Automatically collect and visualize usage statistics in Ubuntu/OSX environments. #### [facebookresearch/adjoint_sampling](https://github.com/facebookresearch/adjoint_sampling) *Python · ★133 · NOASSERTION · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.42 · board:durable · trend:down* code for "Adjoint Sampling: Highly Scalable Diffusion Samplers via Adjoint Matching" #### [vercel/async-retry](https://github.com/vercel/async-retry) *JavaScript · ★1,914 · MIT · — · score:0.00 · hot:0.33 · rising:0.39 · durable:0.49 · board:durable · trend:down* Retrying made simple, easy and async #### [pydantic/pytest-examples](https://github.com/pydantic/pytest-examples) *Python · ★166 · MIT · — · score:0.00 · hot:0.33 · rising:0.40 · durable:0.47 · board:durable · trend:down* Pytest plugin for testing examples in docstrings and markdown files. #### [yuxin-jiang/Anomagic](https://github.com/yuxin-jiang/Anomagic) *Python · ★141 · no-license · — · score:0.00 · hot:0.33 · rising:0.34 · durable:0.42 · board:durable · trend:down* [AAAI 2026] The Official Implementation for "Anomagic: Crossmodal Prompt-driven Zero-shot Anomaly Generation" #### [replicate/llama-chat](https://github.com/replicate/llama-chat) *JavaScript · ★835 · Apache-2.0 · — · score:0.00 · hot:0.33 · rising:0.40 · durable:0.42 · board:durable · trend:down* A boilerplate for creating a Llama 3 chat app #### [yoheinakajima/babyagi-2o](https://github.com/yoheinakajima/babyagi-2o) *Python · ★335 · MIT · — · score:0.00 · hot:0.33 · rising:0.40 · durable:0.46 · board:durable · trend:down* the simplest self-building general autonomous agent #### [Westlake-AGI-Lab/Awesome-Style-Transfer-with-Diffusion-Models](https://github.com/Westlake-AGI-Lab/Awesome-Style-Transfer-with-Diffusion-Models) *? · ★185 · no-license · — · score:0.00 · hot:0.33 · rising:0.37 · durable:0.41 · board:durable · trend:down* A curated list of recent style transfer methods with diffusion models #### [lucidrains/mmdit](https://github.com/lucidrains/mmdit) *Python · ★516 · MIT · — · score:0.00 · hot:0.33 · rising:0.37 · durable:0.51 · board:durable · trend:down* Implementation of a single layer of the MMDiT, proposed in Stable Diffusion 3, in Pytorch #### [lucidrains/segformer-pytorch](https://github.com/lucidrains/segformer-pytorch) *Python · ★373 · MIT · — · score:0.00 · hot:0.33 · rising:0.40 · durable:0.47 · board:durable · trend:down* Implementation of Segformer, Attention + MLP neural network for segmentation, in Pytorch #### [facebookresearch/volumetric_primitives](https://github.com/facebookresearch/volumetric_primitives) *Python · ★140 · MIT · — · score:0.00 · hot:0.33 · rising:0.39 · durable:0.44 · board:durable · trend:down* This repository contains the implementation of our novel approach associated with the paper "Don't Splat Your Gaussians" to modeling and rendering scattering and emissive media using volumetric primitives with the Mitsuba renderer. #### [openai/moderation-api-release](https://github.com/openai/moderation-api-release) *? · ★159 · MIT · — · score:0.00 · hot:0.33 · rising:0.41 · durable:0.45 · board:durable · trend:down* #### [allenai/fm-cheatsheet](https://github.com/allenai/fm-cheatsheet) *JavaScript · ★271 · no-license · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.43 · board:durable · trend:down* Website for hosting the Open Foundation Models Cheat Sheet. #### [anthropics/claude-ai-mcp](https://github.com/anthropics/claude-ai-mcp) *? · ★209 · NOASSERTION · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.37 · board:rising · trend:down* #### [allenai/bolmo-core](https://github.com/allenai/bolmo-core) *Python · ★131 · Apache-2.0 · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.43 · board:durable · trend:down* Code for Bolmo: Byteifying the Next Generation of Language Models #### [lucidrains/autoregressive-diffusion-pytorch](https://github.com/lucidrains/autoregressive-diffusion-pytorch) *Python · ★436 · MIT · — · score:0.00 · hot:0.33 · rising:0.37 · durable:0.51 · board:durable · trend:down* Implementation of Autoregressive Diffusion in Pytorch #### [karpathy/pytorch-normalizing-flows](https://github.com/karpathy/pytorch-normalizing-flows) *Jupyter Notebook · ★914 · no-license · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.40 · board:durable · trend:down* Normalizing flows in PyTorch. Current intended use is education not production. #### [huggingface/finepdfs](https://github.com/huggingface/finepdfs) *Python · ★181 · NOASSERTION · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.43 · board:durable · trend:down* Codebase for FinePDFs #### [NVIDIA/3DObjectReconstruction](https://github.com/NVIDIA/3DObjectReconstruction) *Python · ★167 · NOASSERTION · — · score:0.00 · hot:0.33 · rising:0.37 · durable:0.42 · board:durable · trend:down* 3D Object Reconstruction project is a workflow that takes a set of stereo images and camera info and outputs a textured mesh (i.e., .OBJ file). The purpose is to translate physical items into the digital world in a photorealistic way #### [facebookresearch/eai-vc](https://github.com/facebookresearch/eai-vc) *Python · ★503 · NOASSERTION · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.40 · board:durable · trend:down* The repository for the largest and most comprehensive empirical study of visual foundation models for Embodied AI (EAI). #### [google-deepmind/alphaevolve_results](https://github.com/google-deepmind/alphaevolve_results) *Jupyter Notebook · ★271 · Apache-2.0 · — · score:0.00 · hot:0.33 · rising:0.40 · durable:0.47 · board:durable · trend:down* #### [allenai/SERA](https://github.com/allenai/SERA) *Python · ★140 · Apache-2.0 · — · score:0.00 · hot:0.33 · rising:0.39 · durable:0.45 · board:durable · trend:down* Data generation and training repository for SERA: Soft-Verified Efficient Repository Agents. #### [lucidrains/PEER-pytorch](https://github.com/lucidrains/PEER-pytorch) *Python · ★136 · MIT · — · score:0.00 · hot:0.33 · rising:0.39 · durable:0.48 · board:durable · trend:down* Pytorch implementation of the PEER block from the paper, Mixture of A Million Experts, by Xu Owen He at Deepmind #### [facebookresearch/CRAG](https://github.com/facebookresearch/CRAG) *Jupyter Notebook · ★281 · NOASSERTION · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.42 · board:durable · trend:down* Comprehensive benchmark for RAG #### [modal-labs/doppel-bot](https://github.com/modal-labs/doppel-bot) *Python · ★262 · MIT · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.45 · board:durable · trend:down* Train a language model to answer Slack messages as you. #### [run-llama/fs-explorer](https://github.com/run-llama/fs-explorer) *Python · ★161 · no-license · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.43 · board:durable · trend:down* CLI agent to explore file system, powered by Gemini 3 Flash #### [allenai/sera-cli](https://github.com/allenai/sera-cli) *Python · ★238 · Apache-2.0 · — · score:0.00 · hot:0.33 · rising:0.39 · durable:0.46 · board:durable · trend:down* A tool to use the Ai2 Open Coding Agents Soft-Verified Efficient Repository Agents (SERA) model with Claude Code #### [rasbt/RAGs](https://github.com/rasbt/RAGs) *Jupyter Notebook · ★151 · Apache-2.0 · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.43 · board:durable · trend:down* RAGs: Simple implementations of Retrieval Augmented Generation (RAG) Systems #### [huggingface/fineweb-2](https://github.com/huggingface/fineweb-2) *Python · ★234 · Apache-2.0 · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.44 · board:durable · trend:down* #### [huggingface/olm-datasets](https://github.com/huggingface/olm-datasets) *Python · ★179 · Apache-2.0 · — · score:0.00 · hot:0.33 · rising:0.39 · durable:0.44 · board:durable · trend:down* Pipeline for pulling and processing online language model pretraining data from the web #### [vercel/title](https://github.com/vercel/title) *JavaScript · ★615 · MIT · — · score:0.00 · hot:0.33 · rising:0.41 · durable:0.47 · board:durable · trend:down* A service for capitalizing your title properly #### [karpathy/tsnejs](https://github.com/karpathy/tsnejs) *JavaScript · ★909 · no-license · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.39 · board:durable · trend:down* Implementation of t-SNE visualization algorithm in Javascript. #### [lucidrains/lion-pytorch](https://github.com/lucidrains/lion-pytorch) *Python · ★2,182 · MIT · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.54 · board:durable · trend:down* 🦁 Lion, new optimizer discovered by Google Brain using genetic algorithms that is purportedly better than Adam(w), in Pytorch #### [google-deepmind/perception_test](https://github.com/google-deepmind/perception_test) *Jupyter Notebook · ★247 · Apache-2.0 · — · score:0.00 · hot:0.33 · rising:0.39 · durable:0.44 · board:durable · trend:down* #### [facebookresearch/foundpose](https://github.com/facebookresearch/foundpose) *Python · ★124 · NOASSERTION · — · score:0.00 · hot:0.33 · rising:0.37 · durable:0.40 · board:durable · trend:down* FoundPose: Unseen Object Pose Estimation with Foundation Features, ECCV 2024 #### [run-llama/voice-chat-pdf](https://github.com/run-llama/voice-chat-pdf) *JavaScript · ★329 · MIT · — · score:0.00 · hot:0.33 · rising:0.40 · durable:0.46 · board:durable · trend:down* Use OpenAI's realtime API for a chatting with your documents #### [taylorwilsdon/open-webui-embeddable-widget](https://github.com/taylorwilsdon/open-webui-embeddable-widget) *Svelte · ★112 · no-license · — · score:0.00 · hot:0.33 · rising:0.35 · durable:0.38 · board:durable · trend:down* Lightweight, simple embedded Open WebUI widget, allowing you to easily implement chatbot capabilities and RAG workflows into your existing tools, apps and webpages! #### [openai/spinningup-workshop](https://github.com/openai/spinningup-workshop) *TeX · ★206 · no-license · — · score:0.00 · hot:0.33 · rising:0.40 · durable:0.42 · board:durable · trend:down* For educational materials related to the spinning up workshops. #### [swyxio/devtools-angels](https://github.com/swyxio/devtools-angels) *? · ★640 · no-license · — · score:0.00 · hot:0.33 · rising:0.37 · durable:0.36 · board:rising · trend:down* active angel investors in developer tools! #### [microsoft/MPNet](https://github.com/microsoft/MPNet) *Python · ★299 · MIT · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.41 · board:durable · trend:down* MPNet: Masked and Permuted Pre-training for Language Understanding https://arxiv.org/pdf/2004.09297.pdf #### [google-deepmind/transformer_grammars](https://github.com/google-deepmind/transformer_grammars) *Python · ★136 · Apache-2.0 · — · score:0.00 · hot:0.33 · rising:0.37 · durable:0.37 · board:durable · trend:down* Transformer Grammars: Augmenting Transformer Language Models with Syntactic Inductive Biases at Scale, TACL (2022) #### [facebookresearch/dadaptation](https://github.com/facebookresearch/dadaptation) *Python · ★531 · MIT · — · score:0.00 · hot:0.33 · rising:0.39 · durable:0.46 · board:durable · trend:down* D-Adaptation for SGD, Adam and AdaGrad #### [karpathy/karpathy](https://github.com/karpathy/karpathy) *? · ★110 · no-license · — · score:0.00 · hot:0.33 · rising:0.36 · durable:0.38 · board:durable · trend:down* root repo #### [simonw/git-history](https://github.com/simonw/git-history) *Python · ★222 · Apache-2.0 · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.43 · board:durable · trend:down* Tools for analyzing Git history using SQLite #### [openai/openai-support-agent-demo](https://github.com/openai/openai-support-agent-demo) *TypeScript · ★182 · NOASSERTION · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.43 · board:durable · trend:down* Demo of a customer support agent interface using NextJS and the OpenAI Responses API with File Search #### [simonw/google-drive-to-sqlite](https://github.com/simonw/google-drive-to-sqlite) *Python · ★164 · Apache-2.0 · — · score:0.00 · hot:0.33 · rising:0.38 · durable:0.45 · board:durable · trend:down* Create a SQLite database containing metadata from Google Drive #### [huggingface/large-scale-image-deduplication](https://github.com/huggingface/large-scale-image-deduplication) *Python · ★200 · no-license · — · score:0.00 · hot:0.33 · rising:0.37 · durable:0.43 · board:durable · trend:down* #### [Event-AHU/Medical_Image_Analysis](https://github.com/Event-AHU/Medical_Image_Analysis) *Python · ★221 · BSD-3-Clause · — · score:0.00 · hot:0.33 · rising:0.36 · durable:0.44 · board:durable · trend:down* Foundation models based medical image analysis #### [karpathy/covid-sanity](https://github.com/karpathy/covid-sanity) *Python · ★391 · MIT · — · score:0.00 · hot:0.33 · rising:0.39 · durable:0.42 · board:durable · trend:down* Aspires to help the influx of bioRxiv / medRxiv papers on COVID-19 #### [taylorwilsdon/netshow](https://github.com/taylorwilsdon/netshow) *Python · ★361 · NOASSERTION · — · score:0.00 · hot:0.32 · rising:0.36 · durable:0.48 · board:durable · trend:down* Lightweight, performant interactive network connection monitor with friendly service names #### [lucidrains/sinkhorn-transformer](https://github.com/lucidrains/sinkhorn-transformer) *Python · ★270 · MIT · — · score:0.00 · hot:0.32 · rising:0.38 · durable:0.50 · board:durable · trend:down* Sinkhorn Transformer - Practical implementation of Sparse Sinkhorn Attention #### [jina-ai/cli](https://github.com/jina-ai/cli) *Python · ★125 · Apache-2.0 · — · score:0.00 · hot:0.32 · rising:0.33 · durable:0.45 · board:durable · trend:down* All Jina AI APIs as Unix CLI commands. Search, read, embed, rerank - with pipes. #### [weaviate/mcp-server-weaviate](https://github.com/weaviate/mcp-server-weaviate) *Go · ★161 · no-license · — · score:0.00 · hot:0.32 · rising:0.38 · durable:0.41 · board:durable · trend:down* MCP (Model Context Protocol) server for Weaviate #### [google-deepmind/language_modeling_is_compression](https://github.com/google-deepmind/language_modeling_is_compression) *Python · ★177 · Apache-2.0 · — · score:0.00 · hot:0.32 · rising:0.39 · durable:0.45 · board:durable · trend:down* #### [modal-labs/devlooper](https://github.com/modal-labs/devlooper) *Python · ★468 · MIT · — · score:0.00 · hot:0.32 · rising:0.38 · durable:0.48 · board:durable · trend:down* A program synthesis agent that autonomously fixes its output by running tests! #### [google-research/mt-metrics-eval](https://github.com/google-research/mt-metrics-eval) *Python · ★129 · Apache-2.0 · — · score:0.00 · hot:0.32 · rising:0.38 · durable:0.41 · board:durable · trend:down* Tools for evaluating the performance of MT metrics on data from recent WMT metrics shared tasks. #### [microsoft/CSWin-Transformer](https://github.com/microsoft/CSWin-Transformer) *Python · ★586 · MIT · — · score:0.00 · hot:0.32 · rising:0.38 · durable:0.39 · board:durable · trend:down* CSWin Transformer: A General Vision Transformer Backbone with Cross-Shaped, CVPR 2022 #### [allenai/scifact](https://github.com/allenai/scifact) *Python · ★255 · NOASSERTION · — · score:0.00 · hot:0.32 · rising:0.37 · durable:0.39 · board:durable · trend:down* Data and models for the SciFact verification task. #### [facebookresearch/ScaDiver](https://github.com/facebookresearch/ScaDiver) *Python · ★165 · NOASSERTION · — · score:0.00 · hot:0.32 · rising:0.37 · durable:0.39 · board:durable · trend:down* Project for the paper "A Scalable Approach to Control Diverse Behaviors for Physically Simulated Characters" #### [simonw/llm-llama-cpp](https://github.com/simonw/llm-llama-cpp) *Python · ★145 · Apache-2.0 · — · score:0.00 · hot:0.32 · rising:0.38 · durable:0.44 · board:durable · trend:down* LLM plugin for running models using llama.cpp #### [karpathy/find-birds](https://github.com/karpathy/find-birds) *Python · ★362 · no-license · — · score:0.00 · hot:0.32 · rising:0.36 · durable:0.38 · board:durable · trend:down* Find people you should follow on Twitter based on who the people you follow follow #### [dair-ai/awesome-ML-projects-guide](https://github.com/dair-ai/awesome-ML-projects-guide) *? · ★248 · no-license · — · score:0.00 · hot:0.32 · rising:0.37 · durable:0.40 · board:durable · trend:down* A guide to building awesome machine learning projects. #### [microsoft/vscode-data-wrangler](https://github.com/microsoft/vscode-data-wrangler) *? · ★591 · NOASSERTION · — · score:0.00 · hot:0.32 · rising:0.37 · durable:0.37 · board:durable · trend:down* Data Wrangler extension for Visual Studio Code #### [huggingface/OBELICS](https://github.com/huggingface/OBELICS) *Python · ★213 · Apache-2.0 · — · score:0.00 · hot:0.32 · rising:0.37 · durable:0.43 · board:durable · trend:down* Code used for the creation of OBELICS, an open, massive and curated collection of interleaved image-text web documents, containing 141M documents, 115B text tokens and 353M images. #### [microsoft/SmartPlay](https://github.com/microsoft/SmartPlay) *Python · ★146 · CC-BY-4.0 · — · score:0.00 · hot:0.32 · rising:0.37 · durable:0.42 · board:durable · trend:down* SmartPlay is a benchmark for Large Language Models (LLMs). Uses a variety of games to test various important LLM capabilities as agents. SmartPlay is designed to be easy to use, and to support future development of LLMs. #### [simonw/click-app](https://github.com/simonw/click-app) *Python · ★408 · Apache-2.0 · — · score:0.00 · hot:0.32 · rising:0.37 · durable:0.42 · board:durable · trend:down* Cookiecutter template for creating new Click command-line tools #### [microsoft/LLM2CLIP](https://github.com/microsoft/LLM2CLIP) *Python · ★647 · MIT · — · score:0.00 · hot:0.32 · rising:0.36 · durable:0.41 · board:durable · trend:down* LLM2CLIP significantly improves already state-of-the-art CLIP models. #### [rasbt/algorithms_in_ipython_notebooks](https://github.com/rasbt/algorithms_in_ipython_notebooks) *Jupyter Notebook · ★512 · GPL-3.0 · — · score:0.00 · hot:0.32 · rising:0.40 · durable:0.43 · board:durable · trend:down* A repository with IPython notebooks of algorithms implemented in Python. #### [simonw/csvs-to-sqlite](https://github.com/simonw/csvs-to-sqlite) *Python · ★926 · Apache-2.0 · — · score:0.00 · hot:0.32 · rising:0.39 · durable:0.44 · board:durable · trend:down* Convert CSV files into a SQLite database #### [dair-ai/m2-deep-research](https://github.com/dair-ai/m2-deep-research) *Python · ★206 · no-license · — · score:0.00 · hot:0.32 · rising:0.36 · durable:0.41 · board:durable · trend:down* Deep research agents using MiniMax M2.1 interleaved thinking #### [yoheinakajima/prettygraph](https://github.com/yoheinakajima/prettygraph) *HTML · ★779 · MIT · — · score:0.00 · hot:0.32 · rising:0.38 · durable:0.44 · board:durable · trend:down* An experimental UI for text-to-knowledge-graph generation #### [lucidrains/TimeSformer-pytorch](https://github.com/lucidrains/TimeSformer-pytorch) *Python · ★729 · MIT · — · score:0.00 · hot:0.32 · rising:0.39 · durable:0.49 · board:durable · trend:down* Implementation of TimeSformer from Facebook AI, a pure attention-based solution for video classification #### [vercel/opentelemetry-collector-dev-setup](https://github.com/vercel/opentelemetry-collector-dev-setup) *Shell · ★159 · Apache-2.0 · — · score:0.00 · hot:0.32 · rising:0.38 · durable:0.43 · board:durable · trend:down* #### [google-deepmind/digraph_transformer](https://github.com/google-deepmind/digraph_transformer) *Python · ★122 · NOASSERTION · — · score:0.00 · hot:0.32 · rising:0.36 · durable:0.40 · board:durable · trend:down* #### [facebookresearch/ava-256](https://github.com/facebookresearch/ava-256) *Jupyter Notebook · ★215 · NOASSERTION · — · score:0.00 · hot:0.32 · rising:0.36 · durable:0.38 · board:durable · trend:down* Train universal codec avatars #### [dair-ai/dair-ai.github.io](https://github.com/dair-ai/dair-ai.github.io) *HTML · ★245 · MIT · — · score:0.00 · hot:0.32 · rising:0.37 · durable:0.38 · board:durable · trend:down* Home of DAIR.AI #### [vercel/nextjs-portfolio-starter](https://github.com/vercel/nextjs-portfolio-starter) *JavaScript · ★730 · no-license · — · score:0.00 · hot:0.32 · rising:0.37 · durable:0.38 · board:durable · trend:down* Easily create a portfolio with Next.js and Markdown. #### [BerriAI/liteLLM-proxy](https://github.com/BerriAI/liteLLM-proxy) *Python · ★208 · MIT · — · score:0.00 · hot:0.32 · rising:0.37 · durable:0.40 · board:durable · trend:down* #### [simonw/pelican-bicycle](https://github.com/simonw/pelican-bicycle) *Shell · ★163 · no-license · — · score:0.00 · hot:0.32 · rising:0.35 · durable:0.39 · board:durable · trend:down* LLM benchmark: Generate an SVG of a pelican riding a bicycle #### [HorizonWind2004/reconstruction-alignment](https://github.com/HorizonWind2004/reconstruction-alignment) *Python · ★388 · Apache-2.0 · — · score:0.00 · hot:0.32 · rising:0.35 · durable:0.44 · board:durable · trend:down* [ICLR 2026] Official repo of paper "Reconstruction Alignment Improves Unified Multimodal Models". Unlocking the Massive Zero-shot Potential in Unified Multimodal Models through Self-supervised Learning. #### [replicate/quirky](https://github.com/replicate/quirky) *JavaScript · ★223 · no-license · — · score:0.00 · hot:0.32 · rising:0.37 · durable:0.41 · board:durable · trend:down* An open-source tool for making really cool QR codes with AI #### [lucidrains/clinical-calculator-tooluse](https://github.com/lucidrains/clinical-calculator-tooluse) *Python · ★315 · MIT · — · score:0.00 · hot:0.32 · rising:0.38 · durable:0.45 · board:durable · trend:down* Explorations into training LLMs to use clinical calculators from patient history, using open sourced models. Will start with Wells' Criteria #### [facebookresearch/Mixture-of-Transformers](https://github.com/facebookresearch/Mixture-of-Transformers) *Python · ★198 · BSD-3-Clause · — · score:0.00 · hot:0.32 · rising:0.37 · durable:0.45 · board:durable · trend:down* Mixture-of-Transformers: A Sparse and Scalable Architecture for Multi-Modal Foundation Models. TMLR 2025. #### [kehanlu/DeSTA2.5-Audio](https://github.com/kehanlu/DeSTA2.5-Audio) *Python · ★137 · no-license · — · score:0.00 · hot:0.32 · rising:0.35 · durable:0.39 · board:durable · trend:down* Code for DeSTA2.5-Audio, general-purpose LALM #### [karpathy/EigenLibSVM](https://github.com/karpathy/EigenLibSVM) *C++ · ★109 · no-license · — · score:0.00 · hot:0.32 · rising:0.36 · durable:0.36 · board:durable · trend:down* A wrapper for LibSVM that lets you train SVM's directly on Eigen library matrices in C++ #### [google-research/optformer](https://github.com/google-research/optformer) *Python · ★242 · Apache-2.0 · — · score:0.00 · hot:0.32 · rising:0.40 · durable:0.49 · board:durable · trend:down* #### [karpathy/researchlei](https://github.com/karpathy/researchlei) *Python · ★257 · no-license · — · score:0.00 · hot:0.32 · rising:0.37 · durable:0.37 · board:durable · trend:down* An Academic Papers Management and Discovery System #### [modal-labs/turbo-art](https://github.com/modal-labs/turbo-art) *Svelte · ★238 · no-license · — · score:0.00 · hot:0.32 · rising:0.36 · durable:0.40 · board:durable · trend:down* A playground for creative exploration that uses SDXL Turbo. #### [facebookresearch/digit-design](https://github.com/facebookresearch/digit-design) *Python · ★206 · NOASSERTION · — · score:0.00 · hot:0.32 · rising:0.36 · durable:0.36 · board:rising · trend:down* Design files for the DIGIT tactile sensor #### [karpathy/nipspreview](https://github.com/karpathy/nipspreview) *Python · ★185 · no-license · — · score:0.00 · hot:0.32 · rising:0.37 · durable:0.37 · board:durable · trend:down* Scripts that generate .html to more easily see NIPS papers #### [jina-ai/thinkgpt](https://github.com/jina-ai/thinkgpt) *Python · ★1,579 · Apache-2.0 · — · score:0.00 · hot:0.32 · rising:0.40 · durable:0.47 · board:durable · trend:down* Agent techniques to augment your LLM and push it beyong its limits #### [google-research/parti](https://github.com/google-research/parti) *? · ★1,590 · Apache-2.0 · — · score:0.00 · hot:0.32 · rising:0.41 · durable:0.48 · board:durable · trend:down* #### [facebookresearch/SpinQuant](https://github.com/facebookresearch/SpinQuant) *Python · ★387 · NOASSERTION · — · score:0.00 · hot:0.32 · rising:0.36 · durable:0.38 · board:durable · trend:down* Code repo for the paper "SpinQuant LLM quantization with learned rotations" #### [huggingface/datablations](https://github.com/huggingface/datablations) *Jupyter Notebook · ★343 · Apache-2.0 · — · score:0.00 · hot:0.32 · rising:0.38 · durable:0.45 · board:durable · trend:down* Scaling Data-Constrained Language Models #### [huggingface/VLAb](https://github.com/huggingface/VLAb) *Python · ★161 · NOASSERTION · — · score:0.00 · hot:0.32 · rising:0.36 · durable:0.42 · board:durable · trend:down* #### [allenai/scidocs](https://github.com/allenai/scidocs) *Python · ★146 · NOASSERTION · — · score:0.00 · hot:0.32 · rising:0.36 · durable:0.39 · board:durable · trend:down* Dataset accompanying the SPECTER model #### [huggingface/sharp-transformers](https://github.com/huggingface/sharp-transformers) *C# · ★197 · Apache-2.0 · — · score:0.00 · hot:0.32 · rising:0.38 · durable:0.42 · board:durable · trend:down* A Unity plugin for using Transformers models in Unity. #### [allenai/satlaspretrain_models](https://github.com/allenai/satlaspretrain_models) *Jupyter Notebook · ★140 · Apache-2.0 · — · score:0.00 · hot:0.32 · rising:0.37 · durable:0.42 · board:durable · trend:down* #### [lucidrains/block-recurrent-transformer-pytorch](https://github.com/lucidrains/block-recurrent-transformer-pytorch) *Python · ★224 · MIT · — · score:0.00 · hot:0.32 · rising:0.38 · durable:0.51 · board:durable · trend:down* Implementation of Block Recurrent Transformer - Pytorch #### [karpathy/calorie](https://github.com/karpathy/calorie) *HTML · ★142 · no-license · — · score:0.00 · hot:0.32 · rising:0.35 · durable:0.39 · board:durable · trend:down* nice and effective super simple calorie counter web app #### [swyxio/swyxkit](https://github.com/swyxio/swyxkit) *Svelte · ★712 · MIT · — · score:0.00 · hot:0.32 · rising:0.36 · durable:0.38 · board:durable · trend:down* An opinionated blog starter for SvelteKit + Tailwind + Netlify. Refreshed for SvelteKit 1.0! #### [taylorwilsdon/reclaimed](https://github.com/taylorwilsdon/reclaimed) *Python · ★134 · MIT · — · score:0.00 · hot:0.32 · rising:0.36 · durable:0.49 · board:durable · trend:down* Fast, light and useful - reclaimed is a disk space utilization & cleanup application for macOS, Linux & Windows #### [weaviate/healthsearch-demo](https://github.com/weaviate/healthsearch-demo) *TypeScript · ★181 · MIT · — · score:0.00 · hot:0.32 · rising:0.38 · durable:0.44 · board:durable · trend:down* Discover Healthsearch: Unlocking Health with Semantic Search ✨ #### [allenai/medicat](https://github.com/allenai/medicat) *Python · ★170 · Apache-2.0 · — · score:0.00 · hot:0.32 · rising:0.37 · durable:0.41 · board:durable · trend:down* Dataset of medical images, captions, subfigure-subcaption annotations, and inline textual references #### [replicate/cog-llama-template](https://github.com/replicate/cog-llama-template) *Python · ★302 · Apache-2.0 · — · score:0.00 · hot:0.32 · rising:0.38 · durable:0.39 · board:durable · trend:down* LLaMA Cog template #### [NVIDIA/Deep-Learning-Accelerator-SW](https://github.com/NVIDIA/Deep-Learning-Accelerator-SW) *Python · ★231 · NOASSERTION · — · score:0.00 · hot:0.32 · rising:0.36 · durable:0.39 · board:durable · trend:down* NVIDIA DLA-SW, the recipes and tools for running deep learning workloads on NVIDIA DLA cores for inference applications. #### [lucidrains/jax2torch](https://github.com/lucidrains/jax2torch) *Python · ★262 · MIT · — · score:0.00 · hot:0.31 · rising:0.36 · durable:0.48 · board:durable · trend:down* Use Jax functions in Pytorch #### [facebookresearch/dynamic_stereo](https://github.com/facebookresearch/dynamic_stereo) *Jupyter Notebook · ★233 · NOASSERTION · — · score:0.00 · hot:0.31 · rising:0.36 · durable:0.39 · board:durable · trend:down* [CVPR 2023] DynamicStereo: Consistent Dynamic Depth from Stereo Videos. #### [simonw/django-sql-dashboard](https://github.com/simonw/django-sql-dashboard) *Python · ★465 · Apache-2.0 · — · score:0.00 · hot:0.31 · rising:0.38 · durable:0.44 · board:durable · trend:down* Django app for building dashboards using raw SQL queries #### [allenai/multimodalqa](https://github.com/allenai/multimodalqa) *Python · ★154 · no-license · — · score:0.00 · hot:0.31 · rising:0.36 · durable:0.38 · board:durable · trend:down* #### [simonw/shot-scraper-template](https://github.com/simonw/shot-scraper-template) *? · ★262 · no-license · — · score:0.00 · hot:0.31 · rising:0.35 · durable:0.40 · board:durable · trend:down* Template repository for setting up shot-scraper #### [google-research/proteinfer](https://github.com/google-research/proteinfer) *Jupyter Notebook · ★188 · Apache-2.0 · — · score:0.00 · hot:0.31 · rising:0.36 · durable:0.36 · board:durable · trend:down* Deep networks for protein functional inference #### [lucidrains/progen](https://github.com/lucidrains/progen) *Python · ★114 · MIT · — · score:0.00 · hot:0.31 · rising:0.38 · durable:0.47 · board:durable · trend:down* Implementation and replication of ProGen, Language Modeling for Protein Generation, in Jax #### [lucidrains/q-transformer](https://github.com/lucidrains/q-transformer) *Python · ★405 · MIT · — · score:0.00 · hot:0.31 · rising:0.38 · durable:0.49 · board:durable · trend:down* Implementation of Q-Transformer, Scalable Offline Reinforcement Learning via Autoregressive Q-Functions, out of Google Deepmind #### [lucidrains/routing-transformer](https://github.com/lucidrains/routing-transformer) *Python · ★300 · MIT · — · score:0.00 · hot:0.31 · rising:0.37 · durable:0.48 · board:durable · trend:down* Fully featured implementation of Routing Transformer #### [lucidrains/x-clip](https://github.com/lucidrains/x-clip) *Python · ★722 · MIT · — · score:0.00 · hot:0.31 · rising:0.40 · durable:0.50 · board:durable · trend:down* A concise but complete implementation of CLIP with various experimental improvements from recent papers #### [vercel/sveltekit-commerce](https://github.com/vercel/sveltekit-commerce) *Svelte · ★442 · MIT · — · score:0.00 · hot:0.31 · rising:0.39 · durable:0.44 · board:durable · trend:down* SvelteKit Commerce #### [weaviate/weaviate-examples](https://github.com/weaviate/weaviate-examples) *HTML · ★329 · MIT · — · score:0.00 · hot:0.31 · rising:0.38 · durable:0.42 · board:durable · trend:down* Weaviate vector database – examples #### [allenai/WildBench](https://github.com/allenai/WildBench) *Python · ★248 · Apache-2.0 · — · score:0.00 · hot:0.31 · rising:0.38 · durable:0.45 · board:durable · trend:down* Benchmarking LLMs with Challenging Tasks from Real Users #### [openai/consistency_models_cifar10](https://github.com/openai/consistency_models_cifar10) *Jupyter Notebook · ★152 · Apache-2.0 · — · score:0.00 · hot:0.31 · rising:0.36 · durable:0.42 · board:durable · trend:down* Consistency models trained on CIFAR-10, in JAX. #### [rasbt/cvpr2023](https://github.com/rasbt/cvpr2023) *Python · ★130 · Apache-2.0 · — · score:0.00 · hot:0.31 · rising:0.36 · durable:0.41 · board:durable · trend:down* #### [replicate/replicate-swift](https://github.com/replicate/replicate-swift) *Swift · ★189 · Apache-2.0 · — · score:0.00 · hot:0.31 · rising:0.40 · durable:0.50 · board:durable · trend:down* Swift client for Replicate #### [Li-Jinsong/DAEDAL](https://github.com/Li-Jinsong/DAEDAL) *Python · ★165 · Apache-2.0 · — · score:0.00 · hot:0.31 · rising:0.34 · durable:0.43 · board:durable · trend:down* [ICLR 2026] Official repository of "Beyond Fixed: Training-Free Variable-Length Denoising for Diffusion Large Language Models" #### [lucidrains/fast-transformer-pytorch](https://github.com/lucidrains/fast-transformer-pytorch) *Python · ★176 · MIT · — · score:0.00 · hot:0.31 · rising:0.38 · durable:0.50 · board:durable · trend:down* Implementation of Fast Transformer in Pytorch #### [karpathy/gitstats](https://github.com/karpathy/gitstats) *HTML · ★136 · no-license · — · score:0.00 · hot:0.31 · rising:0.34 · durable:0.37 · board:durable · trend:down* A lightweight/pretty visualizer for recent work on a git code base in multiple branches. Helps stay up to date with teams working on one git repo in many branches. #### [replicate/cog-comfyui](https://github.com/replicate/cog-comfyui) *Python · ★765 · MIT · — · score:0.00 · hot:0.31 · rising:0.38 · durable:0.41 · board:durable · trend:down* Run ComfyUI with an API #### [lucidrains/ETSformer-pytorch](https://github.com/lucidrains/ETSformer-pytorch) *Python · ★155 · MIT · — · score:0.00 · hot:0.31 · rising:0.38 · durable:0.49 · board:durable · trend:down* Implementation of ETSformer, state of the art time-series Transformer, in Pytorch #### [openai/chz](https://github.com/openai/chz) *Python · ★229 · MIT · — · score:0.00 · hot:0.31 · rising:0.38 · durable:0.50 · board:durable · trend:down* #### [taylorwilsdon/reddacted](https://github.com/taylorwilsdon/reddacted) *Python · ★122 · MIT · — · score:0.00 · hot:0.31 · rising:0.38 · durable:0.50 · board:durable · trend:down* reddacted lets you analyze & sanitize your online footprint using LLMs, PII detection & sentiment analysis to identify anything that might reveal personal info you may not want correlated with your anonymous profile #### [lucidrains/axial-attention](https://github.com/lucidrains/axial-attention) *Python · ★396 · MIT · — · score:0.00 · hot:0.31 · rising:0.38 · durable:0.49 · board:durable · trend:down* Implementation of Axial attention - attending to multi-dimensional data efficiently #### [allenai/scicite](https://github.com/allenai/scicite) *Python · ★129 · Apache-2.0 · — · score:0.00 · hot:0.31 · rising:0.37 · durable:0.38 · board:durable · trend:down* Repository for NAACL 2019 paper on Citation Intent prediction #### [microsoft/rust-guidelines](https://github.com/microsoft/rust-guidelines) *Rust · ★151 · MIT · — · score:0.00 · hot:0.31 · rising:0.36 · durable:0.38 · board:durable · trend:down* Write idiomatic Rust that scales. #### [openai/safety-rbr-code-and-data](https://github.com/openai/safety-rbr-code-and-data) *Jupyter Notebook · ★208 · MIT · — · score:0.00 · hot:0.31 · rising:0.40 · durable:0.47 · board:durable · trend:down* Code and example data for the paper: Rule Based Rewards for Language Model Safety #### [vercel/title-site](https://github.com/vercel/title-site) *JavaScript · ★137 · MIT · — · score:0.00 · hot:0.31 · rising:0.40 · durable:0.48 · board:durable · trend:down* A website for capitalizing your titles #### [bilalonur/awesome-llm-os](https://github.com/bilalonur/awesome-llm-os) *? · ★158 · CC0-1.0 · — · score:0.00 · hot:0.31 · rising:0.35 · durable:0.44 · board:durable · trend:down* A curated list of awesome resources, tools, research papers, and projects related to the concept of Large Language Model Operating Systems (LLM-OS). #### [vercel/server-components-notes-demo](https://github.com/vercel/server-components-notes-demo) *TypeScript · ★749 · MIT · — · score:0.00 · hot:0.31 · rising:0.40 · durable:0.47 · board:durable · trend:down* Demo of React Server Components with Next.js. Deployed on Vercel. #### [karpathy/pytorch-made](https://github.com/karpathy/pytorch-made) *Python · ★593 · no-license · — · score:0.00 · hot:0.31 · rising:0.37 · durable:0.40 · board:durable · trend:down* MADE (Masked Autoencoder Density Estimation) implementation in PyTorch #### [karpathy/tf-agent](https://github.com/karpathy/tf-agent) *Python · ★144 · no-license · — · score:0.00 · hot:0.31 · rising:0.35 · durable:0.36 · board:durable · trend:down* tensorflow reinforcement learning agents for OpenAI gym environments #### [vercel/react-transition-progress](https://github.com/vercel/react-transition-progress) *TypeScript · ★325 · MIT · — · score:0.00 · hot:0.31 · rising:0.36 · durable:0.40 · board:durable · trend:down* Show a progress bar while React Transitions run #### [lucidrains/Mega-pytorch](https://github.com/lucidrains/Mega-pytorch) *Python · ★207 · MIT · — · score:0.00 · hot:0.31 · rising:0.37 · durable:0.50 · board:durable · trend:down* Implementation of Mega, the Single-head Attention with Multi-headed EMA architecture that currently holds SOTA on Long Range Arena #### [allenai/objaverse-rendering](https://github.com/allenai/objaverse-rendering) *Python · ★262 · Apache-2.0 · — · score:0.00 · hot:0.31 · rising:0.36 · durable:0.41 · board:durable · trend:down* 📷 Scripts for rendering Objaverse #### [lucidrains/parti-pytorch](https://github.com/lucidrains/parti-pytorch) *Python · ★537 · MIT · — · score:0.00 · hot:0.31 · rising:0.38 · durable:0.51 · board:durable · trend:down* Implementation of Parti, Google's pure attention-based text-to-image neural network, in Pytorch #### [lucidrains/triton-transformer](https://github.com/lucidrains/triton-transformer) *Python · ★279 · MIT · — · score:0.00 · hot:0.31 · rising:0.38 · durable:0.49 · board:durable · trend:down* Implementation of a Transformer, but completely in Triton #### [rasbt/matplotlib-gallery](https://github.com/rasbt/matplotlib-gallery) *Jupyter Notebook · ★1,203 · GPL-3.0 · — · score:0.00 · hot:0.31 · rising:0.39 · durable:0.45 · board:durable · trend:down* Examples of matplotlib codes and plots #### [lucidrains/perfusion-pytorch](https://github.com/lucidrains/perfusion-pytorch) *Python · ★238 · MIT · — · score:0.00 · hot:0.31 · rising:0.37 · durable:0.49 · board:durable · trend:down* Implementation of Key-Locked Rank One Editing, from Nvidia AI #### [allenai/spoc-robot-training](https://github.com/allenai/spoc-robot-training) *Python · ★153 · NOASSERTION · — · score:0.00 · hot:0.31 · rising:0.35 · durable:0.37 · board:durable · trend:down* SPOC: Imitating Shortest Paths in Simulation Enables Effective Navigation and Manipulation in the Real World #### [huggingface/frp](https://github.com/huggingface/frp) *Go · ★180 · Apache-2.0 · — · score:0.00 · hot:0.31 · rising:0.38 · durable:0.44 · board:durable · trend:down* FRP Fork #### [lucidrains/compressive-transformer-pytorch](https://github.com/lucidrains/compressive-transformer-pytorch) *Python · ★164 · MIT · — · score:0.00 · hot:0.31 · rising:0.37 · durable:0.48 · board:durable · trend:down* Pytorch implementation of Compressive Transformers, from Deepmind #### [microsoft/FigmaSharp](https://github.com/microsoft/FigmaSharp) *C# · ★496 · MIT · — · score:0.00 · hot:0.31 · rising:0.36 · durable:0.37 · board:durable · trend:down* Create apps with Figma #### [run-llama/auto_rfp](https://github.com/run-llama/auto_rfp) *TypeScript · ★202 · MIT · — · score:0.00 · hot:0.31 · rising:0.39 · durable:0.46 · board:durable · trend:down* #### [lucidrains/med-seg-diff-pytorch](https://github.com/lucidrains/med-seg-diff-pytorch) *Python · ★238 · MIT · — · score:0.00 · hot:0.31 · rising:0.37 · durable:0.47 · board:durable · trend:down* Implementation of MedSegDiff in Pytorch - SOTA medical segmentation using DDPM and filtering of features in fourier space #### [google-deepmind/3d-shapes](https://github.com/google-deepmind/3d-shapes) *Jupyter Notebook · ★158 · Apache-2.0 · — · score:0.00 · hot:0.31 · rising:0.37 · durable:0.41 · board:durable · trend:down* This repository contains the 3D shapes dataset, used in Kim, Hyunjik and Mnih, Andriy. "Disentangling by Factorising." In Proceedings of the 35th International Conference on Machine Learning (ICML). 2018. to assess the disentanglement properties of unsupervised learning methods. #### [lucidrains/soft-moe-pytorch](https://github.com/lucidrains/soft-moe-pytorch) *Python · ★345 · MIT · — · score:0.00 · hot:0.31 · rising:0.34 · durable:0.50 · board:durable · trend:down* Implementation of Soft MoE, proposed by Brain's Vision team, in Pytorch #### [google-deepmind/magiclens](https://github.com/google-deepmind/magiclens) *Python · ★207 · Apache-2.0 · — · score:0.00 · hot:0.31 · rising:0.37 · durable:0.43 · board:durable · trend:down* [ICML'24 Oral] "MagicLens: Self-Supervised Image Retrieval with Open-Ended Instructions" #### [lucidrains/FLASH-pytorch](https://github.com/lucidrains/FLASH-pytorch) *Python · ★372 · MIT · — · score:0.00 · hot:0.31 · rising:0.37 · durable:0.49 · board:durable · trend:down* Implementation of the Transformer variant proposed in "Transformer Quality in Linear Time" #### [facebookresearch/DistDepth](https://github.com/facebookresearch/DistDepth) *Python · ★236 · NOASSERTION · — · score:0.00 · hot:0.31 · rising:0.34 · durable:0.36 · board:durable · trend:down* Repository for "Toward Practical Monocular Indoor Depth Estimation" (CVPR 2022) #### [allenai/asta-theorizer](https://github.com/allenai/asta-theorizer) *HTML · ★160 · Apache-2.0 · — · score:0.00 · hot:0.31 · rising:0.37 · durable:0.45 · board:durable · trend:down* Staging area for a public release of Theorizer #### [allenai/deepfigures-open](https://github.com/allenai/deepfigures-open) *Python · ★148 · Apache-2.0 · — · score:0.00 · hot:0.31 · rising:0.38 · durable:0.38 · board:durable · trend:down* Companion code to the paper "Extracting Scientific Figures with Distantly Supervised Neural Networks" 🤖 #### [google-research/semivl](https://github.com/google-research/semivl) *Python · ★144 · Apache-2.0 · — · score:0.00 · hot:0.31 · rising:0.38 · durable:0.48 · board:durable · trend:down* [ECCV'24] Official Implementation of SemiVL: Semi-Supervised Semantic Segmentation with Vision-Language Guidance #### [replicate/comfyui-replicate](https://github.com/replicate/comfyui-replicate) *Python · ★207 · MIT · — · score:0.00 · hot:0.30 · rising:0.37 · durable:0.42 · board:durable · trend:down* Run Replicate models as nodes in ComfyUI #### [jina-ai/annlite](https://github.com/jina-ai/annlite) *Python · ★237 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.37 · durable:0.46 · board:durable · trend:down* ⚡ A fast embedded library for approximate nearest neighbor search #### [lucidrains/mlm-pytorch](https://github.com/lucidrains/mlm-pytorch) *Python · ★181 · MIT · — · score:0.00 · hot:0.30 · rising:0.37 · durable:0.48 · board:durable · trend:down* An implementation of masked language modeling for Pytorch, made as concise and simple as possible #### [google-research/visu3d](https://github.com/google-research/visu3d) *Python · ★165 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.38 · durable:0.48 · board:durable · trend:down* 3d without friction (Torch, TF, Jax, Numpy) #### [run-llama/chat-llamaindex](https://github.com/run-llama/chat-llamaindex) *TypeScript · ★984 · MIT · — · score:0.00 · hot:0.30 · rising:0.38 · durable:0.42 · board:durable · trend:down* #### [allenai/pixmo-docs](https://github.com/allenai/pixmo-docs) *Python · ★158 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.37 · durable:0.45 · board:durable · trend:down* ACL 2025: Synthetic data generation pipelines for text-rich images. #### [replicate/pget](https://github.com/replicate/pget) *Go · ★147 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.36 · durable:0.44 · board:durable · trend:down* parallel fetch #### [lucidrains/deformable-attention](https://github.com/lucidrains/deformable-attention) *Python · ★370 · MIT · — · score:0.00 · hot:0.30 · rising:0.38 · durable:0.49 · board:durable · trend:down* Implementation of Deformable Attention in Pytorch from the paper "Vision Transformer with Deformable Attention" #### [vercel/async-sema](https://github.com/vercel/async-sema) *TypeScript · ★654 · MIT · — · score:0.00 · hot:0.30 · rising:0.39 · durable:0.48 · board:durable · trend:down* Semaphore using `async` and `await` #### [simonw/djp](https://github.com/simonw/djp) *Python · ★118 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.34 · durable:0.46 · board:durable · trend:down* A plugin system for Django #### [vercel/on-demand-isr](https://github.com/vercel/on-demand-isr) *TypeScript · ★895 · MIT · — · score:0.00 · hot:0.30 · rising:0.36 · durable:0.40 · board:durable · trend:down* #### [replicate/cog-flux](https://github.com/replicate/cog-flux) *Python · ★369 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.37 · durable:0.41 · board:durable · trend:down* Cog inference for flux models #### [openai/CLIP-featurevis](https://github.com/openai/CLIP-featurevis) *Python · ★310 · no-license · — · score:0.00 · hot:0.30 · rising:0.38 · durable:0.42 · board:durable · trend:down* code for reproducing some of the diagrams in the paper "Multimodal Neurons in Artificial Neural Networks" #### [simonw/s3-credentials](https://github.com/simonw/s3-credentials) *Python · ★252 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.35 · durable:0.42 · board:durable · trend:down* A tool for creating credentials for accessing S3 buckets #### [simonw/justjshtml](https://github.com/simonw/justjshtml) *JavaScript · ★174 · MIT · — · score:0.00 · hot:0.30 · rising:0.35 · durable:0.41 · board:durable · trend:down* JavaScript port of EmilStenstrom/justhtml #### [allenai/ai2thor-rearrangement](https://github.com/allenai/ai2thor-rearrangement) *Python · ★128 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.39 · durable:0.47 · board:durable · trend:down* 🔀 Visual Room Rearrangement #### [lucidrains/h-transformer-1d](https://github.com/lucidrains/h-transformer-1d) *Python · ★166 · MIT · — · score:0.00 · hot:0.30 · rising:0.36 · durable:0.48 · board:durable · trend:down* Implementation of H-Transformer-1D, Hierarchical Attention for Sequence Learning #### [google-deepmind/jaxline](https://github.com/google-deepmind/jaxline) *Python · ★165 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.37 · durable:0.41 · board:durable · trend:down* #### [allenai/cartography](https://github.com/allenai/cartography) *Jupyter Notebook · ★218 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.37 · durable:0.41 · board:durable · trend:down* Dataset Cartography: Mapping and Diagnosing Datasets with Training Dynamics #### [yoheinakajima/GPTvsGPT](https://github.com/yoheinakajima/GPTvsGPT) *Python · ★204 · MIT · — · score:0.00 · hot:0.30 · rising:0.37 · durable:0.43 · board:durable · trend:down* A playful script to get two AI assistants to converse using OpenAI Assistants API #### [lucidrains/STAM-pytorch](https://github.com/lucidrains/STAM-pytorch) *Python · ★134 · MIT · — · score:0.00 · hot:0.30 · rising:0.36 · durable:0.48 · board:durable · trend:down* Implementation of STAM (Space Time Attention Model), a pure and simple attention model that reaches SOTA for video classification #### [lucidrains/recurrent-memory-transformer-pytorch](https://github.com/lucidrains/recurrent-memory-transformer-pytorch) *Python · ★422 · MIT · — · score:0.00 · hot:0.30 · rising:0.36 · durable:0.49 · board:durable · trend:down* Implementation of Recurrent Memory Transformer, Neurips 2022 paper, in Pytorch #### [facebookresearch/cowtracker](https://github.com/facebookresearch/cowtracker) *Python · ★141 · NOASSERTION · — · score:0.00 · hot:0.30 · rising:0.33 · durable:0.40 · board:durable · trend:down* CoWTracker: Tracking by Warping instead of Correlation #### [simonw/python-lib](https://github.com/simonw/python-lib) *Python · ★226 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.35 · durable:0.41 · board:durable · trend:down* Opinionated cookiecutter template for creating a new Python library #### [google-deepmind/dm_memorytasks](https://github.com/google-deepmind/dm_memorytasks) *Python · ★225 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.39 · durable:0.50 · board:durable · trend:down* A set of 13 diverse machine-learning tasks that require memory to solve. #### [allenai/lm-explorer](https://github.com/allenai/lm-explorer) *Python · ★134 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.36 · durable:0.39 · board:durable · trend:down* interactive explorer