[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-0xSojalSec--LLMs-local":3,"tool-0xSojalSec--LLMs-local":61},[4,18,26,36,44,53],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":17},4358,"openclaw","openclaw\u002Fopenclaw","OpenClaw 是一款专为个人打造的本地化 AI 助手，旨在让你在自己的设备上拥有完全可控的智能伙伴。它打破了传统 AI 助手局限于特定网页或应用的束缚，能够直接接入你日常使用的各类通讯渠道，包括微信、WhatsApp、Telegram、Discord、iMessage 等数十种平台。无论你在哪个聊天软件中发送消息，OpenClaw 都能即时响应，甚至支持在 macOS、iOS 和 Android 设备上进行语音交互，并提供实时的画布渲染功能供你操控。\n\n这款工具主要解决了用户对数据隐私、响应速度以及“始终在线”体验的需求。通过将 AI 部署在本地，用户无需依赖云端服务即可享受快速、私密的智能辅助，真正实现了“你的数据，你做主”。其独特的技术亮点在于强大的网关架构，将控制平面与核心助手分离，确保跨平台通信的流畅性与扩展性。\n\nOpenClaw 非常适合希望构建个性化工作流的技术爱好者、开发者，以及注重隐私保护且不愿被单一生态绑定的普通用户。只要具备基础的终端操作能力（支持 macOS、Linux 及 Windows WSL2），即可通过简单的命令行引导完成部署。如果你渴望拥有一个懂你",349277,3,"2026-04-06T06:32:30",[13,14,15,16],"Agent","开发框架","图像","数据工具","ready",{"id":19,"name":20,"github_repo":21,"description_zh":22,"stars":23,"difficulty_score":10,"last_commit_at":24,"category_tags":25,"status":17},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,"2026-04-05T11:01:52",[14,15,13],{"id":27,"name":28,"github_repo":29,"description_zh":30,"stars":31,"difficulty_score":32,"last_commit_at":33,"category_tags":34,"status":17},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",150720,2,"2026-04-11T11:33:10",[14,13,35],"语言模型",{"id":37,"name":38,"github_repo":39,"description_zh":40,"stars":41,"difficulty_score":32,"last_commit_at":42,"category_tags":43,"status":17},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",108322,"2026-04-10T11:39:34",[14,15,13],{"id":45,"name":46,"github_repo":47,"description_zh":48,"stars":49,"difficulty_score":32,"last_commit_at":50,"category_tags":51,"status":17},6121,"gemini-cli","google-gemini\u002Fgemini-cli","gemini-cli 是一款由谷歌推出的开源 AI 命令行工具，它将强大的 Gemini 大模型能力直接集成到用户的终端环境中。对于习惯在命令行工作的开发者而言，它提供了一条从输入提示词到获取模型响应的最短路径，无需切换窗口即可享受智能辅助。\n\n这款工具主要解决了开发过程中频繁上下文切换的痛点，让用户能在熟悉的终端界面内直接完成代码理解、生成、调试以及自动化运维任务。无论是查询大型代码库、根据草图生成应用，还是执行复杂的 Git 操作，gemini-cli 都能通过自然语言指令高效处理。\n\n它特别适合广大软件工程师、DevOps 人员及技术研究人员使用。其核心亮点包括支持高达 100 万 token 的超长上下文窗口，具备出色的逻辑推理能力；内置 Google 搜索、文件操作及 Shell 命令执行等实用工具；更独特的是，它支持 MCP（模型上下文协议），允许用户灵活扩展自定义集成，连接如图像生成等外部能力。此外，个人谷歌账号即可享受免费的额度支持，且项目基于 Apache 2.0 协议完全开源，是提升终端工作效率的理想助手。",100752,"2026-04-10T01:20:03",[52,13,15,14],"插件",{"id":54,"name":55,"github_repo":56,"description_zh":57,"stars":58,"difficulty_score":32,"last_commit_at":59,"category_tags":60,"status":17},4721,"markitdown","microsoft\u002Fmarkitdown","MarkItDown 是一款由微软 AutoGen 团队打造的轻量级 Python 工具，专为将各类文件高效转换为 Markdown 格式而设计。它支持 PDF、Word、Excel、PPT、图片（含 OCR）、音频（含语音转录）、HTML 乃至 YouTube 链接等多种格式的解析，能够精准提取文档中的标题、列表、表格和链接等关键结构信息。\n\n在人工智能应用日益普及的今天，大语言模型（LLM）虽擅长处理文本，却难以直接读取复杂的二进制办公文档。MarkItDown 恰好解决了这一痛点，它将非结构化或半结构化的文件转化为模型“原生理解”且 Token 效率极高的 Markdown 格式，成为连接本地文件与 AI 分析 pipeline 的理想桥梁。此外，它还提供了 MCP（模型上下文协议）服务器，可无缝集成到 Claude Desktop 等 LLM 应用中。\n\n这款工具特别适合开发者、数据科学家及 AI 研究人员使用，尤其是那些需要构建文档检索增强生成（RAG）系统、进行批量文本分析或希望让 AI 助手直接“阅读”本地文件的用户。虽然生成的内容也具备一定可读性，但其核心优势在于为机器",93400,"2026-04-06T19:52:38",[52,14],{"id":62,"github_repo":63,"name":64,"description_en":65,"description_zh":66,"ai_summary_zh":66,"readme_en":67,"readme_zh":68,"quickstart_zh":69,"use_case_zh":70,"hero_image_url":71,"owner_login":72,"owner_name":73,"owner_avatar_url":74,"owner_bio":75,"owner_company":76,"owner_location":77,"owner_email":76,"owner_twitter":78,"owner_website":79,"owner_url":80,"languages":76,"stars":81,"forks":82,"last_commit_at":83,"license":76,"difficulty_score":32,"env_os":84,"env_gpu":85,"env_ram":86,"env_deps":87,"category_tags":98,"github_topics":76,"view_count":32,"oss_zip_url":76,"oss_zip_packed_at":76,"status":17,"created_at":100,"updated_at":101,"faqs":102,"releases":103},6630,"0xSojalSec\u002FLLMs-local","LLMs-local"," list of awesome platforms, tools, and resources   run for LLMs locally","LLMs-local 是一个精心整理的开源资源清单，旨在帮助用户在本地设备上运行大型语言模型（LLM）。它汇集了从推理平台、底层引擎到用户界面、模型库及硬件指南的全方位工具，让使用者无需依赖云端服务，即可在个人电脑上安全、离线地部署和使用 AI 能力。\n\n这一资源库有效解决了数据隐私担忧、云端 API 成本高昂以及网络延迟等痛点，让大模型的运行完全掌握在用户手中。无论是希望保护敏感数据的普通用户，还是追求低延迟和高定制化的开发者与研究人员，都能在此找到适合的解决方案。清单中不仅收录了 LM Studio、Ollama 等易于上手的桌面应用，还涵盖了 llama.cpp、vllm 等高性能推理引擎，甚至支持利用日常设备组建家庭 AI 集群的独特方案。\n\n此外，LLMs-local 还提供了关于模型微调、智能体框架、检索增强生成（RAG）以及提示词工程的专业教程与社区链接。无论你是想尝试本地运行代码助手的技术爱好者，还是需要构建复杂 AI 应用的专业团队，这份指南都能为你提供从零开始到进阶优化的完整路径，助力轻松开启本地大模型之旅。","# LLMs-local\n list of awesome platforms, tools, and resources   run for LLMs locally\n\n \n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002F0xSojalSec_LLMs-local_readme_81d345b929d7.jpeg\">\n\n## Table of Contents\n\n- [Inference platforms](#inference-platforms)\n- [Inference engines](#inference-engines)\n- [User Interfaces](#user-interfaces)\n- [Large Language Models](#large-language-models)\n  - [Explorers, Benchmarks, Leaderboards](#explorers-benchmarks-leaderboards)\n  - [Model providers](#model-providers)\n  - [Specific models](#specific-models)\n    - [General purpose](#general-purpose)\n    - [Coding](#coding)\n    - [Multimodal](#multimodal)\n    - [Image](#image)\n    - [Audio](#audio)\n    - [Miscellaneous](#miscellaneous)\n- [Tools](#tools)\n  - [Models](#models)\n  - [Agent Frameworks](#agent-frameworks)\n  - [Model Context Protocol](#model-context-protocol)\n  - [Retrieval-Augmented Generation](#retrieval-augmented-generation)\n  - [Coding Agents](#coding-agents)\n  - [Computer Use](#computer-use)\n  - [Browser Automation](#browser-automation)\n  - [Memory Management](#memory-management)\n  - [Testing, Evaluation, and Observability](#testing-evaluation-and-observability)\n  - [Research](#research)\n  - [Training and Fine-tuning](#training-and-fine-tuning)\n  - [Miscellaneous](#miscellaneous-1)\n- [Hardware](#hardware)\n- [Tutorials](#tutorials)\n  - [Models](#models-1)\n  - [Prompt Engineering](#prompt-engineering)\n  - [Context Engineering](#context-engineering)\n  - [Inference](#inference)\n  - [Agents](#agents)\n  - [Retrieval-Augmented Generation](#retrieval-augmented-generation-1)\n  - [Miscellaneous](#miscellaneous-2)\n- [Communities](#communities)\n\n## Inference platforms\n\n- [LM Studio](https:\u002F\u002Flmstudio.ai\u002F) - discover, download and run local LLMs\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmenloresearch\u002Fjan?style=social\" height=\"17\" align=\"texttop\"\u002F> [jan](https:\u002F\u002Fgithub.com\u002Fmenloresearch\u002Fjan) - an open source alternative to ChatGPT that runs 100% offline on your computer\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmudler\u002FLocalAI?style=social\" height=\"17\" align=\"texttop\"\u002F> [LocalAI](https:\u002F\u002Fgithub.com\u002Fmudler\u002FLocalAI) -  the free, open-source alternative to OpenAI, Claude and others\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FChatBoxAI\u002FChatBox?style=social\" height=\"17\" align=\"texttop\"\u002F> [ChatBox](https:\u002F\u002Fgithub.com\u002FChatBoxAI\u002FChatBox) - user-friendly desktop client app for AI models\u002FLLMs\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flemonade-sdk\u002Flemonade?style=social\" height=\"17\" align=\"texttop\"\u002F> [lemonade](https:\u002F\u002Fgithub.com\u002Flemonade-sdk\u002Flemonade) - a local LLM server with GPU and NPU Acceleration\n\n[Back to Table of Contents](#table-of-contents)\n\n## Inference engines\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Follama\u002Follama?style=social\" height=\"17\" align=\"texttop\"\u002F> [ollama](https:\u002F\u002Fgithub.com\u002Follama\u002Follama) - get up and running with LLMs\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fggml-org\u002Fllama.cpp?style=social\" height=\"17\" align=\"texttop\"\u002F> [llama.cpp](https:\u002F\u002Fgithub.com\u002Fggml-org\u002Fllama.cpp) - LLM inference in C\u002FC++\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fvllm-project\u002Fvllm?style=social\" height=\"17\" align=\"texttop\"\u002F> [vllm](https:\u002F\u002Fgithub.com\u002Fvllm-project\u002Fvllm) - a high-throughput and memory-efficient inference and serving engine for LLMs\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fexo-explore\u002Fexo?style=social\" height=\"17\" align=\"texttop\"\u002F> [exo](https:\u002F\u002Fgithub.com\u002Fexo-explore\u002Fexo) - run your own AI cluster at home with everyday devices\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002FBitNet?style=social\" height=\"17\" align=\"texttop\"\u002F> [BitNet](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FBitNet) - official inference framework for 1-bit LLMs\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsgl-project\u002Fsglang?style=social\" height=\"17\" align=\"texttop\"\u002F> [sglang](https:\u002F\u002Fgithub.com\u002Fsgl-project\u002Fsglang) - a fast serving framework for large language models and vision language models\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FGeeeekExplorer\u002Fnano-vllm?style=social\" height=\"17\" align=\"texttop\"\u002F> [Nano-vLLM](https:\u002F\u002Fgithub.com\u002FGeeeekExplorer\u002Fnano-vllm) - a lightweight vLLM implementation built from scratch\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FLostRuins\u002Fkoboldcpp?style=social\" height=\"17\" align=\"texttop\"\u002F> [koboldcpp](https:\u002F\u002Fgithub.com\u002FLostRuins\u002Fkoboldcpp) - run GGUF models easily with a KoboldAI UI\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fgpustack\u002Fgpustack?style=social\" height=\"17\" align=\"texttop\"\u002F> [gpustack](https:\u002F\u002Fgithub.com\u002Fgpustack\u002Fgpustack) - simple, scalable AI model deployment on GPU clusters\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fml-explore\u002Fmlx-lm?style=social\" height=\"17\" align=\"texttop\"\u002F> [mlx-lm](https:\u002F\u002Fgithub.com\u002Fml-explore\u002Fmlx-lm) - generate text and fine-tune large language models on Apple silicon with MLX\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fb4rtaz\u002Fdistributed-llama?style=social\" height=\"17\" align=\"texttop\"\u002F> [distributed-llama](https:\u002F\u002Fgithub.com\u002Fb4rtaz\u002Fdistributed-llama) - connect home devices into a powerful cluster to accelerate LLM inference\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fikawrakow\u002Fik_llama.cpp?style=social\" height=\"17\" align=\"texttop\"\u002F> [ik_llama.cpp](https:\u002F\u002Fgithub.com\u002Fikawrakow\u002Fik_llama.cpp) - llama.cpp fork with additional SOTA quants and improved performance\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FFastFlowLM\u002FFastFlowLM?style=social\" height=\"17\" align=\"texttop\"\u002F> [FastFlowLM](https:\u002F\u002Fgithub.com\u002FFastFlowLM\u002FFastFlowLM) - run LLMs on AMD Ryzen™ AI NPUs\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fnlzy\u002Fvllm-gfx906?style=social\" height=\"17\" align=\"texttop\"\u002F> [vllm-gfx906](https:\u002F\u002Fgithub.com\u002Fnlzy\u002Fvllm-gfx906) - vLLM for AMD gfx906 GPUs, e.g. Radeon VII \u002F MI50 \u002F MI60\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fintel\u002Fllm-scaler?style=social\" height=\"17\" align=\"texttop\"\u002F> [llm-scaler](https:\u002F\u002Fgithub.com\u002Fintel\u002Fllm-scaler) - run LLMs on Intel Arc™ Pro B60 GPUs\n\n[Back to Table of Contents](#table-of-contents)\n\n## User Interfaces\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fopen-webui\u002Fopen-webui?style=social\" height=\"17\" align=\"texttop\"\u002F> [Open WebUI](https:\u002F\u002Fgithub.com\u002Fopen-webui\u002Fopen-webui) - User-friendly AI Interface (Supports Ollama, OpenAI API, ...)\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flobehub\u002Flobe-chat?style=social\" height=\"17\" align=\"texttop\"\u002F> [Lobe Chat](https:\u002F\u002Fgithub.com\u002Flobehub\u002Flobe-chat) - an open-source, modern design AI chat framework\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Foobabooga\u002Ftext-generation-webui?style=social\" height=\"17\" align=\"texttop\"\u002F> [Text generation web UI](https:\u002F\u002Fgithub.com\u002Foobabooga\u002Ftext-generation-webui) - LLM UI with advanced features, easy setup, and multiple backend support\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FSillyTavern\u002FSillyTavern?style=social\" height=\"17\" align=\"texttop\"\u002F> [SillyTavern](https:\u002F\u002Fgithub.com\u002FSillyTavern\u002FSillyTavern) - LLM Frontend for Power Users\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fn4ze3m\u002Fpage-assist?style=social\" height=\"17\" align=\"texttop\"\u002F> [Page Assist](https:\u002F\u002Fgithub.com\u002Fn4ze3m\u002Fpage-assist) - Use your locally running AI models to assist you in your web browsing\n\n[Back to Table of Contents](#table-of-contents)\n\n## Large Language Models\n\n### Explorers, Benchmarks, Leaderboards\n\n- [AI Models & API Providers Analysis](https:\u002F\u002Fartificialanalysis.ai\u002F) - understand the AI landscape to choose the best model and provider for your use case\n- [LLM Explorer](https:\u002F\u002Fllm-explorer.com\u002F) - explore list of the open-source LLM models\n- [Dubesor LLM Benchmark table](https:\u002F\u002Fdubesor.de\u002Fbenchtable) - small-scale manual performance comparison benchmark\n- [oobabooga benchmark](https:\u002F\u002Foobabooga.github.io\u002Fbenchmark.html) - a list sorted by size (on disk) for each score\n\n[Back to Table of Contents](#table-of-contents)\n\n### Model providers\n\n- [Qwen](https:\u002F\u002Fhuggingface.co\u002FQwen) - powered by Alibaba Cloud\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FMistral%20AI-%23FA520F?logo=mistralai&logoColor=%23FFFFFF\" height=\"17\" align=\"texttop\"\u002F> [Mistral AI](https:\u002F\u002Fhuggingface.co\u002Fmistralai) - a pioneering French artificial intelligence startup\n- [Tencent](https:\u002F\u002Fhuggingface.co\u002Ftencent) - a profile of a Chinese multinational technology conglomerate and holding company\n- [Unsloth AI](https:\u002F\u002Fhuggingface.co\u002Funsloth) - focusing on making AI more accessible to everyone (GGUFs etc.)\n- [bartowski](https:\u002F\u002Fhuggingface.co\u002Fbartowski) - providing GGUF versions of popular LLMs\n- [Beijing Academy of Artificial Intelligence](https:\u002F\u002Fhuggingface.co\u002FBAAI) - a private non-profit organization engaged in AI research and development\n- [Open Thoughts](https:\u002F\u002Fhuggingface.co\u002Fopen-thoughts) - a team of researchers and engineers curating the best open reasoning datasets\n\n[Back to Table of Contents](#table-of-contents)\n\n### Specific models\n\n#### General purpose\n\n- [Qwen3-Next](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002FQwen\u002Fqwen3-next-68c25fd6838e585db8eeea9d) - a collection of the latest generation Qwen LLMs\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FGoogle-%234285F4?logo=google&logoColor=red\" height=\"17\" align=\"texttop\"\u002F> [Gemma 3](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fgoogle\u002Fgemma-3-release-67c6c6f89c4f76621268bb6d) - a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FOpenAI-%23412991?logo=openai\" height=\"17\" align=\"texttop\"\u002F> [gpt-oss](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fopenai\u002Fgpt-oss-68911959590a1634ba11c7a4) - a collection of open-weight models from OpenAI, designed for powerful reasoning, agentic tasks, and versatile developer use cases\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FMistral%20AI-%23FA520F?logo=mistralai&logoColor=%23FFFFFF\" height=\"17\" align=\"texttop\"\u002F> [Ministral 3](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fmistralai\u002Fministral-3) - a collection of edge models, with base, instruct and reasoning variants, in 3 different sizes: 3B, 8B and 14B, all with vision capabilities\n- [GLM-4.5](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fzai-org\u002Fglm-45-687c621d34bda8c9e4bf503b) - a collection of hybrid reasoning models designed for intelligent agents\n- [Hunyuan](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Ftencent\u002Fhunyuan-dense-model-6890632cda26b19119c9c5e7) - a collection of Tencent's open-source efficient LLMs designed for versatile deployment across diverse computational environments\n- [Phi-4-mini-instruct](https:\u002F\u002Fhuggingface.co\u002Fmicrosoft\u002FPhi-4-mini-instruct) - a lightweight open model built upon synthetic data and filtered publicly available websites\n- [NVIDIA Nemotron v3](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fnvidia\u002Fnvidia-nemotron-v3) - a family of open models from NVIDIA with open weights, training data and recipes, delivering leading efficiency and accuracy for building specialized AI agents\n- [Llama Nemotron](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fnvidia\u002Fllama-nemotron-67d92346030a2691293f200b) - a collection of open, production-ready enterprise models from NVIDIA\n- [OpenReasoning-Nemotron](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fnvidia\u002Fopenreasoning-nemotron-687730dae0170059860f1f01) - a collection of models from NVIDIA, trained on 5M reasoning traces for math, code and science\n- [Granite 4.0](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fibm-granite\u002Fgranite-40-language-models-6811a18b820ef362d9e5a82c) - a collection of lightweight, state-of-the-art open foundation models from IBM that natively support multilingual capabilities, a wide range of coding tasks—including fill-in-the-middle (FIM) code completion—retrieval-augmented generation (RAG), tool usage and structured JSON output\n- [EXAONE-4.0](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002FLGAI-EXAONE\u002Fexaone-40-686b2e0069800c835ed48375) - a collection of LLMs from LG AI Research, integrating non-reasoning and reasoning modes\n- [ERNIE 4.5](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fbaidu\u002Fernie-45-6861cd4c9be84540645f35c9) - a collection of large-scale multimodal models from Baidu\n- [Seed-OSS](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002FByteDance-Seed\u002Fseed-oss-68a609f4201e788db05b5dcd) - a collection of LLMs developed by ByteDance's Seed Team, designed for powerful long-context, reasoning, agent and general capabilities, and versatile developer-friendly features\n\n[Back to Table of Contents](#table-of-contents)\n\n#### Coding\n\n- [Qwen3-Coder](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002FQwen\u002Fqwen3-coder-687fc861e53c939e52d52d10) - a collection of the Qwen's most agentic code models to date\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FMistral%20AI-%23FA520F?logo=mistralai&logoColor=%23FFFFFF\" height=\"17\" align=\"texttop\"\u002F> [Devstral 2](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fmistralai\u002Fdevstral-2) - a couple of agentic LLMs for software engineering tasks, excelling at using tools to explore codebases, edit multiple files, and power SWE Agents\n- [Mellum-4b-base](https:\u002F\u002Fhuggingface.co\u002FJetBrains\u002FMellum-4b-base) - an LLM from JetBrains, optimized for code-related tasks\n- [OlympicCoder-32B](https:\u002F\u002Fhuggingface.co\u002Fopen-r1\u002FOlympicCoder-32B) - a code model that achieves very strong performance on competitive coding benchmarks such as LiveCodeBench and the 2024 International Olympiad in Informatics\n- [NextCoder](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fmicrosoft\u002Fnextcoder-6815ee6bfcf4e42f20d45028) - a family of code-editing LLMs developed using the Qwen2.5-Coder Instruct variants as base\n\n[Back to Table of Contents](#table-of-contents)\n\n#### Multimodal\n\n- [Qwen3-Omni](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002FQwen\u002Fqwen3-omni-68d100a86cd0906843ceccbe) - a collection of the natively end-to-end multilingual omni-modal foundation models from Qwen\n\n[Back to Table of Contents](#table-of-contents)\n\n#### Image\n\n- [Qwen-Image](https:\u002F\u002Fhuggingface.co\u002FQwen\u002FQwen-Image) - an image generation foundation model in the Qwen series that achieves significant advances in complex text rendering and precise image editing\n- [Qwen-Image-Edit-2509](https:\u002F\u002Fhuggingface.co\u002FQwen\u002FQwen-Image-Edit-2509) - the image editing version of Qwen-Image extending the base model's unique text rendering capabilities to image editing tasks, enabling precise text editing\n- [Qwen3-VL](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002FQwen\u002Fqwen3-vl-68d2a7c1b8a8afce4ebd2dbe) - a collection of the most powerful vision-language models in the Qwen series to date\n- [GLM-4.5V](https:\u002F\u002Fhuggingface.co\u002Fzai-org\u002FGLM-4.5V) - a VLLM based on ZhipuAI’s next-generation flagship text foundation model GLM-4.5-Air\n- [HunyuanImage-2.1](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHunyuanImage-2.1) - an efficient diffusion model for high-resolution (2K) text-to-image generation​ \n- [FastVLM](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fapple\u002Ffastvlm-68ac97b9cd5cacefdd04872e) - a collection of VLMs with efficient vision encoding from Apple\n- [MiniCPM-V-4_5](https:\u002F\u002Fhuggingface.co\u002Fopenbmb\u002FMiniCPM-V-4_5) - a GPT-4o Level MLLM for single image, multi image and high-FPS video understanding on your phone\n- [LFM2-VL](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002FLiquidAI\u002Flfm2-vl-68963bbc84a610f7638d5ffa) - a colection of vision-language models, designed for on-device deployment\n- [ClipTagger-12b](https:\u002F\u002Fhuggingface.co\u002Finference-net\u002FClipTagger-12b) -  a vision-language model (VLM) designed for video understanding at massive scale\n\n[Back to Table of Contents](#table-of-contents)\n\n#### Audio\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FMistral%20AI-%23FA520F?logo=mistralai&logoColor=%23FFFFFF\" height=\"17\" align=\"texttop\"\u002F> [Voxtral-Small-24B-2507](https:\u002F\u002Fhuggingface.co\u002Fmistralai\u002FVoxtral-Small-24B-2507) - an enhancement of Mistral Small 3, incorporating state-of-the-art audio input capabilities while retaining best-in-class text performance\n- [chatterbox](https:\u002F\u002Fhuggingface.co\u002FResembleAI\u002Fchatterbox) - first production-grade open-source TTS model\n- [VibeVoice](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fmicrosoft\u002Fvibevoice-68a2ef24a875c44be47b034f) - a collection of frontier text-to-speech models from Microsoft\n- [canary-1b-v2](https:\u002F\u002Fhuggingface.co\u002Fnvidia\u002Fcanary-1b-v2) - a multitask speech transcription and translation model from NVIDIA\n- [parakeet-tdt-0.6b-v3](https:\u002F\u002Fhuggingface.co\u002Fnvidia\u002Fparakeet-tdt-0.6b-v3) - a multilingual speech-to-text model from NVIDIA\n- [Kitten TTS](https:\u002F\u002Fhuggingface.co\u002FKittenML\u002Fmodels) - a collection of open-source realistic text-to-speech models designed for lightweight deployment and high-quality voice synthesis\n\n[Back to Table of Contents](#table-of-contents)\n\n#### Miscellaneous\n\n- [Jan-v1-4B](https:\u002F\u002Fhuggingface.co\u002Fjanhq\u002FJan-v1-4B) - the first release in the Jan Family, designed for agentic reasoning and problem-solving within the Jan App\n- [Jan-nano](https:\u002F\u002Fhuggingface.co\u002FMenlo\u002FJan-nano) - a compact 4-billion parameter language model specifically designed and trained for deep research tasks\n- [Jan-nano-128k](https:\u002F\u002Fhuggingface.co\u002FMenlo\u002FJan-nano-128k) - an enhanced version of Jan-nano features a native 128k context window that enables deeper, more comprehensive research capabilities without the performance degradation typically associated with context extension method\n- [Arch-Router-1.5B](https:\u002F\u002Fhuggingface.co\u002Fkatanemo\u002FArch-Router-1.5B) - the fastest LLM router model that aligns to subjective usage preferences\n- [gpt-oss-safeguard](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fopenai\u002Fgpt-oss-safeguard) - a collection of safety reasoning models built-upon gpt-oss\n- [Qwen3Guard](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002FQwen\u002Fqwen3guard-68d2729abbfae4716f3343a1) - a collection of safety moderation models built upon Qwen3\n- [HunyuanWorld-1](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHunyuanWorld-1) - an open-source 3D world generation model\n- [Hunyuan-GameCraft-1.0](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHunyuan-GameCraft-1.0) - a novel framework for high-dynamic interactive video generation in game environments\n\n[Back to Table of Contents](#table-of-contents)\n\n## Tools\n\n### Models\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Funslothai\u002Funsloth?style=social\" height=\"17\" align=\"texttop\"\u002F> [unsloth](https:\u002F\u002Fgithub.com\u002Funslothai\u002Funsloth) - fine-tuning & reinforcement learning for LLMs\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdottxt-ai\u002Foutlines?style=social\" height=\"17\" align=\"texttop\"\u002F> [outlines](https:\u002F\u002Fgithub.com\u002Fdottxt-ai\u002Foutlines) - structured outputs for LLMs\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fp-e-w\u002Fheretic?style=social\" height=\"17\" align=\"texttop\"\u002F> [heretic](https:\u002F\u002Fgithub.com\u002Fp-e-w\u002Fheretic) - fully automatic censorship removal for language models\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmostlygeek\u002Fllama-swap?style=social\" height=\"17\" align=\"texttop\"\u002F> [llama-swap](https:\u002F\u002Fgithub.com\u002Fmostlygeek\u002Fllama-swap) - reliable model swapping for any local OpenAI compatible server - llama.cpp, vllm, etc.\n\n[Back to Table of Contents](#table-of-contents)\n\n### Agent Frameworks\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FSignificant-Gravitas\u002FAutoGPT?style=social\" height=\"17\" align=\"texttop\"\u002F> [AutoGPT](https:\u002F\u002Fgithub.com\u002FSignificant-Gravitas\u002FAutoGPT) - a powerful platform that allows you to create, deploy, and manage continuous AI agents that automate complex workflows\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flangflow-ai\u002Flangflow?style=social\" height=\"17\" align=\"texttop\"\u002F> [langflow](https:\u002F\u002Fgithub.com\u002Flangflow-ai\u002Flangflow) - a powerful tool for building and deploying AI-powered agents and workflows\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flangchain-ai\u002Flangchain?style=social\" height=\"17\" align=\"texttop\"\u002F> [langchain](https:\u002F\u002Fgithub.com\u002Flangchain-ai\u002Flangchain) - build context-aware reasoning applications\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002Fautogen?style=social\" height=\"17\" align=\"texttop\"\u002F> [autogen](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fautogen) - a programming framework for agentic AI\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMintplex-Labs\u002Fanything-llm?style=social\" height=\"17\" align=\"texttop\"\u002F> [anything-llm](https:\u002F\u002Fgithub.com\u002FMintplex-Labs\u002Fanything-llm) - the all-in-one Desktop & Docker AI application with built-in RAG, AI agents, No-code agent builder, MCP compatibility, and more\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FFlowiseAI\u002FFlowise?style=social\" height=\"17\" align=\"texttop\"\u002F> [Flowise](https:\u002F\u002Fgithub.com\u002FFlowiseAI\u002FFlowise) - build AI agents, visually\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Frun-llama\u002Fllama_index?style=social\" height=\"17\" align=\"texttop\"\u002F> [llama_index](https:\u002F\u002Fgithub.com\u002Frun-llama\u002Fllama_index) - the leading framework for building LLM-powered agents over your data\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FcrewAIInc\u002FcrewAI?style=social\" height=\"17\" align=\"texttop\"\u002F> [crewAI](https:\u002F\u002Fgithub.com\u002FcrewAIInc\u002FcrewAI) - a framework for orchestrating role-playing, autonomous AI agents\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fagno-agi\u002Fagno?style=social\" height=\"17\" align=\"texttop\"\u002F> [agno](https:\u002F\u002Fgithub.com\u002Fagno-agi\u002Fagno) - a full-stack framework for building Multi-Agent Systems with memory, knowledge and reasoning\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsimstudioai\u002Fsim?style=social\" height=\"17\" align=\"texttop\"\u002F> [sim](https:\u002F\u002Fgithub.com\u002Fsimstudioai\u002Fsim) - open-source platform to build and deploy AI agent workflows\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FOpenAI-%23412991?logo=openai\" height=\"17\" align=\"texttop\"\u002F> \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fopenai\u002Fopenai-agents-python?style=social\" height=\"17\" align=\"texttop\"\u002F> [openai-agents-python](https:\u002F\u002Fgithub.com\u002Fopenai\u002Fopenai-agents-python) - a lightweight, powerful framework for multi-agent workflows\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTransformerOptimus\u002FSuperAGI?style=social\" height=\"17\" align=\"texttop\"\u002F> [SuperAGI](https:\u002F\u002Fgithub.com\u002FTransformerOptimus\u002FSuperAGI) - an open-source framework to build, manage and run useful Autonomous AI Agents\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fcamel-ai\u002Fcamel?style=social\" height=\"17\" align=\"texttop\"\u002F> [camel](https:\u002F\u002Fgithub.com\u002Fcamel-ai\u002Fcamel) - the first and the best multi-agent framework\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpydantic\u002Fpydantic-ai?style=social\" height=\"17\" align=\"texttop\"\u002F> [pydantic-ai](https:\u002F\u002Fgithub.com\u002Fpydantic\u002Fpydantic-ai) - a Python agent framework designed to help you quickly, confidently, and painlessly build production grade applications and workflows with Generative AI\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fneuml\u002Ftxtai?style=social\" height=\"17\" align=\"texttop\"\u002F> [txtai](https:\u002F\u002Fgithub.com\u002Fneuml\u002Ftxtai) - all-in-one open-source AI framework for semantic search, LLM orchestration and language model workflows\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002Fagent-framework?style=social\" height=\"17\" align=\"texttop\"\u002F> [agent-framework](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fagent-framework) - a framework for building, orchestrating and deploying AI agents and multi-agent workflows with support for Python and .NET\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fkatanemo\u002Farchgw?style=social\" height=\"17\" align=\"texttop\"\u002F> [archgw](https:\u002F\u002Fgithub.com\u002Fkatanemo\u002Farchgw) - a high-performance proxy server that handles the low-level work in building agents: like applying guardrails, routing prompts to the right agent, and unifying access to LLMs, etc.