for language models #### [vercel/bidc](https://github.com/vercel/bidc) *TypeScript · ★1,261 · no-license · — · score:0.00 · hot:0.30 · rising:0.33 · durable:0.52 · board:durable · trend:down* Bidirectional Channels for JavaScript #### [simonw/ttok](https://github.com/simonw/ttok) *Python · ★387 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.35 · durable:0.45 · board:durable · trend:down* Count and truncate text based on tokens #### [swyxio/svelte-actions](https://github.com/swyxio/svelte-actions) *TypeScript · ★222 · MIT · — · score:0.00 · hot:0.30 · rising:0.35 · durable:0.44 · board:durable · trend:down* prototype official actions for Svelte #### [dair-ai/d2l-study-group](https://github.com/dair-ai/d2l-study-group) *? · ★395 · MIT · — · score:0.00 · hot:0.30 · rising:0.38 · durable:0.44 · board:durable · trend:down* 🧠 Material for the Deep Learning Study Group #### [openai/imitation](https://github.com/openai/imitation) *Python · ★731 · MIT · — · score:0.00 · hot:0.30 · rising:0.39 · durable:0.43 · board:durable · trend:down* Code for the paper "Generative Adversarial Imitation Learning" #### [lucidrains/local-attention](https://github.com/lucidrains/local-attention) *Python · ★498 · MIT · — · score:0.00 · hot:0.30 · rising:0.38 · durable:0.49 · board:durable · trend:down* An implementation of local windowed attention for language modeling #### [simonw/llm-mlx](https://github.com/simonw/llm-mlx) *Python · ★242 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.37 · durable:0.46 · board:durable · trend:down* Support for MLX models in LLM #### [dair-ai/awesome-research-proposals-guide](https://github.com/dair-ai/awesome-research-proposals-guide) *? · ★205 · no-license · — · score:0.00 · hot:0.30 · rising:0.34 · durable:0.39 · board:durable · trend:down* A guide to improve your research proposals. #### [EleutherAI/lm_perplexity](https://github.com/EleutherAI/lm_perplexity) *Python · ★164 · MIT · — · score:0.00 · hot:0.30 · rising:0.36 · durable:0.43 · board:durable · trend:down* #### [google-research/meliad](https://github.com/google-research/meliad) *Python · ★260 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.37 · durable:0.45 · board:durable · trend:down* #### [replicate/hype](https://github.com/replicate/hype) *TypeScript · ★235 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.37 · durable:0.44 · board:durable · trend:down* A feed of trending repos/models from GitHub, Replicate, HuggingFace, and Reddit. #### [allenai/acl2022-zerofewshot-tutorial](https://github.com/allenai/acl2022-zerofewshot-tutorial) *? · ★291 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.35 · durable:0.44 · board:durable · trend:down* #### [allenai/SPECTER2](https://github.com/allenai/SPECTER2) *Python · ★129 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.35 · durable:0.40 · board:durable · trend:down* #### [huggingface/swift-chat](https://github.com/huggingface/swift-chat) *Swift · ★598 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.37 · durable:0.42 · board:durable · trend:down* Mac app to demonstrate swift-transformers #### [google-deepmind/PGMax](https://github.com/google-deepmind/PGMax) *Jupyter Notebook · ★165 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.36 · durable:0.44 · board:durable · trend:down* Loopy belief propagation for factor graphs on discrete variables in JAX #### [simonw/db-to-sqlite](https://github.com/simonw/db-to-sqlite) *Python · ★490 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.36 · durable:0.43 · board:durable · trend:down* CLI tool for exporting tables or queries from any SQL database to a SQLite file #### [google-research/url-nlp](https://github.com/google-research/url-nlp) *Python · ★272 · no-license · — · score:0.00 · hot:0.30 · rising:0.34 · durable:0.35 · board:durable · trend:down* #### [EleutherAI/concept-erasure](https://github.com/EleutherAI/concept-erasure) *Python · ★246 · MIT · — · score:0.00 · hot:0.30 · rising:0.38 · durable:0.50 · board:durable · trend:down* Erasing concepts from neural representations with provable guarantees #### [EleutherAI/oslo](https://github.com/EleutherAI/oslo) *Python · ★174 · no-license · — · score:0.00 · hot:0.30 · rising:0.34 · durable:0.36 · board:durable · trend:down* OSLO: Open Source for Large-scale Optimization #### [google-deepmind/gemini_icpc2025](https://github.com/google-deepmind/gemini_icpc2025) *C++ · ★171 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.36 · durable:0.45 · board:durable · trend:down* Gemini 2025 ICPC World Finals Code Submissions #### [vercel/micro-dev](https://github.com/vercel/micro-dev) *JavaScript · ★710 · MIT · — · score:0.00 · hot:0.30 · rising:0.39 · durable:0.45 · board:durable · trend:down* The development environment for `micro` #### [run-llama/multi-agent-concierge](https://github.com/run-llama/multi-agent-concierge) *Jupyter Notebook · ★445 · MIT · — · score:0.00 · hot:0.30 · rising:0.38 · durable:0.47 · board:durable · trend:down* An example of multi-agent orchestration with llama-index #### [openai/dallify-discord-bot](https://github.com/openai/dallify-discord-bot) *TypeScript · ★151 · MIT · — · score:0.00 · hot:0.30 · rising:0.37 · durable:0.43 · board:durable · trend:down* Example code for using OpenAI’s NodeJS SDK with discord.js SDK to create a Discord Bot that uses Slash Commands. #### [google-research/corenet](https://github.com/google-research/corenet) *Python · ★120 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.36 · durable:0.40 · board:durable · trend:down* CoReNet is a technique for joint multi-object 3D reconstruction from a single RGB image. #### [lucidrains/bidirectional-cross-attention](https://github.com/lucidrains/bidirectional-cross-attention) *Python · ★195 · MIT · — · score:0.00 · hot:0.30 · rising:0.37 · durable:0.49 · board:durable · trend:down* A simple cross attention that updates both the source and target in one step #### [vercel/remote-cache](https://github.com/vercel/remote-cache) *TypeScript · ★191 · MPL-2.0 · — · score:0.00 · hot:0.30 · rising:0.38 · durable:0.50 · board:durable · trend:down* The Vercel Remote Cache SDK #### [simonw/ospeak](https://github.com/simonw/ospeak) *Python · ★171 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.36 · durable:0.46 · board:durable · trend:down* CLI tool for running text through OpenAI Text to speech #### [rasbt/pyprind](https://github.com/rasbt/pyprind) *Python · ★545 · BSD-3-Clause · — · score:0.00 · hot:0.30 · rising:0.37 · durable:0.46 · board:durable · trend:down* PyPrind - Python Progress Indicator Utility #### [allenai/comet-atomic-2020](https://github.com/allenai/comet-atomic-2020) *Python · ★261 · no-license · — · score:0.00 · hot:0.30 · rising:0.35 · durable:0.40 · board:durable · trend:down* #### [simonw/sqlite-diffable](https://github.com/simonw/sqlite-diffable) *Python · ★137 · Apache-2.0 · — · score:0.00 · hot:0.30 · rising:0.34 · durable:0.44 · board:durable · trend:down* Tools for dumping/loading a SQLite database to diffable directory structure #### [huggingface/naacl_transfer_learning_tutorial](https://github.com/huggingface/naacl_transfer_learning_tutorial) *Python · ★722 · MIT · — · score:0.00 · hot:0.30 · rising:0.38 · durable:0.45 · board:durable · trend:down* Repository of code for the tutorial on Transfer Learning in NLP held at NAACL 2019 in Minneapolis, MN, USA #### [google-research/jaxpruner](https://github.com/google-research/jaxpruner) *Python · ★235 · Apache-2.0 · — · score:0.00 · hot:0.29 · rising:0.37 · durable:0.44 · board:durable · trend:down* #### [google-deepmind/jmp](https://github.com/google-deepmind/jmp) *Python · ★212 · Apache-2.0 · — · score:0.00 · hot:0.29 · rising:0.38 · durable:0.49 · board:durable · trend:down* JMP is a Mixed Precision library for JAX. #### [lucidrains/contrastive-learner](https://github.com/lucidrains/contrastive-learner) *Python · ★151 · MIT · — · score:0.00 · hot:0.29 · rising:0.37 · durable:0.47 · board:durable · trend:down* A simple to use pytorch wrapper for contrastive self-supervised learning on any neural network #### [rasbt/pycon2024](https://github.com/rasbt/pycon2024) *Jupyter Notebook · ★247 · MIT · — · score:0.00 · hot:0.29 · rising:0.37 · durable:0.45 · board:durable · trend:down* Tutorial Materials for "The Fundamentals of Modern Deep Learning with PyTorch" workshop at PyCon 2024 #### [karpathy/svmjs](https://github.com/karpathy/svmjs) *JavaScript · ★706 · MIT · — · score:0.00 · hot:0.29 · rising:0.37 · durable:0.42 · board:durable · trend:down* Support Vector Machine in Javascript (SMO algorithm, supports arbitrary kernels) + GUI demo #### [google-research/nested-transformer](https://github.com/google-research/nested-transformer) *Jupyter Notebook · ★203 · Apache-2.0 · — · score:0.00 · hot:0.29 · rising:0.35 · durable:0.42 · board:durable · trend:down* Nested Hierarchical Transformer https://arxiv.org/pdf/2105.12723.pdf #### [openai/openai-deno-build](https://github.com/openai/openai-deno-build) *TypeScript · ★145 · Apache-2.0 · — · score:0.00 · hot:0.29 · rising:0.36 · durable:0.46 · board:durable · trend:down* Deno build of the official Typescript library for the OpenAI API. #### [openai/lean-gym](https://github.com/openai/lean-gym) *Lean · ★202 · Apache-2.0 · — · score:0.00 · hot:0.29 · rising:0.36 · durable:0.41 · board:durable · trend:down* #### [rasbt/One-Python-benchmark-per-day](https://github.com/rasbt/One-Python-benchmark-per-day) *HTML · ★134 · GPL-3.0 · — · score:0.00 · hot:0.29 · rising:0.36 · durable:0.40 · board:durable · trend:down* An ongoing fun challenge where I'll try to post one Python benchmark per day. #### [run-llama/modal_finetune_sql](https://github.com/run-llama/modal_finetune_sql) *Jupyter Notebook · ★323 · MIT · — · score:0.00 · hot:0.29 · rising:0.36 · durable:0.44 · board:durable · trend:down* #### [google-deepmind/neural_networks_chomsky_hierarchy](https://github.com/google-deepmind/neural_networks_chomsky_hierarchy) *Python · ★214 · Apache-2.0 · — · score:0.00 · hot:0.29 · rising:0.37 · durable:0.45 · board:durable · trend:down* Neural Networks and the Chomsky Hierarchy #### [run-llama/finetune-embedding](https://github.com/run-llama/finetune-embedding) *Jupyter Notebook · ★526 · no-license · — · score:0.00 · hot:0.29 · rising:0.35 · durable:0.43 · board:durable · trend:down* Fine-Tuning Embedding for RAG with Synthetic Data #### [vercel/preview-mode-demo](https://github.com/vercel/preview-mode-demo) *TypeScript · ★124 · MIT · — · score:0.00 · hot:0.29 · rising:0.38 · durable:0.43 · board:durable · trend:down* This demo showcases Next.js' next-gen Static Site Generation (SSG) support. #### [simonw/csv-diff](https://github.com/simonw/csv-diff) *Python · ★330 · Apache-2.0 · — · score:0.00 · hot:0.29 · rising:0.36 · durable:0.42 · board:durable · trend:down* Python CLI tool and library for diffing CSV and JSON files #### [allenai/bilm-tf](https://github.com/allenai/bilm-tf) *Python · ★1,613 · Apache-2.0 · — · score:0.00 · hot:0.29 · rising:0.38 · durable:0.42 · board:durable · trend:down* Tensorflow implementation of contextualized word representations from bi-directional language models #### [google-research/neural-structural-optimization](https://github.com/google-research/neural-structural-optimization) *Jupyter Notebook · ★130 · Apache-2.0 · — · score:0.00 · hot:0.29 · rising:0.36 · durable:0.41 · board:durable · trend:down* Neural reparameterization improves structural optimization #### [replicate/flux-fine-tuner](https://github.com/replicate/flux-fine-tuner) *Python · ★429 · Apache-2.0 · — · score:0.00 · hot:0.29 · rising:0.36 · durable:0.42 · board:durable · trend:down* Cog wrapper for ostris/ai-toolkit + post-finetuning cog inference for flux models #### [vercel/email-prompt](https://github.com/vercel/email-prompt) *JavaScript · ★277 · MIT · — · score:0.00 · hot:0.29 · rising:0.37 · durable:0.48 · board:durable · trend:down* CLI email prompt with