\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fbadboysm890\u002FClaraVerse?style=social\" height=\"17\" align=\"texttop\"\u002F> [ClaraVerse](https:\u002F\u002Fgithub.com\u002Fbadboysm890\u002FClaraVerse) - privacy-first, fully local AI workspace with Ollama LLM chat, tool calling, agent builder, Stable Diffusion, and embedded n8n-style automation\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdeepsense-ai\u002Fragbits?style=social\" height=\"17\" align=\"texttop\"\u002F> [ragbits](https:\u002F\u002Fgithub.com\u002Fdeepsense-ai\u002Fragbits) - building blocks for rapid development of GenAI applications\n\n[Back to Table of Contents](#table-of-contents)\n\n### Model Context Protocol\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmindsdb\u002Fmindsdb?style=social\" height=\"17\" align=\"texttop\"\u002F> [mindsdb](https:\u002F\u002Fgithub.com\u002Fmindsdb\u002Fmindsdb) - federated query engine for AI - the only MCP Server you'll ever need \n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fgithub\u002Fgithub-mcp-server?style=social\" height=\"17\" align=\"texttop\"\u002F> [github-mcp-server](https:\u002F\u002Fgithub.com\u002Fgithub\u002Fgithub-mcp-server) - GitHub's official MCP Server\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002Fplaywright-mcp?style=social\" height=\"17\" align=\"texttop\"\u002F> [playwright-mcp](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fplaywright-mcp) - Playwright MCP server\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FChromeDevTools\u002Fchrome-devtools-mcp?style=social\" height=\"17\" align=\"texttop\"\u002F> [chrome-devtools-mcp](https:\u002F\u002Fgithub.com\u002FChromeDevTools\u002Fchrome-devtools-mcp) - Chrome DevTools for coding agents\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fczlonkowski\u002Fn8n-mcp?style=social\" height=\"17\" align=\"texttop\"\u002F> [n8n-mcp](https:\u002F\u002Fgithub.com\u002Fczlonkowski\u002Fn8n-mcp) - a MCP for Claude Desktop \u002F Claude Code \u002F Windsurf \u002F Cursor to build n8n workflows for you\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fawslabs\u002Fmcp?style=social\" height=\"17\" align=\"texttop\"\u002F> [awslabs\u002Fmcp](https:\u002F\u002Fgithub.com\u002Fawslabs\u002Fmcp) - AWS MCP Servers — helping you get the most out of AWS, wherever you use MCP\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsooperset\u002Fmcp-atlassian?style=social\" height=\"17\" align=\"texttop\"\u002F> [mcp-atlassian](https:\u002F\u002Fgithub.com\u002Fsooperset\u002Fmcp-atlassian) - MCP server for Atlassian tools (Confluence, Jira)\n\n[Back to Table of Contents](#table-of-contents)\n\n### Retrieval-Augmented Generation\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpathwaycom\u002Fpathway?style=social\" height=\"17\" align=\"texttop\"\u002F> [pathway](https:\u002F\u002Fgithub.com\u002Fpathwaycom\u002Fpathway) - Python ETL framework for stream processing, real-time analytics, LLM pipelines and RAG\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002Fgraphrag?style=social\" height=\"17\" align=\"texttop\"\u002F> [graphrag](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fgraphrag) - a modular graph-based RAG system\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FHKUDS\u002FLightRAG?style=social\" height=\"17\" align=\"texttop\"\u002F> [LightRAG](https:\u002F\u002Fgithub.com\u002FHKUDS\u002FLightRAG) - simple and fast RAG\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdeepset-ai\u002Fhaystack?style=social\" height=\"17\" align=\"texttop\"\u002F> [haystack](https:\u002F\u002Fgithub.com\u002Fdeepset-ai\u002Fhaystack) - AI orchestration framework to build customizable, production-ready LLM applications, best suited for building RAG, question answering, semantic search or conversational agent chatbots\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fvanna-ai\u002Fvanna?style=social\" height=\"17\" align=\"texttop\"\u002F> [vanna](https:\u002F\u002Fgithub.com\u002Fvanna-ai\u002Fvanna) - an open-source Python RAG framework for SQL generation and related functionality\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fgetzep\u002Fgraphiti?style=social\" height=\"17\" align=\"texttop\"\u002F> [graphiti](https:\u002F\u002Fgithub.com\u002Fgetzep\u002Fgraphiti) - build real-time knowledge graphs for AI Agents\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fonyx-dot-app\u002Fonyx?style=social\" height=\"17\" align=\"texttop\"\u002F> [onyx](https:\u002F\u002Fgithub.com\u002Fonyx-dot-app\u002Fonyx) - the AI platform connected to your company's docs, apps, and people\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fzilliztech\u002Fclaude-context?style=social\" height=\"17\" align=\"texttop\"\u002F> [claude-context](https:\u002F\u002Fgithub.com\u002Fzilliztech\u002Fclaude-context) - make entire codebase the context for any coding agent\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpipeshub-ai\u002Fpipeshub-ai?style=social\" height=\"17\" align=\"texttop\"\u002F> [pipeshub-ai](https:\u002F\u002Fgithub.com\u002Fpipeshub-ai\u002Fpipeshub-ai) - a fully extensible and explainable workplace AI platform for enterprise search and workflow automation\n\n[Back to Table of Contents](#table-of-contents)\n\n### Coding Agents\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fzed-industries\u002Fzed?style=social\" height=\"17\" align=\"texttop\"\u002F> [zed](https:\u002F\u002Fgithub.com\u002Fzed-industries\u002Fzed) - a next-generation code editor designed for high-performance collaboration with humans and AI\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAll-Hands-AI\u002FOpenHands?style=social\" height=\"17\" align=\"texttop\"\u002F> [OpenHands](https:\u002F\u002Fgithub.com\u002FAll-Hands-AI\u002FOpenHands) - a platform for software development agents powered by AI\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fcline\u002Fcline?style=social\" height=\"17\" align=\"texttop\"\u002F> [cline](https:\u002F\u002Fgithub.com\u002Fcline\u002Fcline) - autonomous coding agent right in your IDE, capable of creating\u002Fediting files, executing commands, using the browser, and more with your permission every step of the way\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAider-AI\u002Faider?style=social\" height=\"17\" align=\"texttop\"\u002F> [aider](https:\u002F\u002Fgithub.com\u002FAider-AI\u002Faider) - AI pair programming in your terminal\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsst\u002Fopencode?style=social\" height=\"17\" align=\"texttop\"\u002F> [opencode](https:\u002F\u002Fgithub.com\u002Fsst\u002Fopencode) - a AI coding agent built for the terminal\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTabbyML\u002Ftabby?style=social\" height=\"17\" align=\"texttop\"\u002F> [tabby](https:\u002F\u002Fgithub.com\u002FTabbyML\u002Ftabby) -  an open-source GitHub Copilot alternative, set up your own LLM-powered code completion server\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fcontinuedev\u002Fcontinue?style=social\" height=\"17\" align=\"texttop\"\u002F> [continue](https:\u002F\u002Fgithub.com\u002Fcontinuedev\u002Fcontinue) - create, share, and use custom AI code assistants with our open-source IDE extensions and hub of models, rules, prompts, docs, and other building blocks\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fvoideditor\u002Fvoid?style=social\" height=\"17\" align=\"texttop\"\u002F> [void](https:\u002F\u002Fgithub.com\u002Fvoideditor\u002Fvoid) - an open-source Cursor alternative, use AI agents on your codebase, checkpoint and visualize changes, and bring any model or host locally\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fblock\u002Fgoose?style=social\" height=\"17\" align=\"texttop\"\u002F> [goose](https:\u002F\u002Fgithub.com\u002Fblock\u002Fgoose) - an open-source, extensible AI agent that goes beyond code suggestions \n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRooCodeInc\u002FRoo-Code?style=social\" height=\"17\" align=\"texttop\"\u002F> [Roo-Code](https:\u002F\u002Fgithub.com\u002FRooCodeInc\u002FRoo-Code) - a whole dev team of AI agents in your code editor\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fcharmbracelet\u002Fcrush?style=social\" height=\"17\" align=\"texttop\"\u002F> [crush](https:\u002F\u002Fgithub.com\u002Fcharmbracelet\u002Fcrush) - the glamourous AI coding agent for your favourite terminal\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FKilo-Org\u002Fkilocode?style=social\" height=\"17\" align=\"texttop\"\u002F> [kilocode](https:\u002F\u002Fgithub.com\u002FKilo-Org\u002Fkilocode) - open source AI coding assistant for planning, building, and fixing code\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fhumanlayer\u002Fhumanlayer?style=social\" height=\"17\" align=\"texttop\"\u002F> [humanlayer](https:\u002F\u002Fgithub.com\u002Fhumanlayer\u002Fhumanlayer) - the best way to get AI coding agents to solve hard problems in complex codebases\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fcarlrobertoh\u002FProxyAI?style=social\" height=\"17\" align=\"texttop\"\u002F> [ProxyAI](https:\u002F\u002Fgithub.com\u002Fcarlrobertoh\u002FProxyAI) - the leading open-source AI copilot for JetBrains\n\n[Back to Table of Contents](#table-of-contents)\n\n### Computer Use\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOpenInterpreter\u002Fopen-interpreter?style=social\" height=\"17\" align=\"texttop\"\u002F> [open-interpreter](https:\u002F\u002Fgithub.com\u002FOpenInterpreter\u002Fopen-interpreter) - a natural language interface for computers\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002FOmniParser?style=social\" height=\"17\" align=\"texttop\"\u002F> [OmniParser](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FOmniParser) - a simple screen parsing tool towards pure vision based GUI agent\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ftrycua\u002Fcua?style=social\" height=\"17\" align=\"texttop\"\u002F> [cua](https:\u002F\u002Fgithub.com\u002Ftrycua\u002Fcua) - the Docker Container for Computer-Use AI Agents\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOthersideAI\u002Fself-operating-computer?style=social\" height=\"17\" align=\"texttop\"\u002F> [self-operating-computer](https:\u002F\u002Fgithub.com\u002FOthersideAI\u002Fself-operating-computer) - a framework to enable multimodal models to operate a computer\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsimular-ai\u002FAgent-S?style=social\" height=\"17\" align=\"texttop\"\u002F> [Agent-S](https:\u002F\u002Fgithub.com\u002Fsimular-ai\u002FAgent-S) - an open agentic framework that uses computers like a human\n\n[Back to Table of Contents](#table-of-contents)\n\n### Browser Automation\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpuppeteer\u002Fpuppeteer?style=social\" height=\"17\" align=\"texttop\"\u002F> [puppeteer](https:\u002F\u002Fgithub.com\u002Fpuppeteer\u002Fpuppeteer) - a JavaScript API for Chrome and Firefox\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002Fplaywright?style=social\" height=\"17\" align=\"texttop\"\u002F> [playwright](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fplaywright) - a framework for Web Testing and Automation\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fbrowser-use\u002Fbrowser-use?style=social\" height=\"17\" align=\"texttop\"\u002F> [browser-use](https:\u002F\u002Fgithub.com\u002Fbrowser-use\u002Fbrowser-use) - make websites accessible for AI agents\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmendableai\u002Ffirecrawl?style=social\" height=\"17\" align=\"texttop\"\u002F> [firecrawl](https:\u002F\u002Fgithub.com\u002Fmendableai\u002Ffirecrawl) - turn entire websites into LLM-ready markdown or structured data\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fbrowserbase\u002Fstagehand?style=social\" height=\"17\" align=\"texttop\"\u002F> [stagehand](https:\u002F\u002Fgithub.com\u002Fbrowserbase\u002Fstagehand) -  the AI Browser Automation Framework\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fnanobrowser\u002Fnanobrowser?style=social\" height=\"17\" align=\"texttop\"\u002F> [nanobrowser](https:\u002F\u002Fgithub.com\u002Fnanobrowser\u002Fnanobrowser) -  open-source Chrome extension for AI-powered web automation\n\n[Back to Table of Contents](#table-of-contents)\n\n### Memory Management\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmem0ai\u002Fmem0?style=social\" height=\"17\" align=\"texttop\"\u002F> [mem0](https:\u002F\u002Fgithub.com\u002Fmem0ai\u002Fmem0) - universal memory layer for AI Agents\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fletta-ai\u002Fletta?style=social\" height=\"17\" align=\"texttop\"\u002F> [letta](https:\u002F\u002Fgithub.com\u002Fletta-ai\u002Fletta) - the stateful agents framework with memory, reasoning, and context management\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsupermemoryai\u002Fsupermemory?style=social\" height=\"17\" align=\"texttop\"\u002F> [supermemory](https:\u002F\u002Fgithub.com\u002Fsupermemoryai\u002Fsupermemory) - memory engine and app that is extremely fast, scalable\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ftopoteretes\u002Fcognee?style=social\" height=\"17\" align=\"texttop\"\u002F> [cognee](https:\u002F\u002Fgithub.com\u002Ftopoteretes\u002Fcognee) - memory for AI Agents in 5 lines of code\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FLMCache\u002FLMCache?style=social\" height=\"17\" align=\"texttop\"\u002F> [LMCache](https:\u002F\u002Fgithub.com\u002FLMCache\u002FLMCache) - supercharge your LLM with the fastest KV Cache Layer\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNevaMind-AI\u002FmemU?style=social\" height=\"17\" align=\"texttop\"\u002F> [memU](https:\u002F\u002Fgithub.com\u002FNevaMind-AI\u002FmemU) - an open-source memory framework for AI companions\n\n[Back to Table of Contents](#table-of-contents)\n\n### Testing, Evaluation, and Observability\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flangfuse\u002Flangfuse?style=social\" height=\"17\" align=\"texttop\"\u002F> [langfuse](https:\u002F\u002Fgithub.com\u002Flangfuse\u002Flangfuse) - an open-source LLM engineering platform: LLM Observability, metrics, evals, prompt management, playground, datasets. Integrates with OpenTelemetry, Langchain, OpenAI SDK, LiteLLM, and more\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fcomet-ml\u002Fopik?style=social\" height=\"17\" align=\"texttop\"\u002F> [opik](https:\u002F\u002Fgithub.com\u002Fcomet-ml\u002Fopik) - debug, evaluate, and monitor your LLM applications, RAG systems, and agentic workflows with comprehensive tracing, automated evaluations, and production-ready dashboards\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ftraceloop\u002Fopenllmetry?style=social\" height=\"17\" align=\"texttop\"\u002F> [openllmetry](https:\u002F\u002Fgithub.com\u002Ftraceloop\u002Fopenllmetry) - an open-source observability for your LLM application, based on OpenTelemetry\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNVIDIA\u002Fgarak?style=social\" height=\"17\" align=\"texttop\"\u002F> [garak](https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fgarak) - the LLM vulnerability scanner from NVIDIA\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FGiskard-AI\u002Fgiskard?style=social\" height=\"17\" align=\"texttop\"\u002F> [giskard](https:\u002F\u002Fgithub.com\u002FGiskard-AI\u002Fgiskard) - an open-source evaluation & testing for AI & LLM systems\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAgenta-AI\u002Fagenta?style=social\" height=\"17\" align=\"texttop\"\u002F> [agenta](https:\u002F\u002Fgithub.com\u002FAgenta-AI\u002Fagenta) - an open-source LLMOps platform: prompt playground, prompt management, LLM evaluation, and LLM observability all in one place\n\n[Back to Table of Contents](#table-of-contents)\n\n### Research\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FItzCrazyKns\u002FPerplexica?style=social\" height=\"17\" align=\"texttop\"\u002F> [Perplexica](https:\u002F\u002Fgithub.com\u002FItzCrazyKns\u002FPerplexica) -  an open-source alternative to Perplexity AI, the AI-powered search engine\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fassafelovic\u002Fgpt-researcher?style=social\" height=\"17\" align=\"texttop\"\u002F> [gpt-researcher](https:\u002F\u002Fgithub.com\u002Fassafelovic\u002Fgpt-researcher) - an LLM based autonomous agent that conducts deep local and web research on any topic and generates a long report with citations\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMODSetter\u002FSurfSense?style=social\" height=\"17\" align=\"texttop\"\u002F> [SurfSense](https:\u002F\u002Fgithub.com\u002FMODSetter\u002FSurfSense) - an open-source alternative to NotebookLM \u002F Perplexity \u002F Glean\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flfnovo\u002Fopen-notebook?style=social\" height=\"17\" align=\"texttop\"\u002F> [open-notebook](https:\u002F\u002Fgithub.com\u002Flfnovo\u002Fopen-notebook) - an open-source implementation of Notebook LM with more flexibility and features\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002FRD-Agent?style=social\" height=\"17\" align=\"texttop\"\u002F> [RD-Agent](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FRD-Agent) - automate the most critical and valuable aspects of the industrial R&D process\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flangchain-ai\u002Flocal-deep-researcher?style=social\" height=\"17\" align=\"texttop\"\u002F> [local-deep-researcher](https:\u002F\u002Fgithub.com\u002Flangchain-ai\u002Flocal-deep-researcher) - fully local web research and report writing assistant\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FLearningCircuit\u002Flocal-deep-research?style=social\" height=\"17\" align=\"texttop\"\u002F> [local-deep-research](https:\u002F\u002Fgithub.com\u002FLearningCircuit\u002Flocal-deep-research) - an AI-powered research assistant for deep, iterative research\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmurtaza-nasir\u002Fmaestro?style=social\" height=\"17\" align=\"texttop\"\u002F> [maestro](https:\u002F\u002Fgithub.com\u002Fmurtaza-nasir\u002Fmaestro) - an AI-powered research application designed to streamline complex research tasks\n\n[Back to Table of Contents](#table-of-contents)\n\n### Training and Fine-tuning\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOpenRLHF\u002FOpenRLHF?style=social\" height=\"17\" align=\"texttop\"\u002F> [OpenRLHF](https:\u002F\u002Fgithub.com\u002FOpenRLHF\u002FOpenRLHF) - an easy-to-use, high-performance open-source RLHF framework built on Ray, vLLM, ZeRO-3 and HuggingFace Transformers, designed to make RLHF training simple and accessible\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fkiln-ai\u002Fkiln?style=social\" height=\"17\" align=\"texttop\"\u002F> [Kiln](https:\u002F\u002Fgithub.com\u002Fkiln-ai\u002Fkiln) - the easiest tool for fine-tuning LLM models, synthetic data generation, and collaborating on datasets\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fe-p-armstrong\u002Faugmentoolkit?style=social\" height=\"17\" align=\"texttop\"\u002F> [augmentoolkit](https:\u002F\u002Fgithub.com\u002Fe-p-armstrong\u002Faugmentoolkit) - train an open-source LLM on new facts\n\n[Back to Table of Contents](#table-of-contents)\n\n### Miscellaneous\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fupstash\u002Fcontext7?style=social\" height=\"17\" align=\"texttop\"\u002F> [context7](https:\u002F\u002Fgithub.com\u002Fupstash\u002Fcontext7) - up-to-date code documentation for LLMs and AI code editors\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Faliasrobotics\u002Fcai?style=social\" height=\"17\" align=\"texttop\"\u002F> [cai](https:\u002F\u002Fgithub.com\u002Faliasrobotics\u002Fcai) - Cybersecurity AI (CAI), the framework for AI Security\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmurtaza-nasir\u002Fspeakr?style=social\" height=\"17\" align=\"texttop\"\u002F> [speakr](https:\u002F\u002Fgithub.com\u002Fmurtaza-nasir\u002Fspeakr) - a personal, self-hosted web application designed for transcribing audio recordings\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpresenton\u002Fpresenton?style=social\" height=\"17\" align=\"texttop\"\u002F> [presenton](https:\u002F\u002Fgithub.com\u002Fpresenton\u002Fpresenton) - an open-source AI presentation generator and API\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FVectorSpaceLab\u002FOmniGen2?style=social\" height=\"17\" align=\"texttop\"\u002F> [OmniGen2](https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FOmniGen2) - exploration to advanced multimodal generation\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTheAhmadOsman\u002F4o-ghibli-at-home?style=social\" height=\"17\" align=\"texttop\"\u002F> [4o-ghibli-at-home](https:\u002F\u002Fgithub.com\u002FTheAhmadOsman\u002F4o-ghibli-at-home) - a powerful, self-hosted AI photo stylizer built for performance and privacy\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRoy3838\u002FObserver?style=social\" height=\"17\" align=\"texttop\"\u002F> [Observer](https:\u002F\u002Fgithub.com\u002FRoy3838\u002FObserver) - local open-source micro-agents that observe, log and react, all while keeping your data private and secure\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fminitap-ai\u002Fmobile-use?style=social\" height=\"17\" align=\"texttop\"\u002F> [mobile-use](https:\u002F\u002Fgithub.com\u002Fminitap-ai\u002Fmobile-use) - a powerful, open-source AI agent that controls your Android or IOS device using natural language\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fgabber-dev\u002Fgabber?style=social\" height=\"17\" align=\"texttop\"\u002F> [gabber](https:\u002F\u002Fgithub.com\u002Fgabber-dev\u002Fgabber) - build AI applications that can see, hear, and speak using your screens, microphones, and cameras as inputs\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsevenreasons\u002Fpromptcat?style=social\" height=\"17\" align=\"texttop\"\u002F> [promptcat](https:\u002F\u002Fgithub.com\u002Fsevenreasons\u002Fpromptcat) - a zero-dependency prompt manager\u002Fcatalog\u002Flibrary in a single HTML file\n\n[Back to Table of Contents](#table-of-contents)\n\n## Hardware\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fchannel\u002Fsubscribers\u002FUCajiMK_CY9icRhLepS8_3ug?style=social\" height=\"17\" align=\"texttop\"\u002F> [Alex Ziskind](https:\u002F\u002Fwww.youtube.com\u002F@AZisk) - tests of pcs, laptops, gpus etc. capable of running LLMs\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fchannel\u002Fsubscribers\u002FUCiaQzXI5528Il6r2NNkrkJA?style=social\" height=\"17\" align=\"texttop\"\u002F> [Digital Spaceport](https:\u002F\u002Fwww.youtube.com\u002F@DigitalSpaceport) - reviews of various builds designed for LLM inference\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fchannel\u002Fsubscribers\u002FUCQs0lwV6E4p7LQaGJ6fgy5Q?style=social\" height=\"17\" align=\"texttop\"\u002F> [JetsonHacks](https:\u002F\u002Fwww.youtube.com\u002F@JetsonHacks) - information about developing on NVIDIA Jetson Development Kits\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fchannel\u002Fsubscribers\u002FUC8h2Sf-yyo1WXeEUr-OHgyg?style=social\" height=\"17\" align=\"texttop\"\u002F> [Miyconst](https:\u002F\u002Fwww.youtube.com\u002F@Miyconst) - tests of various types of hardware capable of running LLMs\n- [Kolosal - LLM Memory calculator](https:\u002F\u002Fwww.kolosal.ai\u002Fmemory-calculator) - estimate the RAM requirements of any GGUF model instantly\n- [LLM Inference VRAM & GPU Requirement Calculator](https:\u002F\u002Fapp.linpp2009.com\u002Fen\u002Fllm-gpu-memory-calculator) - calculate how many GPUs you need to deploy LLMs\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fvosen\u002FZLUDA?style=social\" height=\"17\" align=\"texttop\"\u002F> [ZLUDA](https:\u002F\u002Fgithub.com\u002Fvosen\u002FZLUDA) - CUDA on non-NVIDIA GPUs\n\n[Back to Table of Contents](#table-of-contents)\n\n## Tutorials\n\n### Models\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002Fl8pRSuU81PU?style=social\" height=\"17\" align=\"texttop\"\u002F> [Let's reproduce GPT-2 (124M)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=l8pRSuU81PU)\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fkarpathy\u002Fnanochat?style=social\" height=\"17\" align=\"texttop\"\u002F> [nanochat](https:\u002F\u002Fgithub.com\u002Fkarpathy\u002Fnanochat) - a full-stack implementation of an LLM like ChatGPT in a single, clean, minimal, hackable, dependency-lite codebase, designed to run on a single 8XH100 node via scripts like speedrun.sh, that run the entire pipeline start to end\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FjrJKRYAdh7I?style=social\" height=\"17\" align=\"texttop\"\u002F> [Knowledge Distillation: How LLMs train each other](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=jrJKRYAdh7I)\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fiuliaturc\u002Fgguf-docs?style=social\" height=\"17\" align=\"texttop\"\u002F> [gguf-docs](https:\u002F\u002Fgithub.com\u002Fiuliaturc\u002Fgguf-docs) - Docs for GGUF quantization (unofficial)\n\n[Back to Table of Contents](#table-of-contents)\n\n### Prompt Engineering\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdair-ai\u002FPrompt-Engineering-Guide?style=social\" height=\"17\" align=\"texttop\"\u002F> [Prompt Engineering Guide](https:\u002F\u002Fgithub.com\u002Fdair-ai\u002FPrompt-Engineering-Guide) - guides, papers, lecture, notebooks and resources for prompt engineering\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNirDiamant\u002FPrompt_Engineering?style=social\" height=\"17\" align=\"texttop\"\u002F> [Prompt Engineering by NirDiamant](https:\u002F\u002Fgithub.com\u002FNirDiamant\u002FPrompt_Engineering) - a comprehensive collection of tutorials and implementations for Prompt Engineering techniques, ranging from fundamental concepts to advanced strategies\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FGoogle-%234285F4?logo=google&logoColor=red\" height=\"17\" align=\"texttop\"\u002F> [Prompting guide 101](https:\u002F\u002Fservices.google.com\u002Ffh\u002Ffiles\u002Fmisc\u002Fgemini-for-google-workspace-prompting-guide-101.pdf) - a quick-start handbook for effective prompts by Google\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FGoogle-%234285F4?logo=google&logoColor=red\" height=\"17\" align=\"texttop\"\u002F> [Prompt Engineering by Google](https:\u002F\u002Fdrive.google.com\u002Ffile\u002Fd\u002F1AbaBYbEa_EbPelsT40-vj64L-2IwUJHy\u002Fview) - prompt engineering by Google\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FAnthropic-%23191919?logo=anthropic\" height=\"17\" align=\"texttop\"\u002F> [Prompt Engineering by Anthropic](https:\u002F\u002Fdocs.anthropic.com\u002Fen\u002Fdocs\u002Fbuild-with-claude\u002Fprompt-engineering\u002Foverview) - prompt engineering by Anthropic\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FAnthropic-%23191919?logo=anthropic\" height=\"17\" align=\"texttop\"\u002F> [Prompt Engineering Interactive Tutorial](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fcourses\u002Fblob\u002Fmaster\u002Fprompt_engineering_interactive_tutorial\u002FREADME.md) - Prompt Engineering Interactive Tutorial by Anthropic\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FAnthropic-%23191919?logo=anthropic\" height=\"17\" align=\"texttop\"\u002F> [Real world prompting](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fcourses\u002Fblob\u002Fmaster\u002Freal_world_prompting\u002FREADME.md) - real world prompting tutorial by Anthropic\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FAnthropic-%23191919?logo=anthropic\" height=\"17\" align=\"texttop\"\u002F> [Prompt evaluations](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fcourses\u002Fblob\u002Fmaster\u002Fprompt_evaluations\u002FREADME.md) - prompt evaluations course by Anthropic\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fx1xhlol\u002Fsystem-prompts-and-models-of-ai-tools?style=social\" height=\"17\" align=\"texttop\"\u002F> [system-prompts-and-models-of-ai-tools](https:\u002F\u002Fgithub.com\u002Fx1xhlol\u002Fsystem-prompts-and-models-of-ai-tools) - a collection of system prompts extracted from AI tools\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fasgeirtj\u002Fsystem_prompts_leaks?style=social\" height=\"17\" align=\"texttop\"\u002F> [system_prompts_leaks](https:\u002F\u002Fgithub.com\u002Fasgeirtj\u002Fsystem_prompts_leaks) - a collection of extracted System Prompts from popular chatbots like ChatGPT, Claude & Gemini\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FOpenAI-%23412991?logo=openai\" height=\"17\" align=\"texttop\"\u002F> \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fopenai\u002Fcodex?style=social\" height=\"17\" align=\"texttop\"\u002F> [Prompt from Codex](https:\u002F\u002Fgithub.com\u002Fopenai\u002Fcodex\u002Fblob\u002Fmain\u002Fcodex-rs\u002Fcore\u002Fprompt.md) - Prompt used to steer behavior of OpenAI's Codex\n\n[Back to Table of Contents](#table-of-contents)\n\n### Context