autocompletion and built-in validation #### [jina-ai/fastapi-serve](https://github.com/jina-ai/fastapi-serve) *Python · ★145 · Apache-2.0 · — · score:0.00 · hot:0.29 · rising:0.37 · durable:0.51 · board:durable · trend:down* FastAPI to the Cloud, Batteries Included! ☁️🔋🚀 #### [rasbt/data-science-tutorial](https://github.com/rasbt/data-science-tutorial) *Jupyter Notebook · ★197 · MIT · — · score:0.00 · hot:0.29 · rising:0.37 · durable:0.41 · board:durable · trend:down* Code material for a data science tutorial #### [rasbt/protein-science](https://github.com/rasbt/protein-science) *Python · ★109 · GPL-3.0 · — · score:0.00 · hot:0.29 · rising:0.35 · durable:0.39 · board:durable · trend:down* A collection of useful tutorials for Protein Science #### [lm-sys/llm-decontaminator](https://github.com/lm-sys/llm-decontaminator) *Python · ★319 · Apache-2.0 · — · score:0.00 · hot:0.29 · rising:0.35 · durable:0.42 · board:durable · trend:down* Code for the paper "Rethinking Benchmark and Contamination for Language Models with Rephrased Samples" #### [huggingface/torchMoji](https://github.com/huggingface/torchMoji) *Python · ★921 · MIT · — · score:0.00 · hot:0.29 · rising:0.36 · durable:0.42 · board:durable · trend:down* 😇A pyTorch implementation of the DeepMoji model: state-of-the-art deep learning model for analyzing sentiment, emotion, sarcasm etc #### [facebookresearch/pixio](https://github.com/facebookresearch/pixio) *Python · ★368 · NOASSERTION · — · score:0.00 · hot:0.29 · rising:0.31 · durable:0.43 · board:durable · trend:down* Pixio: a capable vision encoder dedicated to dense prediction, simply by pixel reconstruction #### [dair-ai/emotion_dataset](https://github.com/dair-ai/emotion_dataset) *? · ★222 · no-license · — · score:0.00 · hot:0.29 · rising:0.33 · durable:0.40 · board:durable · trend:down* :smile: Dataset for Emotion Recognition Research #### [lucidrains/improving-transformers-world-model-for-rl](https://github.com/lucidrains/improving-transformers-world-model-for-rl) *Python · ★153 · MIT · — · score:0.00 · hot:0.29 · rising:0.34 · durable:0.50 · board:durable · trend:down* Implementation of the new SOTA for model based RL, from the paper "Improving Transformer World Models for Data-Efficient RL", in Pytorch #### [jina-ai/jcloud](https://github.com/jina-ai/jcloud) *Python · ★298 · Apache-2.0 · — · score:0.00 · hot:0.29 · rising:0.37 · durable:0.51 · board:durable · trend:down* Simplify deploying and managing Jina projects on Jina Cloud #### [vercel/test-listen](https://github.com/vercel/test-listen) *JavaScript · ★157 · MIT · — · score:0.00 · hot:0.29 · rising:0.37 · durable:0.44 · board:durable · trend:down* Quick ephemeral URLs for your tests #### [QwenLM/ProcessBench](https://github.com/QwenLM/ProcessBench) *Python · ★189 · no-license · — · score:0.00 · hot:0.29 · rising:0.33 · durable:0.40 · board:durable · trend:down* Official repository for ACL 2025 paper "ProcessBench: Identifying Process Errors in Mathematical Reasoning" #### [simonw/optfunc](https://github.com/simonw/optfunc) *Python · ★135 · BSD-2-Clause · — · score:0.00 · hot:0.29 · rising:0.34 · durable:0.38 · board:durable · trend:down* Syntactic sugar for creating Python command line scripts by introspecting a function definition #### [viiika/Meissonic](https://github.com/viiika/Meissonic) *Python · ★344 · Apache-2.0 · — · score:0.00 · hot:0.29 · rising:0.32 · durable:0.44 · board:durable · trend:down* [ICLR 2025] Official Implementation of Meissonic: Revitalizing Masked Generative Transformers for Efficient High-Resolution Text-to-Image Synthesis #### [google-deepmind/mathematics_conjectures](https://github.com/google-deepmind/mathematics_conjectures) *Jupyter Notebook · ★383 · Apache-2.0 · — · score:0.00 · hot:0.29 · rising:0.37 · durable:0.46 · board:durable · trend:down* #### [run-llama/llamacloud-mcp](https://github.com/run-llama/llamacloud-mcp) *Python · ★224 · MIT · — · score:0.00 · hot:0.29 · rising:0.37 · durable:0.46 · board:durable · trend:down* #### [rasbt/pydata-chicago2016-ml-tutorial](https://github.com/rasbt/pydata-chicago2016-ml-tutorial) *Jupyter Notebook · ★128 · MIT · — · score:0.00 · hot:0.29 · rising:0.36 · durable:0.41 · board:durable · trend:down* Machine learning with scikit-learn tutorial at PyData Chicago 2016 #### [google-deepmind/representations4d](https://github.com/google-deepmind/representations4d) *Jupyter Notebook · ★125 · Apache-2.0 · — · score:0.00 · hot:0.29 · rising:0.33 · durable:0.43 · board:durable · trend:down* #### [simonw/prompts-js](https://github.com/simonw/prompts-js) *JavaScript · ★154 · Apache-2.0 · — · score:0.00 · hot:0.29 · rising:0.33 · durable:0.48 · board:durable · trend:down* async alternatives to browser alert() and prompt() and confirm() #### [vercel/update-check](https://github.com/vercel/update-check) *JavaScript · ★168 · MIT · — · score:0.00 · hot:0.29 · rising:0.37 · durable:0.47 · board:durable · trend:down* Minimalistic update notifications for command line interfaces #### [allenai/vila](https://github.com/allenai/vila) *Python · ★180 · Apache-2.0 · — · score:0.00 · hot:0.29 · rising:0.36 · durable:0.47 · board:durable · trend:down* Incorporating VIsual LAyout Structures for Scientific Text Classification #### [huggingface/diffusion-fast](https://github.com/huggingface/diffusion-fast) *Python · ★232 · Apache-2.0 · — · score:0.00 · hot:0.29 · rising:0.35 · durable:0.46 · board:durable · trend:down* Faster generation with text-to-image diffusion models. #### [allenai/mmda](https://github.com/allenai/mmda) *Jupyter Notebook · ★165 · Apache-2.0 · — · score:0.00 · hot:0.29 · rising:0.37 · durable:0.44 · board:durable · trend:down* multimodal document analysis #### [allenai/procthor-10k](https://github.com/allenai/procthor-10k) *Python · ★121 · Apache-2.0 · — · score:0.00 · hot:0.29 · rising:0.35 · durable:0.45 · board:durable · trend:down* The ProcTHOR-10K Houses Dataset #### [vercel/zsh-theme](https://github.com/vercel/zsh-theme) *? · ★194 · MIT · — · score:0.00 · hot:0.29 · rising:0.37 · durable:0.42 · board:durable · trend:down* Yet another zsh theme #### [vercel/uid-promise](https://github.com/vercel/uid-promise) *TypeScript · ★257 · MIT · — · score:0.00 · hot:0.29 · rising:0.36 · durable:0.48 · board:durable · trend:down* Creates a cryptographically strong UID #### [google-research/rigl](https://github.com/google-research/rigl) *Python · ★335 · Apache-2.0 · — · score:0.00 · hot:0.29 · rising:0.37 · durable:0.45 · board:durable · trend:down* End-to-end training of sparse deep neural networks with little-to-no performance loss. #### [karpathy/Random-Forest-Matlab](https://github.com/karpathy/Random-Forest-Matlab) *Matlab · ★223 · no-license · — · score:0.00 · hot:0.28 · rising:0.35 · durable:0.38 · board:durable · trend:down* A Random Forest implementation for MATLAB. Supports arbitrary weak learners that you can define. #### [allenai/discoverybench](https://github.com/allenai/discoverybench) *Python · ★137 · NOASSERTION · — · score:0.00 · hot:0.28 · rising:0.34 · durable:0.39 · board:durable · trend:down* Discovering Data-driven Hypotheses in the Wild #### [simonw/django-queryset-transform](https://github.com/simonw/django-queryset-transform) *Python · ★145 · BSD-3-Clause · — · score:0.00 · hot:0.28 · rising:0.33 · durable:0.38 · board:durable · trend:down* Experimental .transform(fn) method for Django QuerySets, for clever lazily evaluated optimisations. #### [rasbt/faster-pytorch-blog](https://github.com/rasbt/faster-pytorch-blog) *Python · ★128 · no-license · — · score:0.00 · hot:0.28 · rising:0.32 · durable:0.40 · board:durable · trend:down* Outlining techniques for improving the training performance of your PyTorch model without compromising its accuracy #### [dair-ai/nlp_newsletter](https://github.com/dair-ai/nlp_newsletter) *? · ★303 · no-license · — · score:0.00 · hot:0.28 · rising:0.35 · durable:0.39 · board:durable · trend:down* 📰Natural language processing (NLP) newsletter #### [rasbt/datacollect](https://github.com/rasbt/datacollect) *Jupyter Notebook · ★209 · GPL-3.0 · — · score:0.00 · hot:0.28 · rising:0.36 · durable:0.39 · board:durable · trend:down* A collection of tools to collect and download various data. #### [rasbt/musicmood](https://github.com/rasbt/musicmood) *OpenEdge ABL · ★422 · GPL-3.0 · — · score:0.00 · hot:0.28 · rising:0.36 · durable:0.41 · board:durable · trend:down* A machine learning approach to classify songs by mood. #### [dair-ai/nlp_fundamentals](https://github.com/dair-ai/nlp_fundamentals) *Jupyter Notebook · ★372 · MIT · — · score:0.00 · hot:0.28 · rising:0.36 · durable:0.41 · board:durable · trend:down* 📘 Contains a series of hands-on notebooks for learning the fundamentals of NLP #### [vercel/hyperpower](https://github.com/vercel/hyperpower) *JavaScript · ★635 · MIT · — · score:0.00 · hot:0.28 · rising:0.36 · durable:0.44 · board:durable · trend:down* Hyper particle effects extension #### [vercel/mongodb-starter](https://github.com/vercel/mongodb-starter) *TypeScript · ★510 · no-license · — · score:0.00 · hot:0.28 · rising:0.35 · durable:0.40 · board:durable · trend:down* A developer directory built on Next.js and MongoDB Atlas, deployed on Vercel with the Vercel + MongoDB integration. #### [jina-ai/deepsearch-ui](https://github.com/jina-ai/deepsearch-ui) *JavaScript · ★128 · Apache-2.0 · — · score:0.00 · hot:0.28 · rising:0.36 · durable:0.45 · board:durable · trend:down* Jina DeepSearch UI #### [allenai/PeerRead](https://github.com/allenai/PeerRead) *Python · ★427 · no-license · — · score:0.00 · hot:0.28 · rising:0.35 · durable:0.39 · board:durable · trend:down* Data and code for Kang et al., NAACL 2018's paper titled "A Dataset of Peer Reviews (PeerRead): Collection, Insights and NLP Applications" #### [lucidrains/Adan-pytorch](https://github.com/lucidrains/Adan-pytorch) *Python · ★253 · MIT · — · score:0.00 · hot:0.28 · rising:0.35 · durable:0.50 · board:durable · trend:down* Implementation of the Adan (ADAptive Nesterov momentum algorithm) Optimizer in Pytorch #### [simonw/llm-jq](https://github.com/simonw/llm-jq) *Python · ★192 · Apache-2.0 · — · score:0.00 · hot:0.28 · rising:0.32 · durable:0.49 · board:durable · trend:down* Write and execute jq programs with the help of LLM #### [allenai/pdffigures](https://github.com/allenai/pdffigures) *C++ · ★130 · GPL-2.0 · — · score:0.00 · hot:0.28 · rising:0.35 · durable:0.38 · board:durable · trend:down* Command line tool to extract figures, tables, and captions from scholarly documents in PDF form. #### [run-llama/gmail-extractor](https://github.com/run-llama/gmail-extractor) *Python · ★124 · MIT · — · score:0.00 · hot:0.28 · rising:0.36 · durable:0.45 · board:durable · trend:down* #### [vercel/git-hooks](https://github.com/vercel/git-hooks) *JavaScript · ★202 · MIT · — · score:0.00 · hot:0.28 · rising:0.36 · durable:0.48 · board:durable · trend:down* No nonsense Git hook management #### [jxnl/n-levels-of-rag](https://github.com/jxnl/n-levels-of-rag) *Python · ★197 · MIT · — · score:0.00 · hot:0.28 · rising:0.34 · durable:0.42 · board:durable · trend:down* #### [google-deepmind/synjax](https://github.com/google-deepmind/synjax) *Python · ★250 · Apache-2.0 · — · score:0.00 · hot:0.28 · rising:0.36 · durable:0.45 · board:durable · trend:down* #### [allenai/macaw](https://github.com/allenai/macaw) *Python · ★458 · Apache-2.0 · — · score:0.00 · hot:0.28 · rising:0.36 · durable:0.43 · board:durable · trend:down* Multi-angle c(q)uestion answering #### [lucidrains/mixture-of-attention](https://github.com/lucidrains/mixture-of-attention) *Python · ★122 · MIT · — · score:0.00 · hot:0.28 · rising:0.33 · durable:0.48 · board:durable · trend:down* Some personal experiments around routing tokens to different autoregressive attention, akin to mixture-of-experts #### [huggingface/100-times-faster-nlp](https://github.com/huggingface/100-times-faster-nlp) *HTML · ★336 · no-license · — · score:0.00 · hot:0.28 · rising:0.35 · durable:0.40 · board:durable · trend:down* 🚀100 Times Faster Natural Language Processing in Python - iPython notebook #### [google-deepmind/grid-cells](https://github.com/google-deepmind/grid-cells) *Python · ★263 · Apache-2.0 · — · score:0.00 · hot:0.28 · rising:0.36 · durable:0.42 · board:durable · trend:down* Implementation of the supervised learning experiments in Vector-based navigation using grid-like representations in artificial agents, as published at https://www.nature.com/articles/s41586-018-0102-6 #### [pydantic/bump-pydantic](https://github.com/pydantic/bump-pydantic) *Python · ★348 · MIT · — · score:0.00 · hot:0.28 · rising:0.36 · durable:0.46 · board:durable · trend:down* Convert Pydantic from V1 to V2 ♻ #### [rasbt/stat451-machine-learning-fs21](https://github.com/rasbt/stat451-machine-learning-fs21) *Jupyter Notebook · ★150 · MIT · — · score:0.00 · hot:0.28 · rising:0.35 · durable:0.42 · board:durable · trend:down* #### [karpathy/forestjs](https://github.com/karpathy/forestjs) *JavaScript · ★335 · MIT · — · score:0.00 · hot:0.28 · rising:0.35 · durable:0.41 · board:durable · trend:down* Random Forest implementation for JavaScript. Supports arbitrary weak learners. Includes interactive demo. #### [google-research/readout_guidance](https://github.com/google-research/readout_guidance) *Jupyter Notebook · ★153 · Apache-2.0 · — · score:0.00 · hot:0.28 · rising:0.34 · durable:0.43 · board:durable · trend:down* Official PyTorch Implementation for Readout Guidance, CVPR 2024 #### [huggingface/community-events](https://github.com/huggingface/community-events) *Jupyter Notebook · ★429 · no-license · — · score:0.00 · hot:0.28 · rising:0.35 · durable:0.38 · board:durable · trend:down* Place where folks can contribute to 🤗 community events #### [run-llama/mongodb-demo](https://github.com/run-llama/mongodb-demo) *Python · ★181 · no-license · — · score:0.00 · hot:0.28 · rising:0.34 · durable:0.42 · board:durable · trend:down* #### [yoheinakajima/pippin](https://github.com/yoheinakajima/pippin) *Python · ★106 · no-license · — · score:0.00 · hot:0.28 · rising:0.33 · durable:0.40 · board:durable · trend:down* a little AI-generated unicorn #### [allenai/FineGrainedRLHF](https://github.com/allenai/FineGrainedRLHF) *Python · ★284 · Apache-2.0 · — · score:0.00 · hot:0.28 · rising:0.35 · durable:0.45 · board:durable · trend:down* #### [run-llama/llamabot](https://github.com/run-llama/llamabot) *Python · ★171 · MIT · — · score:0.00 · hot:0.28 · rising:0.35 · durable:0.44 · board:durable · trend:down* #### [dair-ai/ml-nlp-paper-discussions](https://github.com/dair-ai/ml-nlp-paper-discussions) *? · ★149 · no-license · — · score:0.00 · hot:0.28 · rising:0.34 · durable:0.39 · board:durable · trend:down* 📄 A repo containing notes and discussions for our weekly NLP/ML paper discussions. #### [google-deepmind/language_to_reward_2023](https://github.com/google-deepmind/language_to_reward_2023) *Python · ★160 · Apache-2.0 · — · score:0.00 · hot:0.28 · rising:0.35 · durable:0.43 · board:durable · trend:down* #### [allenai/naacl2021-longdoc-tutorial](https://github.com/allenai/naacl2021-longdoc-tutorial) *Python · ★344 · Apache-2.0 · — · score:0.00 · hot:0.28 · rising:0.35 · durable:0.44 · board:durable · trend:down* #### [allenai/unifiedqa](https://github.com/allenai/unifiedqa) *Python · ★445 · Apache-2.0 · — · score:0.00 · hot:0.28 · rising:0.36 · durable:0.44 · board:durable · trend:down* UnifiedQA: Crossing Format Boundaries With a Single QA System #### [facebookresearch/4DGT](https://github.com/facebookresearch/4DGT) *Python · ★433 · NOASSERTION · — · score:0.00 · hot:0.28 · rising:0.30 · durable:0.42 · board:durable · trend:down* [NeurIPS 2025 (Spotlight)] The implementation for the paper "4DGT Learning a 4D Gaussian Transformer Using Real-World Monocular Videos" #### [run-llama/openai_realtime_client](https://github.com/run-llama/openai_realtime_client) *Python · ★244 · MIT · — · score:0.00 · hot:0.28 · rising:0.35 · durable:0.43 · board:durable · trend:down* A simple client and utils for interacting with OpenAI's Realtime API in Python #### [google-deepmind/interval-bound-propagation](https://github.com/google-deepmind/interval-bound-propagation) *Python · ★161 · Apache-2.0 · — · score:0.00 · hot:0.28 · rising:0.35 · durable:0.41 · board:durable · trend:down* This repository contains a simple implementation of Interval Bound Propagation (IBP) using TensorFlow: https://arxiv.org/abs/1810.12715 #### [rasbt/msu-datascience-ml-tutorial-2018](https://github.com/rasbt/msu-datascience-ml-tutorial-2018) *Jupyter Notebook · ★111 · MIT · — · score:0.00 · hot:0.28 · rising:0.35 · durable:0.40 · board:durable · trend:down* Machine learning with Python tutorial at MSU Data Science 2018 #### [run-llama/llama-slides](https://github.com/run-llama/llama-slides) *TypeScript · ★129 · MIT · — · score:0.00 · hot:0.28 · rising:0.35 · durable:0.45 · board:durable · trend:down* #### [simonw/llm-cmd](https://github.com/simonw/llm-cmd) *Python · ★468 · Apache-2.0 · — · score:0.00 · hot:0.28 · rising:0.32 · durable:0.38 · board:durable · trend:down* Use LLM to generate and execute commands in your shell #### [run-llama/flow-maker](https://github.com/run-llama/flow-maker) *TypeScript · ★212 · MIT · — · score:0.00 · hot:0.28 · rising:0.35 · durable:0.45 · board:durable · trend:down* #### [vercel/little-date](https://github.com/vercel/little-date) *TypeScript · ★1,966 · MIT · — · score:0.00 · hot:0.28 · rising:0.31 · durable:0.54 · board:durable · trend:down* A friendly formatter to make date ranges small & sweet #### [replicate/tilemaker](https://github.com/replicate/tilemaker) *JavaScript · ★151 · no-license · — · score:0.00 · hot:0.27 · rising:0.34 · durable:0.39 · board:durable · trend:down* #### [huggingface/HuggingSnap](https://github.com/huggingface/HuggingSnap) *Swift · ★188 · no-license · — · score:0.00 · hot:0.27 · rising:0.34 · durable:0.42 · board:durable · trend:down* SmolVLM2 Demo #### [simonw/airtable-export](https://github.com/simonw/airtable-export) *Python · ★133 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.34 · durable:0.43 · board:durable · trend:down* Export Airtable data to YAML, JSON or SQLite files on disk #### [run-llama/ai-engineer-workshop](https://github.com/run-llama/ai-engineer-workshop) *Jupyter Notebook · ★187 · no-license · — · score:0.00 · hot:0.27 · rising:0.34 · durable:0.42 · board:durable · trend:down* #### [replicate/cog-sdxl](https://github.com/replicate/cog-sdxl) *Python · ★231 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.35 · durable:0.40 · board:durable · trend:down* Stable Diffusion XL training and inference as a cog model #### [google-deepmind/spiral](https://github.com/google-deepmind/spiral) *C++ · ★332 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.35 · durable:0.42 · board:durable · trend:down* We provide a pre-trained model for unconditional 19-step generation of CelebA-HQ images #### [simonw/datasette-lite](https://github.com/simonw/datasette-lite) *CSS · ★399 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.33 · durable:0.37 · board:durable · trend:down* Datasette running in your browser using WebAssembly and Pyodide #### [allenai/cord19](https://github.com/allenai/cord19) *? · ★185 · NOASSERTION · — · score:0.00 · hot:0.27 · rising:0.34 · durable:0.40 · board:durable · trend:down* Get started with CORD-19 #### [google-deepmind/mishax](https://github.com/google-deepmind/mishax) *Python · ★156 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.35 · durable:0.44 · board:durable · trend:down* #### [simonw/openai-to-sqlite](https://github.com/simonw/openai-to-sqlite) *Python · ★236 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.34 · durable:0.48 · board:durable · trend:down* Save OpenAI API results to a SQLite database #### [EleutherAI/knowledge-neurons](https://github.com/EleutherAI/knowledge-neurons) *Python · ★159 · MIT · — · score:0.00 · hot:0.27 · rising:0.33 · durable:0.43 · board:durable · trend:down* A library for finding knowledge neurons in pretrained transformer models. #### [google-deepmind/neural_testbed](https://github.com/google-deepmind/neural_testbed) *Jupyter Notebook · ★191 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.35 · durable:0.43 · board:durable · trend:down* #### [huggingface/pytorch_block_sparse](https://github.com/huggingface/pytorch_block_sparse) *C++ · ★550 · NOASSERTION · — · score:0.00 · hot:0.27 · rising:0.34 · durable:0.38 · board:durable · trend:down* Fast Block Sparse Matrices for Pytorch #### [simonw/symbex](https://github.com/simonw/symbex) *Python · ★315 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.33 · durable:0.49 · board:durable · trend:down* Find the Python code for specified symbols #### [google-deepmind/multidim-image-augmentation](https://github.com/google-deepmind/multidim-image-augmentation) *C++ · ★139 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.35 · durable:0.41 · board:durable · trend:down* This package provides TensorFlow Ops for multidimensional volumetric image augmentation. #### [huggingface/chug](https://github.com/huggingface/chug) *Python · ★161 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.35 · durable:0.44 · board:durable · trend:down* Minimal sharded dataset loaders, decoders, and utils for multi-modal document, image, and text datasets. #### [simonw/git-scraper-template](https://github.com/simonw/git-scraper-template) *Shell · ★127 · no-license · — · score:0.00 · hot:0.27 · rising:0.31 · durable:0.39 · board:durable · trend:down* Template repository for setting up a new git scraper #### [allenai/spv2](https://github.com/allenai/spv2) *Python · ★255 · no-license · — · score:0.00 · hot:0.27 · rising:0.34 · durable:0.38 · board:durable · trend:down* Science-parse version 2 #### [lucidrains/MaMMUT-pytorch](https://github.com/lucidrains/MaMMUT-pytorch) *Python · ★104 · MIT · — · score:0.00 · hot:0.27 · rising:0.32 · durable:0.48 · board:durable · trend:down* Implementation of MaMMUT, a simple vision-encoder text-decoder architecture for multimodal tasks from Google, in Pytorch #### [pydantic/pydantic.run](https://github.com/pydantic/pydantic.run) *TypeScript · ★187 · MIT · — · score:0.00 · hot:0.27 · rising:0.35 · durable:0.43 · board:durable · trend:down* Python browser sandbox. #### [run-llama/llama_docs_bot](https://github.com/run-llama/llama_docs_bot) *Jupyter Notebook · ★192 · MIT · — · score:0.00 · hot:0.27 · rising:0.35 · durable:0.43 · board:durable · trend:down* Bottoms Up Development with LlamaIndex - Building a Documentation Chatbot #### [openai/distribution_augmentation](https://github.com/openai/distribution_augmentation) *Python · ★132 · MIT · — · score:0.00 · hot:0.27 · rising:0.35 · durable:0.42 · board:durable · trend:down* Code for the paper, "Distribution Augmentation for Generative Modeling", ICML 2020. #### [allenai/catwalk](https://github.com/allenai/catwalk) *Python · ★154 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.34 · durable:0.43 · board:durable · trend:down* This project studies the performance and robustness of language models and task-adaptation methods. #### [simonw/llm-gpt4all](https://github.com/simonw/llm-gpt4all) *Python · ★264 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.34 · durable:0.43 · board:durable · trend:down* Plugin for LLM adding support for the GPT4All collection of models #### [google-deepmind/mc_gradients](https://github.com/google-deepmind/mc_gradients) *Jupyter Notebook · ★173 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.35 · durable:0.42 · board:durable · trend:down* #### [allenai/document-qa](https://github.com/allenai/document-qa) *Python · ★437 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.35 · durable:0.38 · board:durable · trend:down* #### [google-research/cad-estate](https://github.com/google-research/cad-estate) *Python · ★126 