Engineering\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdavidkimai\u002FContext-Engineering?style=social\" height=\"17\" align=\"texttop\"\u002F> [Context-Engineering](https:\u002F\u002Fgithub.com\u002Fdavidkimai\u002FContext-Engineering) - a frontier, first-principles handbook inspired by Karpathy and 3Blue1Brown for moving beyond prompt engineering to the wider discipline of context design, orchestration, and optimization\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMeirtz\u002FAwesome-Context-Engineering?style=social\" height=\"17\" align=\"texttop\"\u002F> [Awesome-Context-Engineering](https:\u002F\u002Fgithub.com\u002FMeirtz\u002FAwesome-Context-Engineering) - a comprehensive survey on Context Engineering: from prompt engineering to production-grade AI systems\n\n[Back to Table of Contents](#table-of-contents)\n\n### Inference\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fvllm-project\u002Fproduction-stack?style=social\" height=\"17\" align=\"texttop\"\u002F> [vLLM Production Stack](https:\u002F\u002Fgithub.com\u002Fvllm-project\u002Fproduction-stack) - vLLM’s reference system for K8S-native cluster-wide deployment with community-driven performance optimization\n\n[Back to Table of Contents](#table-of-contents)\n\n### Agents\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNirDiamant\u002FGenAI_Agents?style=social\" height=\"17\" align=\"texttop\"\u002F> [GenAI Agents](https:\u002F\u002Fgithub.com\u002FNirDiamant\u002FGenAI_Agents) - tutorials and implementations for various Generative AI Agent techniques\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fashishpatel26\u002F500-AI-Agents-Projects?style=social\" height=\"17\" align=\"texttop\"\u002F> [500+ AI Agent Projects](https:\u002F\u002Fgithub.com\u002Fashishpatel26\u002F500-AI-Agents-Projects) - a curated collection of AI agent use cases across various industries\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fhumanlayer\u002F12-factor-agents?style=social\" height=\"17\" align=\"texttop\"\u002F> [12-Factor Agents](https:\u002F\u002Fgithub.com\u002Fhumanlayer\u002F12-factor-agents) - principles for building reliable LLM applications\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNirDiamant\u002Fagents-towards-production?style=social\" height=\"17\" align=\"texttop\"\u002F> [Agents towards production](https:\u002F\u002Fgithub.com\u002FNirDiamant\u002Fagents-towards-production) - end-to-end, code-first tutorials covering every layer of production-grade GenAI agents, guiding you from spark to scale with proven patterns and reusable blueprints for real-world launches\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Foxbshw\u002FLLM-Agents-Ecosystem-Handbook?style=social\" height=\"17\" align=\"texttop\"\u002F> [LLM Agents & Ecosystem Handbook](https:\u002F\u002Fgithub.com\u002Foxbshw\u002FLLM-Agents-Ecosystem-Handbook) - one-stop handbook for building, deploying, and understanding LLM agents with 60+ skeletons, tutorials, ecosystem guides, and evaluation tools\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FGoogle-%234285F4?logo=google&logoColor=red\" height=\"17\" align=\"texttop\"\u002F> [601 real-world gen AI use cases](https:\u002F\u002Fcloud.google.com\u002Ftransform\u002F101-real-world-generative-ai-use-cases-from-industry-leaders) - 601 real-world gen AI use cases from the world's leading organizations by Google\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FOpenAI-%23412991?logo=openai\" height=\"17\" align=\"texttop\"\u002F> [A practical guide to building agents](https:\u002F\u002Fcdn.openai.com\u002Fbusiness-guides-and-resources\u002Fa-practical-guide-to-building-agents.pdf) - a practical guide to building agents by OpenAI\n\n[Back to Table of Contents](#table-of-contents)\n\n### Retrieval-Augmented Generation\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpathwaycom\u002Fllm-app?style=social\" height=\"17\" align=\"texttop\"\u002F> [Pathway AI Pipelines](https:\u002F\u002Fgithub.com\u002Fpathwaycom\u002Fllm-app) - ready-to-run cloud templates for RAG, AI pipelines, and enterprise search with live data\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNirDiamant\u002FRAG_Techniques?style=social\" height=\"17\" align=\"texttop\"\u002F> [RAG Techniques](https:\u002F\u002Fgithub.com\u002FNirDiamant\u002FRAG_Techniques) - various advanced techniques for Retrieval-Augmented Generation (RAG) systems\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNirDiamant\u002FControllable-RAG-Agent?style=social\" height=\"17\" align=\"texttop\"\u002F> [Controllable RAG Agent](https:\u002F\u002Fgithub.com\u002FNirDiamant\u002FControllable-RAG-Agent) - an advanced Retrieval-Augmented Generation (RAG) solution for complex question answering that uses sophisticated graph based algorithm to handle the tasks\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flokeswaran-aj\u002Flangchain-rag-cookbook?style=social\" height=\"17\" align=\"texttop\"\u002F> [LangChain RAG Cookbook](https:\u002F\u002Fgithub.com\u002Flokeswaran-aj\u002Flangchain-rag-cookbook) - a collection of modular RAG techniques, implemented in LangChain + Python\n\n[Back to Table of Contents](#table-of-contents)\n\n### Miscellaneous\n\n- [Self-hosted AI coding that just works](https:\u002F\u002Fwww.reddit.com\u002Fr\u002FLocalLLaMA\u002Fcomments\u002F1lt4y1z\u002Fselfhosted_ai_coding_that_just_works\u002F)\n\n[Back to Table of Contents](#table-of-contents)\n\n","# LLMs-本地\n用于在本地运行大型语言模型的优秀平台、工具和资源列表\n\n \n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002F0xSojalSec_LLMs-local_readme_81d345b929d7.jpeg\">\n\n## 目录\n\n- [推理平台](#inference-platforms)\n- [推理引擎](#inference-engines)\n- [用户界面](#user-interfaces)\n- [大型语言模型](#large-language-models)\n  - [探索工具、基准测试、排行榜](#explorers-benchmarks-leaderboards)\n  - [模型提供商](#model-providers)\n  - [特定模型](#specific-models)\n    - [通用模型](#general-purpose)\n    - [编程模型](#coding)\n    - [多模态模型](#multimodal)\n    - [图像模型](#image)\n    - [音频模型](#audio)\n    - [其他](#miscellaneous)\n- [工具](#tools)\n  - [模型](#models)\n  - [智能体框架](#agent-frameworks)\n  - [模型上下文协议](#model-context-protocol)\n  - [检索增强生成](#retrieval-augmented-generation)\n  - [编程智能体](#coding-agents)\n  - [计算机使用](#computer-use)\n  - [浏览器自动化](#browser-automation)\n  - [内存管理](#memory-management)\n  - [测试、评估与可观测性](#testing-evaluation-and-observability)\n  - [研究](#research)\n  - [训练与微调](#training-and-fine-tuning)\n  - [其他](#miscellaneous-1)\n- [硬件](#hardware)\n- [教程](#tutorials)\n  - [模型](#models-1)\n  - [提示工程](#prompt-engineering)\n  - [上下文工程](#context-engineering)\n  - [推理](#inference)\n  - [智能体](#agents)\n  - [检索增强生成](#retrieval-augmented-generation-1)\n  - [其他](#miscellaneous-2)\n- [社区](#communities)\n\n## 推理平台\n\n- [LM Studio](https:\u002F\u002Flmstudio.ai\u002F) - 发现、下载并运行本地大型语言模型\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmenloresearch\u002Fjan?style=social\" height=\"17\" align=\"texttop\"\u002F> [jan](https:\u002F\u002Fgithub.com\u002Fmenloresearch\u002Fjan) - 一款开源的ChatGPT替代品，可在您的电脑上100%离线运行\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmudler\u002FLocalAI?style=social\" height=\"17\" align=\"texttop\"\u002F> [LocalAI](https:\u002F\u002Fgithub.com\u002Fmudler\u002FLocalAI) - OpenAI、Claude等服务的免费开源替代方案\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FChatBoxAI\u002FChatBox?style=social\" height=\"17\" align=\"texttop\"\u002F> [ChatBox](https:\u002F\u002Fgithub.com\u002FChatBoxAI\u002FChatBox) - 一款用户友好的桌面客户端应用，用于运行AI模型\u002F大型语言模型\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flemonade-sdk\u002Flemonade?style=social\" height=\"17\" align=\"texttop\"\u002F> [lemonade](https:\u002F\u002Fgithub.com\u002Flemonade-sdk\u002Flemonade) - 一个支持GPU和NPU加速的本地大型语言模型服务器\n\n[返回目录](#table-of-contents)\n\n## 推理引擎\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Follama\u002Follama?style=social\" height=\"17\" align=\"texttop\"\u002F> [ollama](https:\u002F\u002Fgithub.com\u002Follama\u002Follama) - 快速启动并运行大型语言模型\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fggml-org\u002Fllama.cpp?style=social\" height=\"17\" align=\"texttop\"\u002F> [llama.cpp](https:\u002F\u002Fgithub.com\u002Fggml-org\u002Fllama.cpp) - 使用C\u002FC++进行大型语言模型推理\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fvllm-project\u002Fvllm?style=social\" height=\"17\" align=\"texttop\"\u002F> [vllm](https:\u002F\u002Fgithub.com\u002Fvllm-project\u002Fvllm) - 高吞吐量且内存高效的大型语言模型推理与服务引擎\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fexo-explore\u002Fexo?style=social\" height=\"17\" align=\"texttop\"\u002F> [exo](https:\u002F\u002Fgithub.com\u002Fexo-explore\u002Fexo) - 使用日常设备在家搭建自己的AI集群\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002FBitNet?style=social\" height=\"17\" align=\"texttop\"\u002F> [BitNet](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FBitNet) - 1位大型语言模型的官方推理框架\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsgl-project\u002Fsglang?style=social\" height=\"17\" align=\"texttop\"\u002F> [sglang](https:\u002F\u002Fgithub.com\u002Fsgl-project\u002Fsglang) - 一种用于大型语言模型和视觉语言模型的快速服务框架\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FGeeeekExplorer\u002Fnano-vllm?style=social\" height=\"17\" align=\"texttop\"\u002F> [Nano-vLLM](https:\u002F\u002Fgithub.com\u002FGeeeekExplorer\u002Fnano-vllm) - 从头开始构建的轻量级vLLM实现\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FLostRuins\u002Fkoboldcpp?style=social\" height=\"17\" align=\"texttop\"\u002F> [koboldcpp](https:\u002F\u002Fgithub.com\u002FLostRuins\u002Fkoboldcpp) - 轻松运行GGUF模型，并配备KoboldAI界面\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fgpustack\u002Fgpustack?style=social\" height=\"17\" align=\"texttop\"\u002F> [gpustack](https:\u002F\u002Fgithub.com\u002Fgpustack\u002Fgpustack) - 在GPU集群上简单、可扩展地部署AI模型\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fml-explore\u002Fmlx-lm?style=social\" height=\"17\" align=\"texttop\"\u002F> [mlx-lm](https:\u002F\u002Fgithub.com\u002Fml-explore\u002Fmlx-lm) - 使用MLX在Apple芯片上生成文本并微调大型语言模型\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fb4rtaz\u002Fdistributed-llama?style=social\" height=\"17\" align=\"texttop\"\u002F> [distributed-llama](https:\u002F\u002Fgithub.com\u002Fb4rtaz\u002Fdistributed-llama) - 将家用设备连接成强大的集群，以加速大型语言模型的推理\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fikawrakow\u002Fik_llama.cpp?style=social\" height=\"17\" align=\"texttop\"\u002F> [ik_llama.cpp](https:\u002F\u002Fgithub.com\u002Fikawrakow\u002Fik_llama.cpp) - llama.cpp的分支，增加了最新的量化技术并提升了性能\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FFastFlowLM\u002FFastFlowLM?style=social\" height=\"17\" align=\"texttop\"\u002F> [FastFlowLM](https:\u002F\u002Fgithub.com\u002FFastFlowLM\u002FFastFlowLM) - 在AMD Ryzen™ AI NPU上运行大型语言模型\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fnlzy\u002Fvllm-gfx906?style=social\" height=\"17\" align=\"texttop\"\u002F> [vllm-gfx906](https:\u002F\u002Fgithub.com\u002Fnlzy\u002Fvllm-gfx906) - 适用于AMD gfx906 GPU（如Radeon VII \u002F MI50 \u002F MI60）的vLLM\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fintel\u002Fllm-scaler?style=social\" height=\"17\" align=\"texttop\"\u002F> [llm-scaler](https:\u002F\u002Fgithub.com\u002Fintel\u002Fllm-scaler) - 在Intel Arc™ Pro B60 GPU上运行大型语言模型\n\n[返回目录](#table-of-contents)\n\n## 用户界面\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fopen-webui\u002Fopen-webui?style=social\" height=\"17\" align=\"texttop\"\u002F> [Open WebUI](https:\u002F\u002Fgithub.com\u002Fopen-webui\u002Fopen-webui) - 友好易用的AI界面（支持Ollama、OpenAI API等）\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flobehub\u002Flobe-chat?style=social\" height=\"17\" align=\"texttop\"\u002F> [Lobe Chat](https:\u002F\u002Fgithub.com\u002Flobehub\u002Flobe-chat) - 一个开源、现代设计的AI聊天框架\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Foobabooga\u002Ftext-generation-webui?style=social\" height=\"17\" align=\"texttop\"\u002F> [文本生成Web UI](https:\u002F\u002Fgithub.com\u002Foobabooga\u002Ftext-generation-webui) - 具有高级功能、易于部署且支持多种后端的LLM用户界面\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FSillyTavern\u002FSillyTavern?style=social\" height=\"17\" align=\"texttop\"\u002F> [SillyTavern](https:\u002F\u002Fgithub.com\u002FSillyTavern\u002FSillyTavern) - 面向高级用户的LLM前端\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fn4ze3m\u002Fpage-assist?style=social\" height=\"17\" align=\"texttop\"\u002F> [Page Assist](https:\u002F\u002Fgithub.com\u002Fn4ze3m\u002Fpage-assist) - 使用您本地运行的AI模型来辅助您的网页浏览\n\n[返回目录](#table-of-contents)\n\n## 大型语言模型\n\n### 探索工具、基准测试与排行榜\n\n- [AI模型与API提供商分析](https:\u002F\u002Fartificialanalysis.ai\u002F) - 帮助您了解AI领域，从而为您的应用场景选择最佳模型和提供商\n- [LLM Explorer](https:\u002F\u002Fllm-explorer.com\u002F) - 探索开源LLM模型列表\n- [Dubesor LLM基准测试表](https:\u002F\u002Fdubesor.de\u002Fbenchtable) - 小规模手动性能对比基准\n- [oobabooga基准测试](https:\u002F\u002Foobabooga.github.io\u002Fbenchmark.html) - 按照每个分数的磁盘大小排序的列表\n\n[返回目录](#table-of-contents)\n\n### 模型提供商\n\n- [Qwen](https:\u002F\u002Fhuggingface.co\u002FQwen) - 由阿里云提供支持\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FMistral%20AI-%23FA520F?logo=mistralai&logoColor=%23FFFFFF\" height=\"17\" align=\"texttop\"\u002F> [Mistral AI](https:\u002F\u002Fhuggingface.co\u002Fmistralai) - 一家领先的法国人工智能初创公司\n- [腾讯](https:\u002F\u002Fhuggingface.co\u002Ftencent) - 中国一家跨国科技集团及控股公司的简介\n- [Unsloth AI](https:\u002F\u002Fhuggingface.co\u002Funsloth) - 致力于让每个人都能更方便地使用AI（如GGUF等）\n- [bartowski](https:\u002F\u002Fhuggingface.co\u002Fbartowski) - 提供流行LLM的GGUF版本\n- [北京人工智能研究院](https:\u002F\u002Fhuggingface.co\u002FBAAI) - 一家从事AI研究与开发的非营利性私人机构\n- [Open Thoughts](https:\u002F\u002Fhuggingface.co\u002Fopen-thoughts) - 一个由研究人员和工程师组成的团队，致力于整理最佳的开放推理数据集\n\n[返回目录](#table-of-contents)\n\n### 具体模型\n\n#### 通用型\n\n- [Qwen3-Next](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002FQwen\u002Fqwen3-next-68c25fd6838e585db8eeea9d) - 最新一代Qwen LLM系列\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FGoogle-%234285F4?logo=google&logoColor=red\" height=\"17\" align=\"texttop\"\u002F> [Gemma 3](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fgoogle\u002Fgemma-3-release-67c6c6f89c4f76621268bb6d) - 来自谷歌的一系列轻量级、最先进的开源模型，基于与Gemini模型相同的科研和技术打造\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FOpenAI-%23412991?logo=openai\" height=\"17\" align=\"texttop\"\u002F> [gpt-oss](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fopenai\u002Fgpt-oss-68911959590a1634ba11c7a4) - OpenAI提供的开放权重模型集合，专为强大的推理能力、智能体任务以及多用途的开发者场景而设计\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FMistral%20AI-%23FA520F?logo=mistralai&logoColor=%23FFFFFF\" height=\"17\" align=\"texttop\"\u002F> [Ministral 3](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fmistralai\u002Fministral-3) - 边缘模型系列，包含基础版、指令版和推理版，共有3种尺寸：3B、8B和14B，均具备视觉能力\n- [GLM-4.5](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fzai-org\u002Fglm-45-687c621d34bda8c9e4bf503b) - 专为智能代理设计的混合推理模型系列\n- [Hunyuan](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Ftencent\u002Fhunyuan-dense-model-6890632cda26b19119c9c5e7) - 腾讯开源的高效LLM系列，专为在各种计算环境中灵活部署而设计\n- [Phi-4-mini-instruct](https:\u002F\u002Fhuggingface.co\u002Fmicrosoft\u002FPhi-4-mini-instruct) - 基于合成数据和筛选后的公开网站构建的轻量级开放模型\n- [NVIDIA Nemotron v3](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fnvidia\u002Fnvidia-nemotron-v3) - NVIDIA推出的一系列开放模型，拥有开放的权重、训练数据和配方，能够以领先的效率和准确度构建专业化的AI智能体\n- [Llama Nemotron](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fnvidia\u002Fllama-nemotron-67d92346030a2691293f200b) - NVIDIA推出的可用于生产的开放企业级模型系列\n- [OpenReasoning-Nemotron](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fnvidia\u002Fopenreasoning-nemotron-687730dae0170059860f1f01) - NVIDIA推出的一系列模型，基于500万条数学、代码和科学相关的推理轨迹进行训练\n- [Granite 4.0](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fibm-granite\u002Fgranite-40-language-models-6811a18b820ef362d9e5a82c) - IBM推出的一系列轻量级、最先进的开源基础模型，原生支持多语言能力、广泛的编码任务（包括中间填充式代码补全）、检索增强生成（RAG）、工具使用以及结构化JSON输出\n- [EXAONE-4.0](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002FLGAI-EXAONE\u002Fexaone-40-686b2e0069800c835ed48375) - LG AI Research推出的LLM系列，融合了非推理模式和推理模式\n- [ERNIE 4.5](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fbaidu\u002Fernie-45-6861cd4c9be84540645f35c9) - 百度推出的大规模多模态模型系列\n- [Seed-OSS](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002FByteDance-Seed\u002Fseed-oss-68a609f4201e788db05b5dcd) - 字节跳动Seed团队开发的LLM系列，专为强大的长上下文理解、推理能力、智能体功能以及通用能力而设计，并具备多样的开发者友好特性\n\n[返回目录](#table-of-contents)\n\n#### 编码\n\n- [Qwen3-Coder](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002FQwen\u002Fqwen3-coder-687fc861e53c939e52d52d10) - Qwen系列迄今为止最具代理能力的代码模型合集\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FMistral%20AI-%23FA520F?logo=mistralai&logoColor=%23FFFFFF\" height=\"17\" align=\"texttop\"\u002F> [Devstral 2](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fmistralai\u002Fdevstral-2) - 两款用于软件工程任务的代理型LLM，擅长利用工具探索代码库、编辑多个文件，并为SWE智能体提供支持\n- [Mellum-4b-base](https:\u002F\u002Fhuggingface.co\u002FJetBrains\u002FMellum-4b-base) - 来自JetBrains的LLM，专为代码相关任务优化\n- [OlympicCoder-32B](https:\u002F\u002Fhuggingface.co\u002Fopen-r1\u002FOlympicCoder-32B) - 在LiveCodeBench和2024年国际信息学奥林匹克竞赛等竞技编程基准测试中表现极为出色的代码模型\n- [NextCoder](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fmicrosoft\u002Fnextcoder-6815ee6bfcf4e42f20d45028) - 以Qwen2.5-Coder Instruct变体为基础开发的一系列代码编辑LLM\n\n[返回目录](#table-of-contents)\n\n#### 多模态\n\n- [Qwen3-Omni](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002FQwen\u002Fqwen3-omni-68d100a86cd0906843ceccbe) - Qwen原生端到端多语言全模态基础模型合集\n\n[返回目录](#table-of-contents)\n\n#### 图像\n\n- [Qwen-Image](https:\u002F\u002Fhuggingface.co\u002FQwen\u002FQwen-Image) - Qwen系列中的图像生成基础模型，在复杂文本渲染和精准图像编辑方面取得了显著进展\n- [Qwen-Image-Edit-2509](https:\u002F\u002Fhuggingface.co\u002FQwen\u002FQwen-Image-Edit-2509) - Qwen-Image的图像编辑版本，将基础模型独特的文本渲染能力扩展至图像编辑任务，实现精确的文本编辑\n- [Qwen3-VL](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002FQwen\u002Fqwen3-vl-68d2a7c1b8a8afce4ebd2dbe) - Qwen系列迄今为止最强大的视觉-语言模型合集\n- [GLM-4.5V](https:\u002F\u002Fhuggingface.co\u002Fzai-org\u002FGLM-4.5V) - 基于智谱AI新一代旗舰文本基础模型GLM-4.5-Air的VLLM\n- [HunyuanImage-2.1](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHunyuanImage-2.1) - 高效的扩散模型，用于高分辨率（2K）文生图​ \n- [FastVLM](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fapple\u002Ffastvlm-68ac97b9cd5cacefdd04872e) - 苹果公司高效视觉编码的VLM合集\n- [MiniCPM-V-4_5](https:\u002F\u002Fhuggingface.co\u002Fopenbmb\u002FMiniCPM-V-4_5) - GPT-4o级别的小型MLLM，可在手机上实现单张图片、多张图片及高帧率视频的理解\n- [LFM2-VL](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002FLiquidAI\u002Flfm2-vl-68963bbc84a610f7638d5ffa) - 专为设备端部署设计的视觉-语言模型合集\n- [ClipTagger-12b](https:\u002F\u002Fhuggingface.co\u002Finference-net\u002FClipTagger-12b) - 一款专为大规模视频理解设计的视觉-语言模型（VLM）\n\n[返回目录](#table-of-contents)\n\n#### 音频\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FMistral%20AI-%23FA520F?logo=mistralai&logoColor=%23FFFFFF\" height=\"17\" align=\"texttop\"\u002F> [Voxtral-Small-24B-2507](https:\u002F\u002Fhuggingface.co\u002Fmistralai\u002FVoxtral-Small-24B-2507) - Mistral Small 3的升级版，融合了最先进的音频输入能力，同时保持一流的文本性能\n- [chatterbox](https:\u002F\u002Fhuggingface.co\u002FResembleAI\u002Fchatterbox) - 首个生产级开源TTS模型\n- [VibeVoice](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fmicrosoft\u002Fvibevoice-68a2ef24a875c44be47b034f) - 微软前沿文本转语音模型合集\n- [canary-1b-v2](https:\u002F\u002Fhuggingface.co\u002Fnvidia\u002Fcanary-1b-v2) - 英伟达的一款多任务语音转录与翻译模型\n- [parakeet-tdt-0.6b-v3](https:\u002F\u002Fhuggingface.co\u002Fnvidia\u002Fparakeet-tdt-0.6b-v3) - 英伟达的多语言语音转文字模型\n- [Kitten TTS](https:\u002F\u002Fhuggingface.co\u002FKittenML\u002Fmodels) - 开源逼真文本转语音模型合集，专为轻量级部署和高质量语音合成设计\n\n[返回目录](#table-of-contents)\n\n#### 其他\n\n- [Jan-v1-4B](https:\u002F\u002Fhuggingface.co\u002Fjanhq\u002FJan-v1-4B) - Jan家族的首个发布版本，专为Jan App中的代理式推理和问题解决而设计\n- [Jan-nano](https:\u002F\u002Fhuggingface.co\u002FMenlo\u002FJan-nano) - 一款紧凑的40亿参数语言模型，专门针对深度研究任务进行设计和训练\n- [Jan-nano-128k](https:\u002F\u002Fhuggingface.co\u002FMenlo\u002FJan-nano-128k) - Jan-nano的增强版，具备原生128k上下文窗口，可在不降低性能的情况下实现更深入、更全面的研究能力\n- [Arch-Router-1.5B](https:\u002F\u002Fhuggingface.co\u002Fkatanemo\u002FArch-Router-1.5B) - 最快的LLM路由模型，可根据用户的主观偏好进行匹配\n- [gpt-oss-safeguard](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fopenai\u002Fgpt-oss-safeguard) - 基于gpt-oss构建的安全推理模型合集\n- [Qwen3Guard](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002FQwen\u002Fqwen3guard-68d2729abbfae4716f3343a1) - 基于Qwen3构建的安全审核模型合集\n- [HunyuanWorld-1](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHunyuanWorld-1) - 开源3D世界生成模型\n- [Hunyuan-GameCraft-1.0](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHunyuan-GameCraft-1.0) - 一种用于游戏环境中高动态交互式视频生成的新框架\n\n[返回目录](#table-of-contents)\n\n\n\n## 工具\n\n### 模型\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Funslothai\u002Funsloth?style=social\" height=\"17\" align=\"texttop\"\u002F> [unsloth](https:\u002F\u002Fgithub.com\u002Funslothai\u002Funsloth) - LLM的微调与强化学习工具\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdottxt-ai\u002Foutlines?style=social\" height=\"17\" align=\"texttop\"\u002F> [outlines](https:\u002F\u002Fgithub.com\u002Fdottxt-ai\u002Foutlines) - LLM的结构化输出工具\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fp-e-w\u002Fheretic?style=social\" height=\"17\" align=\"texttop\"\u002F> [heretic](https:\u002F\u002Fgithub.com\u002Fp-e-w\u002Fheretic) - 语言模型的全自动内容审查移除工具\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmostlygeek\u002Fllama-swap?style=social\" height=\"17\" align=\"texttop\"\u002F> [llama-swap](https:\u002F\u002Fgithub.com\u002Fmostlygeek\u002Fllama-swap) - 适用于任何本地OpenAI兼容服务器（如llama.cpp、vllm等）的可靠模型切换工具\n\n[返回目录](#table-of-contents)\n\n### 代理框架\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FSignificant-Gravitas\u002FAutoGPT?style=social\" height=\"17\" align=\"texttop\"\u002F> [AutoGPT](https:\u002F\u002Fgithub.com\u002FSignificant-Gravitas\u002FAutoGPT) - 一个强大的平台，允许你创建、部署和管理持续运行的AI代理，以自动化复杂的流程\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flangflow-ai\u002Flangflow?style=social\" height=\"17\" align=\"texttop\"\u002F> [langflow](https:\u002F\u002Fgithub.com\u002Flangflow-ai\u002Flangflow) - 一个功能强大的工具，用于构建和部署基于AI的代理及工作流\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flangchain-ai\u002Flangchain?style=social\" height=\"17\" align=\"texttop\"\u002F> [langchain](https:\u002F\u002Fgithub.com\u002Flangchain-ai\u002Flangchain) - 构建上下文感知的推理应用\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002Fautogen?style=social\" height=\"17\" align=\"texttop\"\u002F> [autogen](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fautogen) - 一个用于智能体式AI的编程框架\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMintplex-Labs\u002Fanything-llm?style=social\" height=\"17\" align=\"texttop\"\u002F> [anything-llm](https:\u002F\u002Fgithub.com\u002FMintplex-Labs\u002Fanything-llm) - 一款一体化的桌面及Docker AI应用，内置RAG、AI代理、无代码代理构建器、MCP兼容性等功能\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FFlowiseAI\u002FFlowise?style=social\" height=\"17\" align=\"texttop\"\u002F> [Flowise](https:\u002F\u002Fgithub.com\u002FFlowiseAI\u002FFlowise) - 可视化方式构建AI代理\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Frun-llama\u002Fllama_index?style=social\" height=\"17\" align=\"texttop\"\u002F> [llama_index](https:\u002F\u002Fgithub.com\u002Frun-llama\u002Fllama_index) - 领先的数据驱动LLM代理构建框架\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FcrewAIInc\u002FcrewAI?style=social\" height=\"17\" align=\"texttop\"\u002F> [crewAI](https:\u002F\u002Fgithub.com\u002FcrewAIInc\u002FcrewAI) - 一个用于编排角色扮演型自主AI代理的框架\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fagno-agi\u002Fagno?style=social\" height=\"17\" align=\"texttop\"\u002F> [agno](https:\u002F\u002Fgithub.com\u002Fagno-agi\u002Fagno) - 一个全栈框架，用于构建具备记忆、知识与推理能力的多智能体系统\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsimstudioai\u002Fsim?style=social\" height=\"17\" align=\"texttop\"\u002F> [sim](https:\u002F\u002Fgithub.com\u002Fsimstudioai\u002Fsim) - 开源平台，用于构建和部署AI代理工作流\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FOpenAI-%23412991?logo=openai\" height=\"17\" align=\"texttop\"\u002F> \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fopenai\u002Fopenai-agents-python?style=social\" height=\"17\" align=\"texttop\"\u002F> [openai-agents-python](https:\u002F\u002Fgithub.com\u002Fopenai\u002Fopenai-agents-python) - 一个轻量级但功能强大的多智能体工作流框架\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTransformerOptimus\u002FSuperAGI?style=social\" height=\"17\" align=\"texttop\"\u002F> [SuperAGI](https:\u002F\u002Fgithub.com\u002FTransformerOptimus\u002FSuperAGI) - 一个开源框架，用于构建、管理和运行实用的自主AI代理\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fcamel-ai\u002Fcamel?style=social\" height=\"17\" align=\"texttop\"\u002F> [camel](https:\u002F\u002Fgithub.com\u002Fcamel-ai\u002Fcamel) - 第一个也是最好的多智能体框架\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpydantic\u002Fpydantic-ai?style=social\" height=\"17\" align=\"texttop\"\u002F> [pydantic-ai](https:\u002F\u002Fgithub.com\u002Fpydantic\u002Fpydantic-ai) - 一个Python代理框架，旨在帮助你快速、自信且轻松地构建生产级别的生成式AI应用和工作流\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fneuml\u002Ftxtai?style=social\" height=\"17\" align=\"texttop\"\u002F> [txtai](https:\u002F\u002Fgithub.com\u002Fneuml\u002Ftxtai) - 一体化开源AI框架，适用于语义搜索、LLM编排及语言模型工作流\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002Fagent-framework?style=social\" height=\"17\" align=\"texttop\"\u002F> [agent-framework](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fagent-framework) - 一个用于构建、编排和部署AI代理及多智能体工作流的框架，支持Python和.NET\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fkatanemo\u002Farchgw?style=social\" height=\"17\" align=\"texttop\"\u002F> [archgw](https:\u002F\u002Fgithub.com\u002Fkatanemo\u002Farchgw) - 一个高性能代理服务器，负责处理构建代理时的底层工作：例如应用护栏机制、将提示路由到合适的代理、统一LLM访问等\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fbadboysm890\u002FClaraVerse?style=social\" height=\"17\" align=\"texttop\"\u002F> [ClaraVerse](https:\u002F\u002Fgithub.com\u002Fbadboysm890\u002FClaraVerse) - 一个以隐私为先、完全本地化的AI工作空间，配备Ollama LLM聊天、工具调用、代理构建器、Stable Diffusion以及嵌入式n8n风格自动化功能\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdeepsense-ai\u002Fragbits?style=social\" height=\"17\" align=\"texttop\"\u002F> [ragbits](https:\u002F\u002Fgithub.com\u002Fdeepsense-ai\u002Fragbits) - 用于快速开发GenAI应用的构建模块\n\n[返回目录](#table-of-contents)\n\n### 模型上下文协议\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmindsdb\u002Fmindsdb?style=social\" height=\"17\" align=\"texttop\"\u002F> [mindsdb](https:\u002F\u002Fgithub.com\u002Fmindsdb\u002Fmindsdb) - 用于AI的联邦查询引擎，是你唯一需要的MCP服务器\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fgithub\u002Fgithub-mcp-server?style=social\" height=\"17\" align=\"texttop\"\u002F> [github-mcp-server](https:\u002F\u002Fgithub.com\u002Fgithub\u002Fgithub-mcp-server) - GitHub官方MCP服务器\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002Fplaywright-mcp?style=social\" height=\"17\" align=\"texttop\"\u002F> [playwright-mcp](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fplaywright-mcp) - Playwright MCP服务器\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FChromeDevTools\u002Fchrome-devtools-mcp?style=social\" height=\"17\" align=\"texttop\"\u002F> [chrome-devtools-mcp](https:\u002F\u002Fgithub.com\u002FChromeDevTools\u002Fchrome-devtools-mcp) - Chrome DevTools，用于编写代理程序\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002F czlonkowski\u002Fn8n-mcp?style=social\" height=\"17\" align=\"texttop\"\u002F> [n8n-mcp](https:\u002F\u002Fgithub.com\u002Fczlonkowski\u002Fn8n-mcp) - 一个MCP，专为Claude Desktop \u002F Claude Code \u002F Windsurf \u002F Cursor设计，可为你构建n8n工作流\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fawslabs\u002Fmcp?style=social\" height=\"17\" align=\"texttop\"\u002F> [awslabs\u002Fmcp](https:\u002F\u002Fgithub.com\u002Fawslabs\u002Fmcp) - AWS MCP服务器，无论你在何处使用MCP，都能帮助你最大化AWS的使用价值\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsooperset\u002Fmcp-atlassian?style=social\" height=\"17\" align=\"texttop\"\u002F> [mcp-atlassian](https:\u002F\u002Fgithub.com\u002Fsooperset\u002Fmcp-atlassian) - Atlassian工具（Confluence、Jira）的MCP服务器\n\n[返回目录](#table-of-contents)\n\n### 检索增强生成\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpathwaycom\u002Fpathway?style=social\" height=\"17\" align=\"texttop\"\u002F> [pathway](https:\u002F\u002Fgithub.com\u002Fpathwaycom\u002Fpathway) - 用于流处理、实时分析、LLM 流程和 RAG 的 Python ETL 框架\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002Fgraphrag?style=social\" height=\"17\" align=\"texttop\"\u002F> [graphrag](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fgraphrag) - 一个模块化的基于图的 RAG 系统\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FHKUDS\u002FLightRAG?style=social\" height=\"17\" align=\"texttop\"\u002F> [LightRAG](https:\u002F\u002Fgithub.com\u002FHKUDS\u002FLightRAG) - 简单快速的 RAG\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdeepset-ai\u002Fhaystack?style=social\" height=\"17\" align=\"texttop\"\u002F> [haystack](https:\u002F\u002Fgithub.com\u002Fdeepset-ai\u002Fhaystack) - 用于构建可定制、生产就绪的 LLM 应用程序的 AI 编排框架，非常适合构建 RAG、问答系统、语义搜索或对话式智能助手聊天机器人\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fvanna-ai\u002Fvanna?style=social\" height=\"17\" align=\"texttop\"\u002F> [vanna](https:\u002F\u002Fgithub.com\u002Fvanna-ai\u002Fvanna) - 一个开源的 Python RAG 框架，用于 SQL 生成及相关功能\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fgetzep\u002Fgraphiti?style=social\" height=\"17\" align=\"texttop\"\u002F> [graphiti](https:\u002F\u002Fgithub.com\u002Fgetzep\u002Fgraphiti) - 为 AI 代理构建实时知识图谱\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fonyx-dot-app\u002Fonyx?style=social\" height=\"17\" align=\"texttop\"\u002F> [onyx](https:\u002F\u002Fgithub.com\u002Fonyx-dot-app\u002Fonyx) - 与贵公司文档、应用程序和人员相连的 AI 平台\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fzilliztech\u002Fclaude-context?style=social\" height=\"17\" align=\"texttop\"\u002F> [claude-context](https:\u002F\u002Fgithub.com\u002Fzilliztech\u002Fclaude-context) - 