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.34 · durable:0.43 · board:durable · trend:down* #### [google-deepmind/Temporal-3D-Pose-Kinetics](https://github.com/google-deepmind/Temporal-3D-Pose-Kinetics) *Python · ★226 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.34 · durable:0.41 · board:durable · trend:down* Exploiting temporal context for 3D human pose estimation in the wild: 3D poses for the Kinetics dataset #### [yoheinakajima/idealist](https://github.com/yoheinakajima/idealist) *Python · ★188 · no-license · — · score:0.00 · hot:0.27 · rising:0.32 · durable:0.41 · board:durable · trend:down* infinite idea generator #### [huggingface/flux-fast](https://github.com/huggingface/flux-fast) *Python · ★166 · no-license · — · score:0.00 · hot:0.27 · rising:0.33 · durable:0.41 · board:durable · trend:down* Making Flux go brrr on GPUs. #### [allenai/real-toxicity-prompts](https://github.com/allenai/real-toxicity-prompts) *Jupyter Notebook · ★230 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.34 · durable:0.42 · board:durable · trend:down* #### [google-deepmind/gqn-datasets](https://github.com/google-deepmind/gqn-datasets) *Python · ★273 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.34 · durable:0.40 · board:durable · trend:down* Datasets used to train Generative Query Networks (GQNs) in the ‘Neural Scene Representation and Rendering’ paper. #### [huggingface/visual-blocks-custom-components](https://github.com/huggingface/visual-blocks-custom-components) *TypeScript · ★124 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.34 · durable:0.43 · board:durable · trend:down* Custom Hugging Face Nodes for Google Visual Blocks for ML #### [lucidrains/titok-pytorch](https://github.com/lucidrains/titok-pytorch) *Python · ★182 · MIT · — · score:0.00 · hot:0.27 · rising:0.31 · durable:0.50 · board:durable · trend:down* Implementation of TiTok, proposed by Bytedance in "An Image is Worth 32 Tokens for Reconstruction and Generation" #### [allenai/hidden-networks](https://github.com/allenai/hidden-networks) *Python · ★196 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.34 · durable:0.41 · board:durable · trend:down* #### [weaviate/gorilla](https://github.com/weaviate/gorilla) *Jupyter Notebook · ★141 · no-license · — · score:0.00 · hot:0.27 · rising:0.33 · durable:0.41 · board:durable · trend:down* Research repository on interfacing LLMs with Weaviate APIs. Inspired by the Berkeley Gorilla LLM. #### [simonw/ca-fires-history](https://github.com/simonw/ca-fires-history) *? · ★217 · no-license · — · score:0.00 · hot:0.27 · rising:0.32 · durable:0.39 · board:durable · trend:down* Tracking fire data from www.fire.ca.gov #### [huggingface/instruction-tuned-sd](https://github.com/huggingface/instruction-tuned-sd) *Python · ★248 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.32 · durable:0.43 · board:durable · trend:down* Code for instruction-tuning Stable Diffusion. #### [allenai/tpu_pretrain](https://github.com/allenai/tpu_pretrain) *Python · ★137 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.34 · durable:0.41 · board:durable · trend:down* LM Pretraining with PyTorch/TPU #### [google-research/spherical-cnn](https://github.com/google-research/spherical-cnn) *Python · ★136 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.32 · durable:0.41 · board:durable · trend:down* #### [simonw/datasette-app](https://github.com/simonw/datasette-app) *JavaScript · ★134 · no-license · — · score:0.00 · hot:0.27 · rising:0.32 · durable:0.39 · board:durable · trend:down* The Datasette macOS application #### [google-research/composed_image_retrieval](https://github.com/google-research/composed_image_retrieval) *Shell · ★195 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.32 · durable:0.39 · board:durable · trend:down* #### [allenai/embodied-clip](https://github.com/allenai/embodied-clip) *Python · ★129 · Apache-2.0 · — · score:0.00 · hot:0.27 · rising:0.34 · durable:0.43 · board:durable · trend:down* Official codebase for EmbCLIP #### [jxnl/magic-text](https://github.com/jxnl/magic-text) *TypeScript · ★153 · MIT · — · score:0.00 · hot:0.26 · rising:0.33 · durable:0.41 · board:durable · trend:down* #### [simonw/cougar-or-not](https://github.com/simonw/cougar-or-not) *Jupyter Notebook · ★127 · no-license · — · score:0.00 · hot:0.26 · rising:0.32 · durable:0.36 · board:durable · trend:down* An API for identifying cougars v.s. bobcats v.s. other USA cat species #### [google-deepmind/pg19](https://github.com/google-deepmind/pg19) *? · ★254 · Apache-2.0 · — · score:0.00 · hot:0.26 · rising:0.34 · durable:0.42 · board:durable · trend:down* #### [replicate/dreambooth](https://github.com/replicate/dreambooth) *Python · ★150 · Apache-2.0 · — · score:0.00 · hot:0.26 · rising:0.34 · durable:0.39 · board:durable · trend:down* A Cog model that takes training images as input and generates custom Stable Diffusion model weights as output #### [allenai/allentune](https://github.com/allenai/allentune) *Python · ★140 · Apache-2.0 · — · score:0.00 · hot:0.26 · rising:0.34 · durable:0.41 · board:durable · trend:down* Hyperparameter Search for AllenNLP #### [yoheinakajima/wrappers_delight](https://github.com/yoheinakajima/wrappers_delight) *Python · ★106 · MIT · — · score:0.00 · hot:0.26 · rising:0.33 · durable:0.41 · board:durable · trend:down* A simple wrapper for OpenAI to log input/outputs. #### [allenai/peS2o](https://github.com/allenai/peS2o) *Python · ★183 · Apache-2.0 · — · score:0.00 · hot:0.26 · rising:0.31 · durable:0.42 · board:durable · trend:down* Pretraining Efficiently on S2ORC! #### [google-deepmind/rgb_stacking](https://github.com/google-deepmind/rgb_stacking) *Python · ★129 · Apache-2.0 · — · score:0.00 · hot:0.26 · rising:0.34 · durable:0.41 · board:durable · trend:down* #### [vercel/reactions](https://github.com/vercel/reactions) *JavaScript · ★313 · MIT · — · score:0.00 · hot:0.26 · rising:0.34 · durable:0.42 · board:durable · trend:down* Next.js Incremental Static Regeneration Demo #### [allenai/unified-io-inference](https://github.com/allenai/unified-io-inference) *Jupyter Notebook · ★231 · Apache-2.0 · — · score:0.00 · hot:0.26 · rising:0.34 · durable:0.40 · board:durable · trend:down* #### [yoheinakajima/predictivechat](https://github.com/yoheinakajima/predictivechat) *Python · ★105 · MIT · — · score:0.00 · hot:0.26 · rising:0.32 · durable:0.42 · board:durable · trend:down* Demo of AI chatbot that predicts user message to generate response quickly. #### [simonw/action-transcription](https://github.com/simonw/action-transcription) *Python · ★189 · Apache-2.0 · — · score:0.00 · hot:0.26 · rising:0.31 · durable:0.36 · board:durable · trend:down* A tool for creating a repository of transcribed videos #### [allenai/SciREX](https://github.com/allenai/SciREX) *Python · ★140 · Apache-2.0 · — · score:0.00 · hot:0.26 · rising:0.33 · durable:0.39 · board:durable · trend:down* Data/Code Repository for https://api.semanticscholar.org/CorpusID:218470122 #### [allenai/gooaq](https://github.com/allenai/gooaq) *Python · ★132 · no-license · — · score:0.00 · hot:0.26 · rising:0.32 · durable:0.39 · board:durable · trend:down* Question-answers, collected from Google #### [allenai/savn](https://github.com/allenai/savn) *Python · ★195 · Apache-2.0 · — · score:0.00 · hot:0.26 · rising:0.33 · durable:0.36 · board:durable · trend:down* Learning to Learn how to Learn: Self-Adaptive Visual Navigation using Meta-Learning (https://arxiv.org/abs/1812.00971) #### [replicate/latent-consistency-model](https://github.com/replicate/latent-consistency-model) *Python · ★196 · MIT · — · score:0.00 · hot:0.26 · rising:0.33 · durable:0.41 · board:durable · trend:down* Run Latent Consistency Models on your Mac #### [simonw/download-esm](https://github.com/simonw/download-esm) *Python · ★125 · Apache-2.0 · — · score:0.00 · hot:0.26 · rising:0.32 · durable:0.40 · board:durable · trend:down* Download ESM modules from npm and jsdelivr #### [simonw/geocoders](https://github.com/simonw/geocoders) *Python · ★183 · BSD-2-Clause · — · score:0.00 · hot:0.26 · rising:0.32 · durable:0.38 · board:durable · trend:down* Ultra simple API for geocoding a single string against various web services. #### [google-deepmind/multi_object_datasets](https://github.com/google-deepmind/multi_object_datasets) *Python · ★284 · Apache-2.0 · — · score:0.00 · hot:0.26 · rising:0.33 · durable:0.40 · board:durable · trend:down* Multi-object image datasets with ground-truth segmentation masks and generative factors. #### [yoheinakajima/asymmetrix](https://github.com/yoheinakajima/asymmetrix) *Python · ★132 · MIT · — · score:0.00 · hot:0.26 · rising:0.32 · durable:0.41 · board:durable · trend:down* #### [vercel/install-node](https://github.com/vercel/install-node) *Shell · ★144 · MIT · — · score:0.00 · hot:0.26 · rising:0.33 · durable:0.40 · board:durable · trend:down* Simple one-liner shell script that installs official Node.js binaries #### [simonw/1991-WWW-NeXT-Implementation](https://github.com/simonw/1991-WWW-NeXT-Implementation) *Objective-C · ★100 · no-license · — · score:0.00 · hot:0.26 · rising:0.27 · durable:0.39 · board:durable · trend:down* #### [simonw/strip-tags](https://github.com/simonw/strip-tags) *Python · ★359 · Apache-2.0 · — · score:0.00 · hot:0.26 · rising:0.31 · durable:0.47 · board:durable · trend:down* CLI tool for stripping tags from HTML #### [simonw/sqlite-history](https://github.com/simonw/sqlite-history) *Python · ★129 · Apache-2.0 · — · score:0.00 · hot:0.25 · rising:0.31 · durable:0.45 · board:durable · trend:down* Track changes to SQLite tables using triggers #### [allenai/dnw](https://github.com/allenai/dnw) *Python · ★138 · NOASSERTION · — · score:0.00 · hot:0.25 · rising:0.31 · durable:0.36 · board:durable · trend:down* Discovering Neural Wirings (https://arxiv.org/abs/1906.00586) #### [allenai/PRIMER](https://github.com/allenai/PRIMER) *Python · ★157 · Apache-2.0 · — · score:0.00 · hot:0.25 · rising:0.32 · durable:0.37 · board:durable · trend:down* The official code for PRIMERA: Pyramid-based Masked Sentence Pre-training for Multi-document Summarization #### [simonw/django-openid](https://github.com/simonw/django-openid) *Python · ★164 · no-license · — · score:0.00 · hot:0.25 · rising:0.29 · durable:0.34 · board:durable · trend:down* A modern library for integrating OpenID with Django - incomplete, but really nearly there (promise) #### [google-deepmind/functa](https://github.com/google-deepmind/functa) *Python · ★161 · Apache-2.0 · — · score:0.00 · hot:0.25 · rising:0.31 · durable:0.41 · board:durable · trend:down* #### [simonw/advent-of-code-2022-in-rust](https://github.com/simonw/advent-of-code-2022-in-rust) *Rust · ★124 · no-license · — · score:0.00 · hot:0.24 · rising:0.28 · durable:0.38 · board:durable · trend:down* Copilot-assisted Advent of Code 2022 to learn Rust #### [lucidrains/gradnorm-pytorch](https://github.com/lucidrains/gradnorm-pytorch) *Python · ★127 · MIT · — · score:0.00 · hot:0.24 · rising:0.28 · durable:0.46 · board:durable · trend:down* A practical implementation of GradNorm, Gradient Normalization for Adaptive Loss Balancing, in Pytorch #### [lucidrains/diffusion-policy](https://github.com/lucidrains/diffusion-policy) *Python · ★135 · MIT · — · score:0.00 · hot:0.23 · rising:0.25 · durable:0.41 · board:durable · trend:down* Implementation of Diffusion Policy, Toyota Research's supposed breakthrough in leveraging DDPMs for learning policies for real-world Robotics #### [swyxio/async-render-toolbox](https://github.com/swyxio/async-render-toolbox) *JavaScript · ★320 · MIT · — · score:0.00 · hot:0.20 · rising:0.23 · durable:0.39 · board:durable · trend:down* BECAUSE PERFORMANCE SHOULD BE SEXY #### [swyxio/fresh](https://github.com/swyxio/fresh) *? · ★135 · no-license · — · score:0.00 · hot:0.19 · rising:0.20 · durable:0.36 · board:durable · trend:down* Community curated lists that never go out of style! 🍅