将整个代码库作为任何编码代理的上下文\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpipeshub-ai\u002Fpipeshub-ai?style=social\" height=\"17\" align=\"texttop\"\u002F> [pipeshub-ai](https:\u002F\u002Fgithub.com\u002Fpipeshub-ai\u002Fpipeshub-ai) - 一个完全可扩展且可解释的企业级搜索与工作流自动化 AI 平台\n\n[返回目录](#table-of-contents)\n\n### 编码代理\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fzed-industries\u002Fzed?style=social\" height=\"17\" align=\"texttop\"\u002F> [zed](https:\u002F\u002Fgithub.com\u002Fzed-industries\u002Fzed) - 一款面向下一代的代码编辑器，专为与人类和 AI 高效协作而设计\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAll-Hands-AI\u002FOpenHands?style=social\" height=\"17\" align=\"texttop\"\u002F> [OpenHands](https:\u002F\u002Fgithub.com\u002FAll-Hands-AI\u002FOpenHands) - 一个由 AI 驱动的软件开发代理平台\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fcline\u002Fcline?style=social\" height=\"17\" align=\"texttop\"\u002F> [cline](https:\u002F\u002Fgithub.com\u002Fcline\u002Fcline) - 一款就在你的 IDE 中的自主编码代理，能够在你每一步的许可下创建\u002F编辑文件、执行命令、使用浏览器等\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAider-AI\u002Faider?style=social\" height=\"17\" align=\"texttop\"\u002F> [aider](https:\u002F\u002Fgithub.com\u002FAider-AI\u002Faider) - 在你的终端中进行 AI 配对编程\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsst\u002Fopencode?style=social\" height=\"17\" align=\"texttop\"\u002F> [opencode](https:\u002F\u002Fgithub.com\u002Fsst\u002Fopencode) - 一款专为终端打造的 AI 编码代理\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTabbyML\u002Ftabby?style=social\" height=\"17\" align=\"texttop\"\u002F> [tabby](https:\u002F\u002Fgithub.com\u002FTabbyML\u002Ftabby) - 一个开源的 GitHub Copilot 替代品，允许你搭建自己的 LLM 驱动的代码补全服务器\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fcontinuedev\u002Fcontinue?style=social\" height=\"17\" align=\"texttop\"\u002F> [continue](https:\u002F\u002Fgithub.com\u002Fcontinuedev\u002Fcontinue) - 使用我们的开源 IDE 扩展和模型、规则、提示、文档等构建块的中心，创建、分享和使用自定义的 AI 代码助手\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fvoideditor\u002Fvoid?style=social\" height=\"17\" align=\"texttop\"\u002F> [void](https:\u002F\u002Fgithub.com\u002Fvoideditor\u002Fvoid) - 一个开源的 Cursor 替代品，在你的代码库上使用 AI 代理，检查点并可视化更改，并将任何模型或主机本地化\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fblock\u002Fgoose?style=social\" height=\"17\" align=\"texttop\"\u002F> [goose](https:\u002F\u002Fgithub.com\u002Fblock\u002Fgoose) - 一个开源、可扩展的 AI 代理，超越了单纯的代码建议\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRooCodeInc\u002FRoo-Code?style=social\" height=\"17\" align=\"texttop\"\u002F> [Roo-Code](https:\u002F\u002Fgithub.com\u002FRooCodeInc\u002FRoo-Code) - 你的代码编辑器中的一整支由 AI 代理组成的开发团队\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fcharmbracelet\u002Fcrush?style=social\" height=\"17\" align=\"texttop\"\u002F> [crush](https:\u002F\u002Fgithub.com\u002Fcharmbracelet\u002Fcrush) - 一款为你喜爱的终端打造的迷人 AI 编码代理\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FKilo-Org\u002Fkilocode?style=social\" height=\"17\" align=\"texttop\"\u002F> [kilocode](https:\u002F\u002Fgithub.com\u002FKilo-Org\u002Fkilocode) - 开源的 AI 编码助手，用于规划、构建和修复代码\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fhumanlayer\u002Fhumanlayer?style=social\" height=\"17\" align=\"texttop\"\u002F> [humanlayer](https:\u002F\u002Fgithub.com\u002Fhumanlayer\u002Fhumanlayer) - 让 AI 编码代理解决复杂代码库中难题的最佳方式\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fcarlrobertoh\u002FProxyAI?style=social\" height=\"17\" align=\"texttop\"\u002F> [ProxyAI](https:\u002F\u002Fgithub.com\u002Fcarlrobertoh\u002FProxyAI) - JetBrains 领先的开源 AI 合作伙伴\n\n[返回目录](#table-of-contents)\n\n### 计算机使用\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOpenInterpreter\u002Fopen-interpreter?style=social\" height=\"17\" align=\"texttop\"\u002F> [open-interpreter](https:\u002F\u002Fgithub.com\u002FOpenInterpreter\u002Fopen-interpreter) - 一种用于计算机的自然语言接口\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002FOmniParser?style=social\" height=\"17\" align=\"texttop\"\u002F> [OmniParser](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FOmniParser) - 一个简单的屏幕解析工具，旨在实现纯视觉驱动的 GUI 代理\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ftrycua\u002Fcua?style=social\" height=\"17\" align=\"texttop\"\u002F> [cua](https:\u002F\u002Fgithub.com\u002Ftrycua\u002Fcua) - 用于计算机使用 AI 代理的 Docker 容器\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOthersideAI\u002Fself-operating-computer?style=social\" height=\"17\" align=\"texttop\"\u002F> [self-operating-computer](https:\u002F\u002Fgithub.com\u002FOthersideAI\u002Fself-operating-computer) - 一个使多模态模型能够操作计算机的框架\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsimular-ai\u002FAgent-S?style=social\" height=\"17\" align=\"texttop\"\u002F> [Agent-S](https:\u002F\u002Fgithub.com\u002Fsimular-ai\u002FAgent-S) - 一个开放的代理框架，像人类一样使用计算机\n\n[返回目录](#table-of-contents)\n\n### 浏览器自动化\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpuppeteer\u002Fpuppeteer?style=social\" height=\"17\" align=\"texttop\"\u002F> [puppeteer](https:\u002F\u002Fgithub.com\u002Fpuppeteer\u002Fpuppeteer) - 一个用于 Chrome 和 Firefox 的 JavaScript API\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002Fplaywright?style=social\" height=\"17\" align=\"texttop\"\u002F> [playwright](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fplaywright) - 一个用于 Web 测试和自动化的框架\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fbrowser-use\u002Fbrowser-use?style=social\" height=\"17\" align=\"texttop\"\u002F> [browser-use](https:\u002F\u002Fgithub.com\u002Fbrowser-use\u002Fbrowser-use) - 让网站对 AI 代理可用\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmendableai\u002Ffirecrawl?style=social\" height=\"17\" align=\"texttop\"\u002F> [firecrawl](https:\u002F\u002Fgithub.com\u002Fmendableai\u002Ffirecrawl) - 将整个网站转换为适合 LLM 的 Markdown 或结构化数据\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fbrowserbase\u002Fstagehand?style=social\" height=\"17\" align=\"texttop\"\u002F> [stagehand](https:\u002F\u002Fgithub.com\u002Fbrowserbase\u002Fstagehand) - AI 浏览器自动化框架\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fnanobrowser\u002Fnanobrowser?style=social\" height=\"17\" align=\"texttop\"\u002F> [nanobrowser](https:\u002F\u002Fgithub.com\u002Fnanobrowser\u002Fnanobrowser) - 开源 Chrome 扩展，用于 AI 驱动的网页自动化\n\n[返回目录](#table-of-contents)\n\n### 内存管理\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmem0ai\u002Fmem0?style=social\" height=\"17\" align=\"texttop\"\u002F> [mem0](https:\u002F\u002Fgithub.com\u002Fmem0ai\u002Fmem0) - 适用于 AI 代理的通用内存层\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fletta-ai\u002Fletta?style=social\" height=\"17\" align=\"texttop\"\u002F> [letta](https:\u002F\u002Fgithub.com\u002Fletta-ai\u002Fletta) - 具有记忆、推理和上下文管理功能的状态型代理框架\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsupermemoryai\u002Fsupermemory?style=social\" height=\"17\" align=\"texttop\"\u002F> [supermemory](https:\u002F\u002Fgithub.com\u002Fsupermemoryai\u002Fsupermemory) - 极其快速且可扩展的内存引擎和应用\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ftopoteretes\u002Fcognee?style=social\" height=\"17\" align=\"texttop\"\u002F> [cognee](https:\u002F\u002Fgithub.com\u002Ftopoteretes\u002Fcognee) - 用 5 行代码实现的 AI 代理内存\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FLMCache\u002FLMCache?style=social\" height=\"17\" align=\"texttop\"\u002F> [LMCache](https:\u002F\u002Fgithub.com\u002FLMCache\u002FLMCache) - 使用最快的 KV 缓存层加速你的 LLM\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNevaMind-AI\u002FmemU?style=social\" height=\"17\" align=\"texttop\"\u002F> [memU](https:\u002F\u002Fgithub.com\u002FNevaMind-AI\u002FmemU) - 一个开源的 AI 伴侣内存框架\n\n[返回目录](#table-of-contents)\n\n### 测试、评估与可观测性\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flangfuse\u002Flangfuse?style=social\" height=\"17\" align=\"texttop\"\u002F> [langfuse](https:\u002F\u002Fgithub.com\u002Flangfuse\u002Flangfuse) - 一个开源的 LLM 工程平台：LLM 可观测性、指标、评估、提示管理、游乐场、数据集。与 OpenTelemetry、Langchain、OpenAI SDK、LiteLLM 等集成\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fcomet-ml\u002Fopik?style=social\" height=\"17\" align=\"texttop\"\u002F> [opik](https:\u002F\u002Fgithub.com\u002Fcomet-ml\u002Fopik) - 使用全面的追踪、自动化评估和生产就绪的仪表板来调试、评估和监控你的 LLM 应用程序、RAG 系统和代理式工作流\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ftraceloop\u002Fopenllmetry?style=social\" height=\"17\" align=\"texttop\"\u002F> [openllmetry](https:\u002F\u002Fgithub.com\u002Ftraceloop\u002Fopenllmetry) - 基于 OpenTelemetry 的开源 LLM 应用程序可观测性工具\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNVIDIA\u002Fgarak?style=social\" height=\"17\" align=\"texttop\"\u002F> [garak](https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fgarak) - NVIDIA 提供的 LLM 漏洞扫描器\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FGiskard-AI\u002Fgiskard?style=social\" height=\"17\" align=\"texttop\"\u002F> [giskard](https:\u002F\u002Fgithub.com\u002FGiskard-AI\u002Fgiskard) - 一个开源的 AI 和 LLM 系统评估与测试工具\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAgenta-AI\u002Fagenta?style=social\" height=\"17\" align=\"texttop\"\u002F> [agenta](https:\u002F\u002Fgithub.com\u002FAgenta-AI\u002Fagenta) - 一个开源的 LLMOps 平台：提示游乐场、提示管理、LLM 评估和 LLM 可观测性，全部集中在一个地方\n\n[返回目录](#table-of-contents)\n\n### 研究\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FItzCrazyKns\u002FPerplexica?style=social\" height=\"17\" align=\"texttop\"\u002F> [Perplexica](https:\u002F\u002Fgithub.com\u002FItzCrazyKns\u002FPerplexica) - 一个开源的 Perplexity AI 替代品，即基于 AI 的搜索引擎\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fassafelovic\u002Fgpt-researcher?style=social\" height=\"17\" align=\"texttop\"\u002F> [gpt-researcher](https:\u002F\u002Fgithub.com\u002Fassafelovic\u002Fgpt-researcher) - 一个基于 LLM 的自主代理，能够针对任何主题进行深入的本地和网络研究，并生成带有引用的长篇报告\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMODSetter\u002FSurfSense?style=social\" height=\"17\" align=\"texttop\"\u002F> [SurfSense](https:\u002F\u002Fgithub.com\u002FMODSetter\u002FSurfSense) - 一个开源的 NotebookLM \u002F Perplexity \u002F Glean 替代品\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flfnovo\u002Fopen-notebook?style=social\" height=\"17\" align=\"texttop\"\u002F> [open-notebook](https:\u002F\u002Fgithub.com\u002Flfnovo\u002Fopen-notebook) - 一个具有更高灵活性和更多功能的 Notebook LM 开源实现\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002FRD-Agent?style=social\" height=\"17\" align=\"texttop\"\u002F> [RD-Agent](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FRD-Agent) - 自动化工业研发流程中最关键和最有价值的部分\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flangchain-ai\u002Flocal-deep-researcher?style=social\" height=\"17\" align=\"texttop\"\u002F> [local-deep-researcher](https:\u002F\u002Fgithub.com\u002Flangchain-ai\u002Flocal-deep-researcher) - 一个完全本地的网络研究和报告撰写助手\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FLearningCircuit\u002Flocal-deep-research?style=social\" height=\"17\" align=\"texttop\"\u002F> [local-deep-research](https:\u002F\u002Fgithub.com\u002FLearningCircuit\u002Flocal-deep-research) - 一个用于深度迭代研究的 AI 驱动的研究助手\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmurtaza-nasir\u002Fmaestro?style=social\" height=\"17\" align=\"texttop\"\u002F> [maestro](https:\u002F\u002Fgithub.com\u002Fmurtaza-nasir\u002Fmaestro) - 一个旨在简化复杂研究任务的 AI 驱动的研究应用\n\n[返回目录](#table-of-contents)\n\n### 训练与微调\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOpenRLHF\u002FOpenRLHF?style=social\" height=\"17\" align=\"texttop\"\u002F> [OpenRLHF](https:\u002F\u002Fgithub.com\u002FOpenRLHF\u002FOpenRLHF) - 一个基于 Ray、vLLM、ZeRO-3 和 HuggingFace Transformers 构建的易用、高性能开源 RLHF 框架，旨在让 RLHF 训练变得简单且易于访问\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fkiln-ai\u002Fkiln?style=social\" height=\"17\" align=\"texttop\"\u002F> [Kiln](https:\u002F\u002Fgithub.com\u002Fkiln-ai\u002Fkiln) - 微调 LLM 模型、生成合成数据以及协作处理数据集的最简单工具\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fe-p-armstrong\u002Faugmentoolkit?style=social\" height=\"17\" align=\"texttop\"\u002F> [augmentoolkit](https:\u002F\u002Fgithub.com\u002Fe-p-armstrong\u002Faugmentoolkit) - 使用新事实训练开源 LLM\n\n[返回目录](#table-of-contents)\n\n### 其他\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fupstash\u002Fcontext7?style=social\" height=\"17\" align=\"texttop\"\u002F> [context7](https:\u002F\u002Fgithub.com\u002Fupstash\u002Fcontext7) - 针对 LLM 和 AI 代码编辑器的最新代码文档\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Faliasrobotics\u002Fcai?style=social\" height=\"17\" align=\"texttop\"\u002F> [cai](https:\u002F\u002Fgithub.com\u002Faliasrobotics\u002Fcai) - 网络安全人工智能（CAI），AI 安全框架\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmurtaza-nasir\u002Fspeakr?style=social\" height=\"17\" align=\"texttop\"\u002F> [speakr](https:\u002F\u002Fgithub.com\u002Fmurtaza-nasir\u002Fspeakr) - 一款个人自托管的 Web 应用程序，专为转录音频记录而设计\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpresenton\u002Fpresenton?style=social\" height=\"17\" align=\"texttop\"\u002F> [presenton](https:\u002F\u002Fgithub.com\u002Fpresenton\u002Fpresenton) - 开源 AI 演示文稿生成器及 API\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FVectorSpaceLab\u002FOmniGen2?style=social\" height=\"17\" align=\"texttop\"\u002F> [OmniGen2](https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FOmniGen2) - 探索高级多模态生成\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTheAhmadOsman\u002F4o-ghibli-at-home?style=social\" height=\"17\" align=\"texttop\"\u002F> [4o-ghibli-at-home](https:\u002F\u002Fgithub.com\u002FTheAhmadOsman\u002F4o-ghibli-at-home) - 一款功能强大、自托管的 AI 照片风格化工具，专为性能和隐私而设计\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRoy3838\u002FObserver?style=social\" height=\"17\" align=\"texttop\"\u002F> [Observer](https:\u002F\u002Fgithub.com\u002FRoy3838\u002FObserver) - 本地开源微型智能体，可观察、记录并作出反应，同时确保您的数据私密且安全\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fminitap-ai\u002Fmobile-use?style=social\" height=\"17\" align=\"texttop\"\u002F> [mobile-use](https:\u002F\u002Fgithub.com\u002Fminitap-ai\u002Fmobile-use) - 一款功能强大的开源 AI 助手，可通过自然语言控制您的 Android 或 iOS 设备\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fgabber-dev\u002Fgabber?style=social\" height=\"17\" align=\"texttop\"\u002F> [gabber](https:\u002F\u002Fgithub.com\u002Fgabber-dev\u002Fgabber) - 利用您的屏幕、麦克风和摄像头作为输入，构建能够看、听、说的 AI 应用程序\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsevenreasons\u002Fpromptcat?style=social\" height=\"17\" align=\"texttop\"\u002F> [promptcat](https:\u002F\u002Fgithub.com\u002Fsevenreasons\u002Fpromptcat) - 一个零依赖的提示管理器\u002F目录\u002F库，集成在一个 HTML 文件中\n\n[返回目录](#table-of-contents)\n\n## 硬件\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fchannel\u002Fsubscribers\u002FUCajiMK_CY9icRhLepS8_3ug?style=social\" height=\"17\" align=\"texttop\"\u002F> [Alex Ziskind](https:\u002F\u002Fwww.youtube.com\u002F@AZisk) - 测试能够运行 LLM 的 PC、笔记本电脑、GPU 等设备\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fchannel\u002Fsubscribers\u002FUCiaQzXI5528Il6r2NNkrkJA?style=social\" height=\"17\" align=\"texttop\"\u002F> [Digital Spaceport](https:\u002F\u002Fwww.youtube.com\u002F@DigitalSpaceport) - 评测各种专为 LLM 推理设计的硬件配置\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fchannel\u002Fsubscribers\u002FUCQs0lwV6E4p7LQaGJ6fgy5Q?style=social\" height=\"17\" align=\"texttop\"\u002F> [JetsonHacks](https:\u002F\u002Fwww.youtube.com\u002F@JetsonHacks) - 关于在 NVIDIA Jetson 开发套件上进行开发的信息\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fchannel\u002Fsubscribers\u002FUC8h2Sf-yyo1WXeEUr-OHgyg?style=social\" height=\"17\" align=\"texttop\"\u002F> [Miyconst](https:\u002F\u002Fwww.youtube.com\u002F@Miyconst) - 测试各种能够运行 LLM 的硬件类型\n- [Kolosal - LLM 内存计算器](https:\u002F\u002Fwww.kolosal.ai\u002Fmemory-calculator) - 即时估算任何 GGUF 模型所需的 RAM 大小\n- [LLM 推理 VRAM & GPU 要求计算器](https:\u002F\u002Fapp.linpp2009.com\u002Fen\u002Fllm-gpu-memory-calculator) - 计算部署 LLM 需要多少块 GPU\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fvosen\u002FZLUDA?style=social\" height=\"17\" align=\"texttop\"\u002F> [ZLUDA](https:\u002F\u002Fgithub.com\u002Fvosen\u002FZLUDA) - 在非 NVIDIA GPU 上使用 CUDA\n\n[返回目录](#table-of-contents)\n\n## 教程\n\n### 模型\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002Fl8pRSuU81PU?style=social\" height=\"17\" align=\"texttop\"\u002F> [让我们重现 GPT-2 (124M)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=l8pRSuU81PU)\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fkarpathy\u002Fnanochat?style=social\" height=\"17\" align=\"texttop\"\u002F> [nanochat](https:\u002F\u002Fgithub.com\u002Fkarpathy\u002Fnanochat) - 一个全栈式实现的类似 ChatGPT 的 LLM，采用单一、简洁、极简、可 hack 且依赖极少的代码库，专为通过 speedrun.sh 等脚本在单个 8XH100 节点上运行整个流水线而设计\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fyoutube\u002Fviews\u002FjrJKRYAdh7I?style=social\" height=\"17\" align=\"texttop\"\u002F> [知识蒸馏：LLM 如何相互训练](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=jrJKRYAdh7I)\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fiuliaturc\u002Fgguf-docs?style=social\" height=\"17\" align=\"texttop\"\u002F> [gguf-docs](https:\u002F\u002Fgithub.com\u002Fiuliaturc\u002Fgguf-docs) - GGUF 量化文档（非官方）\n\n[返回目录](#table-of-contents)\n\n### 提示工程\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdair-ai\u002FPrompt-Engineering-Guide?style=social\" height=\"17\" align=\"texttop\"\u002F> [提示工程指南](https:\u002F\u002Fgithub.com\u002Fdair-ai\u002FPrompt-Engineering-Guide) - 提示工程相关的指南、论文、讲座、笔记本及资源\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNirDiamant\u002FPrompt_Engineering?style=social\" height=\"17\" align=\"texttop\"\u002F> [NirDiamant 的提示工程](https:\u002F\u002Fgithub.com\u002FNirDiamant\u002FPrompt_Engineering) - 一套全面的教程和实现，涵盖从基础概念到高级策略的提示工程技术\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FGoogle-%234285F4?logo=google&logoColor=red\" height=\"17\" align=\"texttop\"\u002F> [提示引导入门指南](https:\u002F\u002Fservices.google.com\u002Ffh\u002Ffiles\u002Fmisc\u002Fgemini-for-google-workspace-prompting-guide-101.pdf) - Google 提供的有效提示快速入门手册\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FGoogle-%234285F4?logo=google&logoColor=red\" height=\"17\" align=\"texttop\"\u002F> [Google 的提示工程](https:\u002F\u002Fdrive.google.com\u002Ffile\u002Fd\u002F1AbaBYbEa_EbPelsT40-vj64L-2IwUJHy\u002Fview) - Google 的提示工程资料\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FAnthropic-%23191919?logo=anthropic\" height=\"17\" align=\"texttop\"\u002F> [Anthropic 的提示工程](https:\u002F\u002Fdocs.anthropic.com\u002Fen\u002Fdocs\u002Fbuild-with-claude\u002Fprompt-engineering\u002Foverview) - Anthropic 的提示工程文档\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FAnthropic-%23191919?logo=anthropic\" height=\"17\" align=\"texttop\"\u002F> [提示工程互动教程](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fcourses\u002Fblob\u002Fmaster\u002Fprompt_engineering_interactive_tutorial\u002FREADME.md) - Anthropic 提供的提示工程互动教程\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FAnthropic-%23191919?logo=anthropic\" height=\"17\" align=\"texttop\"\u002F> [现实世界中的提示](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fcourses\u002Fblob\u002Fmaster\u002Freal_world_prompting\u002FREADME.md) - Anthropic 提供的现实世界提示教程\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FAnthropic-%23191919?logo=anthropic\" height=\"17\" align=\"texttop\"\u002F> [提示评估](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fcourses\u002Fblob\u002Fmaster\u002Fprompt_evaluations\u002FREADME.md) - Anthropic 的提示评估课程\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fx1xhlol\u002Fsystem-prompts-and-models-of-ai-tools?style=social\" height=\"17\" align=\"texttop\"\u002F> [system-prompts-and-models-of-ai-tools](https:\u002F\u002Fgithub.com\u002Fx1xhlol\u002Fsystem-prompts-and-models-of-ai-tools) - 从 AI 工具中提取的系统提示集合\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fasgeirtj\u002Fsystem_prompts_leaks?style=social\" height=\"17\" align=\"texttop\"\u002F> [system_prompts_leaks](https:\u002F\u002Fgithub.com\u002Fasgeirtj\u002Fsystem_prompts_leaks) - 从 ChatGPT、Claude 和 Gemini 等热门聊天机器人中提取的系统提示集合\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FOpenAI-%23412991?logo=openai\" height=\"17\" align=\"texttop\"\u002F> \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fopenai\u002Fcodex?style=social\" height=\"17\" align=\"texttop\"\u002F> [Codex 的提示](https:\u002F\u002Fgithub.com\u002Fopenai\u002Fcodex\u002Fblob\u002Fmain\u002Fcodex-rs\u002Fcore\u002Fprompt.md) - OpenAI Codex 用于引导行为的提示\n\n[返回目录](#table-of-contents)\n\n### 上下文工程\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdavidkimai\u002FContext-Engineering?style=social\" height=\"17\" align=\"texttop\"\u002F> [上下文工程](https:\u002F\u002Fgithub.com\u002Fdavidkimai\u002FContext-Engineering) - 受 Karpathy 和 3Blue1Brown 启发的前沿、基于第一性原理的手册，旨在超越提示工程，进入更广泛的上下文设计、编排和优化领域\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMeirtz\u002FAwesome-Context-Engineering?style=social\" height=\"17\" align=\"texttop\"\u002F> [Awesome-Context-Engineering](https:\u002F\u002Fgithub.com\u002FMeirtz\u002FAwesome-Context-Engineering) - 关于上下文工程的全面综述：从提示工程到生产级 AI 系统\n\n[返回目录](#table-of-contents)\n\n### 推理\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fvllm-project\u002Fproduction-stack?style=social\" height=\"17\" align=\"texttop\"\u002F> [vLLM 生产栈](https:\u002F\u002Fgithub.com\u002Fvllm-project\u002Fproduction-stack) - vLLM 针对 K8S 原生集群部署的参考系统，并由社区驱动进行性能优化\n\n[返回目录](#table-of-contents)\n\n### 代理\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNirDiamant\u002FGenAI_Agents?style=social\" height=\"17\" align=\"texttop\"\u002F> [GenAI 代理](https:\u002F\u002Fgithub.com\u002FNirDiamant\u002FGenAI_Agents) - 各类生成式 AI 代理技术的教程和实现\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fashishpatel26\u002F500-AI-Agents-Projects?style=social\" height=\"17\" align=\"texttop\"\u002F> [500 多个 AI 代理项目](https:\u002F\u002Fgithub.com\u002Fashishpatel26\u002F500-AI-Agents-Projects) - 涵盖各行业的 AI 代理用例精选集\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fhumanlayer\u002F12-factor-agents?style=social\" height=\"17\" align=\"texttop\"\u002F> [12 因子代理](https:\u002F\u002Fgithub.com\u002Fhumanlayer\u002F12-factor-agents) - 构建可靠 LLM 应用的原则\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNirDiamant\u002Fagents-towards-production?style=social\" height=\"17\" align=\"texttop\"\u002F> [迈向生产的代理](https:\u002F\u002Fgithub.com\u002FNirDiamant\u002Fagents-towards-production) - 从头到尾、以代码为导向的教程，覆盖生产级生成式 AI 代理的每一层，通过成熟的模式和可重用蓝图，指导您从萌芽阶段走向规模化落地\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Foxbshw\u002FLLM-Agents-Ecosystem-Handbook?style=social\" height=\"17\" align=\"texttop\"\u002F> [LLM 代理与生态系统手册](https:\u002F\u002Fgithub.com\u002Foxbshw\u002FLLM-Agents-Ecosystem-Handbook) - 一站式手册，包含 60 多种骨架、教程、生态系统指南和评估工具，帮助您构建、部署和理解 LLM 代理\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FGoogle-%234285F4?logo=google&logoColor=red\" height=\"17\" align=\"texttop\"\u002F> [601 个现实世界的生成式 AI 用例](https:\u002F\u002Fcloud.google.com\u002Ftransform\u002F101-real-world-generative-ai-use-cases-from-industry-leaders) - Google 整理的来自全球领先企业的 601 个现实世界生成式 AI 用例\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FOpenAI-%23412991?logo=openai\" height=\"17\" align=\"texttop\"\u002F> [构建代理的实用指南](https:\u002F\u002Fcdn.openai.com\u002Fbusiness-guides-and-resources\u002Fa-practical-guide-to-building-agents.pdf) - OpenAI 提供的构建代理实用指南\n\n[返回目录](#table-of-contents)\n\n### 检索增强生成\n\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpathwaycom\u002Fllm-app?style=social\" height=\"17\" align=\"texttop\"\u002F> [Pathway AI 管道](https:\u002F\u002Fgithub.com\u002Fpathwaycom\u002Fllm-app) - 用于 RAG、AI 流水线和企业级实时数据搜索的开箱即用云模板\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNirDiamant\u002FRAG_Techniques?style=social\" height=\"17\" align=\"texttop\"\u002F> [RAG 技术](https:\u002F\u002Fgithub.com\u002FNirDiamant\u002FRAG_Techniques) - 针对检索增强生成（RAG）系统的各种高级技术\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNirDiamant\u002FControllable-RAG-Agent?style=social\" height=\"17\" align=\"texttop\"\u002F> [可控 RAG 代理](https:\u002F\u002Fgithub.com\u002FNirDiamant\u002FControllable-RAG-Agent) - 一种用于复杂问答任务的先进检索增强生成解决方案，采用复杂的基于图的算法来处理任务\n- \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flokeswaran-aj\u002Flangchain-rag-cookbook?style=social\" height=\"17\" align=\"texttop\"\u002F> [LangChain RAG 烹饪书](https:\u002F\u002Fgithub.com\u002Flokeswaran-aj\u002Flangchain-rag-cookbook) - 一系列模块化的 RAG 技术，使用 LangChain + Python 实现\n\n[返回目录](#table-of-contents)\n\n### 杂项\n\n- [开箱即用的自托管 AI 编码工具](https:\u002F\u002Fwww.reddit.com\u002Fr\u002FLocalLLaMA\u002Fcomments\u002F1lt4y1z\u002Fselfhosted_ai_coding_that_just_works\u002F)\n\n[返回目录](#table-of-contents)","# LLMs-local 快速上手指南\n\n`LLMs-local` 并非单一软件，而是一个精选的本地大语言模型（LLM）生态资源列表。本指南将带你从零开始，利用列表中推荐的工具在本地运行开源大模型。对于中国开发者，我们推荐从 **Ollama**（推理引擎）配合 **Open WebUI**（用户界面）或 **LM Studio**（一体化平台）入手。\n\n## 环境准备\n\n在开始之前，请确保你的硬件和系统满足以下基本要求：\n\n*   **操作系统**：Windows 10\u002F11, macOS (Intel 或 Apple Silicon), 或 Linux (Ubuntu\u002FDebian 推荐)。\n*   **内存 (RAM)**：\n    *   运行 7B 参数模型：建议至少 8GB - 16GB。\n    *   运行更大模型（如 14B+）：建议 32GB 或以上。\n*   **显卡 (GPU)**（可选但强烈推荐）：\n    *   NVIDIA GPU：显存越大越好（8GB 起步），需安装最新显卡驱动。\n    *   Apple Silicon (M1\u002FM2\u002FM3)：统一内存架构对本地推理非常友好。\n    *   AMD GPU：部分工具（如 `vllm-gfx906` 或 `llm-scaler`）提供支持，但配置相对复杂。\n*   **前置依赖**：\n    *   已安装 `Git`。\n    *   已安装 `Python 3.8+` (部分高级工具需要)。\n    *   **网络环境**：由于模型权重托管在 Hugging Face，国内访问可能较慢，建议配置代理或使用国内镜像源。\n\n## 安装步骤\n\n以下提供两种最主流的安装方案：**方案 A** 适合追求极致简便的用户，**方案 B** 适合喜欢命令行和轻量级部署的开发者。\n\n### 方案 A：使用 LM Studio (图形化一体版)\n*适合不想配置环境，希望“下载即运行”的用户。*\n\n1.  **下载安装包**：\n    访问 [LM Studio 官网](https:\u002F\u002Flmstudio.ai\u002F) 下载对应系统的安装包。\n    > **提示**：若官网下载慢，可尝试在国内技术社区搜索离线安装包。\n\n2.  **安装并启动**：\n    运行安装程序，启动 LM Studio。\n\n3.  **搜索并下载模型**：\n    *   在左侧搜索栏输入模型名称（例如 `Qwen2.5` 或 `Gemma`）。\n    *   选择右侧列表中的量化版本（推荐 `GGUF` 格式，如 `Q4_K_M`，平衡速度与精度）。\n    *   点击 \"Download\"。\n\n### 方案 B：使用 Ollama + Open WebUI (推荐开发者)\n*适合需要 API 集成、Docker 部署及更灵活控制的用户。*\n\n#### 1. 安装 Ollama (推理后端)\n在终端执行以下命令：\n\n**macOS \u002F Linux:**\n```bash\ncurl -fsSL https:\u002F\u002Follama.com\u002Finstall.sh | sh\n```\n\n**Windows:**\n下载 [Ollama Setup.exe](https:\u002F\u002Follama.com\u002Fdownload\u002FOllamaSetup.exe) 并运行。\n\n> **国内加速提示**：若上述脚本执行超时，可使用国内镜像变量安装：\n> ```bash\n> export OLLAMA_HOST=\"0.0.0.0\"\n> curl -fsSL https:\u002F\u002Follama.com\u002Finstall.sh | sh\n> # 若下载模型慢，可设置代理：export HTTPS_PROXY=http:\u002F\u002F127.0.0.1:7890\n> ```\n\n#### 2. 安装 Open WebUI (图形化前端)\n推荐使用 Docker 部署，需先安装 Docker Desktop。\n\n```bash\ndocker run -d -p 3000:8080 --add-host=host.docker.internal:host-gateway -v open-webui:\u002Fapp\u002Fbackend\u002Fdata --name open-webui --restart always ghcr.io\u002Fopen-webui\u002Fopen-webui:main\n```\n\n安装完成后，浏览器访问 `http:\u002F\u002Flocalhost:3000` 即可使用。\n\n## 基本使用\n\n### 1. 运行第一个模型 (以 Qwen2.5 为例)\n\n**如果你使用 Ollama：**\n直接在终端运行以下命令，Ollama 会自动拉取并运行模型：\n```bash\nollama run qwen2.5\n```\n*注：`qwen2.5` 是阿里云通义千问系列的热门开源模型，对中文支持极佳。若需其他模型，替换名称即可，如 `llama3`, `gemma2`, `mistral`。*\n\n**如果你使用 LM Studio：**\n1.  点击左侧 \"Local Server\" 标签页。\n2.  在顶部下拉菜单选择已下载的模型。\n3.  点击 \"Start Server\"。\n4.  在聊天界面直接对话，或通过 `http:\u002F\u002Flocalhost:1234\u002Fv1` 调用 API。\n\n### 2. 简单的 API 调用示例\n\n本地服务启动后，你可以像使用 OpenAI API 一样调用本地模型。以下是一个 Python 示例：\n\n```python\nfrom openai import OpenAI\n\n# 指向本地 Ollama 或 LM Studio 的服务地址\nclient = OpenAI(\n    base_url=\"http:\u002F\u002Flocalhost:11434\u002Fv1\", # Ollama 默认端口\n    api_key=\"ollama\" # Ollama 不需要真实 key，填任意字符串即可\n)\n\nresponse = client.chat.completions.create(\n    model=\"qwen2.5\",\n    messages=[\n        {\"role\": \"system\", \"content\": \"你是一个有用的助手。\"},\n        {\"role\": \"user\", \"content\": \"请用中文简要介绍什么是大语言模型？\"}\n    ]\n)\n\nprint(response.choices[0].message.content)\n```\n\n### 3. 常用管理命令 (Ollama)\n\n*   **查看已下载模型**：\n    ```bash\n    ollama list\n    ```\n*   **删除模型**：\n    ```bash\n    ollama rm \u003Cmodel_name>\n    ```\n*   **后台运行服务**：\n    ```bash\n    ollama serve\n    ```\n\n通过以上步骤，你已成功搭建起本地的 AI 开发环境。你可以进一步探索 `LLMs-local` 列表中的高级工具，如 `llama.cpp` 进行底层优化，或使用 `Text generation web UI` 体验更多微调功能。","某金融科技公司的高级数据分析师需要在完全隔离的内网环境中，对每日更新的敏感财报数据进行自动化摘要与风险提取。\n\n### 没有 LLMs-local 时\n- **数据泄露风险高**：由于缺乏本地部署方案，团队被迫将脱敏不彻底的财报数据上传至公有云 API，时刻面临合规审计压力。\n- **环境配置极其繁琐**：尝试自行编译 `llama.cpp` 或配置 `vllm` 时，常因显卡驱动版本、CUDA 依赖冲突导致数天无法跑通推理流程。\n- **模型选型盲目低效**：面对海量开源模型，缺乏统一的基准测试和分类索引，难以快速找到适合“金融文本理解”且能在现有显存下运行的特定模型。\n- **响应延迟不可控**：依赖外部网络调用大模型，一旦网络波动或云端限流，整个自动化分析流水线就会中断，无法保证日报准时产出。\n\n### 使用 LLMs-local 后\n- **实现极致数据安全**：利用列表中推荐的 `LM Studio` 或 `Jan`，分析师直接在本地离线运行高性能模型，确保敏感财报数据从未离开公司内网。\n- **一键部署推理环境**：参考 curated 清单中的成熟平台，几分钟内即可拉起基于 `Ollama` 或 `LocalAI` 的服务，自动适配硬件加速，无需手动解决依赖地狱。\n- **精准匹配业务模型**：通过目录中清晰的“通用”、“编码”及“基准测试”分类，迅速锁定并下载针对金融领域微调过的量化模型，最大化利用有限显存。\n- **稳定高效的本地集群**：借助 `exo` 等工具将多台旧电脑组建成本地 AI 集群，不仅消除了网络延迟，还大幅提升了批量处理财报的吞吐量。\n\nLLMs-local 通过提供一站式的本地大模型生态导航，让企业在零数据泄露风险的前提下，以最低门槛实现了高效、可控的私有化 AI 落地。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002F0xSojalSec_LLMs-local_f164bf16.png","0xSojalSec","MD ISMAIL SOJAL","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002F0xSojalSec_dc12a8a4.jpg","Cyber Security Re-searcher | Malware Analysis | AI Re-searcher | Studying OSCP \r\n| Open-source AI,LLM stuff | Post-training \u002F Reasonign Models \u002F RAG | AI Agent ",null,"INTERNET","0x0SojalSec","\u002Fdev\u002Fnull","https:\u002F\u002Fgithub.com\u002F0xSojalSec",622,63,"2026-04-01T08:49:33","Linux, macOS, Windows","非绝对必需（支持 CPU 推理），但推荐 NVIDIA GPU (CUDA)、AMD GPU (ROCm\u002Fgfx906)、Intel Arc GPU 或 Apple Silicon (M1\u002FM2\u002FM3)。部分工具支持 NPU (AMD Ryzen AI)。显存需求取决于模型大小，通常建议 8GB+ 以运行中等规模模型。","未说明（取决于模型大小，通常建议 16GB+ 以流畅运行本地大模型）",{"notes":88,"python":89,"dependencies":90},"该 README 是一个本地运行大语言模型（LLMs）的工具和资源汇总列表，而非单一软件的具体安装指南。它列出了多种推理平台（如 LM Studio, Jan, LocalAI）、推理引擎（如 ollama, llama.cpp, vllm）和用户界面。硬件需求高度依赖于所选的具体工具和模型大小：Apple 用户可使用 mlx-lm；AMD 和 Intel 用户有特定的分支支持；大多数工具支持 GGUF 格式的量化模型以降低内存和显存需求。建议使用量化工具（如 Unsloth, bartowski 提供的模型）在消费级硬件上运行。","未说明",[91,92,93,94,95,96,97],"llama.cpp","ollama","vllm","mlx-lm (Apple Silicon)","torch","transformers","GGUF format support",[35,14,13,99],"其他","2026-03-27T02:49:30.150509","2026-04-11T23:23:13.377633",[],[]]