[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-infiniflow--ragflow":3,"tool-infiniflow--ragflow":64},[4,17,27,35,43,56],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":16},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。",162132,3,"2026-04-05T11:01:52",[13,14,15],"开发框架","图像","Agent","ready",{"id":18,"name":19,"github_repo":20,"description_zh":21,"stars":22,"difficulty_score":23,"last_commit_at":24,"category_tags":25,"status":16},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",138956,2,"2026-04-05T11:33:21",[13,15,26],"语言模型",{"id":28,"name":29,"github_repo":30,"description_zh":31,"stars":32,"difficulty_score":23,"last_commit_at":33,"category_tags":34,"status":16},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。",107662,"2026-04-03T11:11:01",[13,14,15],{"id":36,"name":37,"github_repo":38,"description_zh":39,"stars":40,"difficulty_score":23,"last_commit_at":41,"category_tags":42,"status":16},3704,"NextChat","ChatGPTNextWeb\u002FNextChat","NextChat 是一款轻量且极速的 AI 助手，旨在为用户提供流畅、跨平台的大模型交互体验。它完美解决了用户在多设备间切换时难以保持对话连续性，以及面对众多 AI 模型不知如何统一管理的痛点。无论是日常办公、学习辅助还是创意激发，NextChat 都能让用户随时随地通过网页、iOS、Android、Windows、MacOS 或 Linux 端无缝接入智能服务。\n\n这款工具非常适合普通用户、学生、职场人士以及需要私有化部署的企业团队使用。对于开发者而言，它也提供了便捷的自托管方案，支持一键部署到 Vercel 或 Zeabur 等平台。\n\nNextChat 的核心亮点在于其广泛的模型兼容性，原生支持 Claude、DeepSeek、GPT-4 及 Gemini Pro 等主流大模型，让用户在一个界面即可自由切换不同 AI 能力。此外，它还率先支持 MCP（Model Context Protocol）协议，增强了上下文处理能力。针对企业用户，NextChat 提供专业版解决方案，具备品牌定制、细粒度权限控制、内部知识库整合及安全审计等功能，满足公司对数据隐私和个性化管理的高标准要求。",87618,"2026-04-05T07:20:52",[13,26],{"id":44,"name":45,"github_repo":46,"description_zh":47,"stars":48,"difficulty_score":23,"last_commit_at":49,"category_tags":50,"status":16},2268,"ML-For-Beginners","microsoft\u002FML-For-Beginners","ML-For-Beginners 是由微软推出的一套系统化机器学习入门课程，旨在帮助零基础用户轻松掌握经典机器学习知识。这套课程将学习路径规划为 12 周，包含 26 节精炼课程和 52 道配套测验，内容涵盖从基础概念到实际应用的完整流程，有效解决了初学者面对庞大知识体系时无从下手、缺乏结构化指导的痛点。\n\n无论是希望转型的开发者、需要补充算法背景的研究人员，还是对人工智能充满好奇的普通爱好者，都能从中受益。课程不仅提供了清晰的理论讲解，还强调动手实践，让用户在循序渐进中建立扎实的技能基础。其独特的亮点在于强大的多语言支持，通过自动化机制提供了包括简体中文在内的 50 多种语言版本，极大地降低了全球不同背景用户的学习门槛。此外，项目采用开源协作模式，社区活跃且内容持续更新，确保学习者能获取前沿且准确的技术资讯。如果你正寻找一条清晰、友好且专业的机器学习入门之路，ML-For-Beginners 将是理想的起点。",84991,"2026-04-05T10:45:23",[14,51,52,53,15,54,26,13,55],"数据工具","视频","插件","其他","音频",{"id":57,"name":58,"github_repo":59,"description_zh":60,"stars":61,"difficulty_score":10,"last_commit_at":62,"category_tags":63,"status":16},519,"PaddleOCR","PaddlePaddle\u002FPaddleOCR","PaddleOCR 是一款基于百度飞桨框架开发的高性能开源光学字符识别工具包。它的核心能力是将图片、PDF 等文档中的文字提取出来，转换成计算机可读取的结构化数据，让机器真正“看懂”图文内容。\n\n面对海量纸质或电子文档，PaddleOCR 解决了人工录入效率低、数字化成本高的问题。尤其在人工智能领域，它扮演着连接图像与大型语言模型（LLM）的桥梁角色，能将视觉信息直接转化为文本输入，助力智能问答、文档分析等应用场景落地。\n\nPaddleOCR 适合开发者、算法研究人员以及有文档自动化需求的普通用户。其技术优势十分明显：不仅支持全球 100 多种语言的识别，还能在 Windows、Linux、macOS 等多个系统上运行，并灵活适配 CPU、GPU、NPU 等各类硬件。作为一个轻量级且社区活跃的开源项目，PaddleOCR 既能满足快速集成的需求，也能支撑前沿的视觉语言研究，是处理文字识别任务的理想选择。",74913,"2026-04-05T10:44:17",[26,14,13,54],{"id":65,"github_repo":66,"name":67,"description_en":68,"description_zh":69,"ai_summary_zh":69,"readme_en":70,"readme_zh":71,"quickstart_zh":72,"use_case_zh":73,"hero_image_url":74,"owner_login":75,"owner_name":76,"owner_avatar_url":77,"owner_bio":78,"owner_company":79,"owner_location":79,"owner_email":79,"owner_twitter":80,"owner_website":79,"owner_url":81,"languages":82,"stars":121,"forks":122,"last_commit_at":123,"license":124,"difficulty_score":10,"env_os":125,"env_gpu":126,"env_ram":127,"env_deps":128,"category_tags":135,"github_topics":136,"view_count":157,"oss_zip_url":79,"oss_zip_packed_at":79,"status":16,"created_at":158,"updated_at":159,"faqs":160,"releases":190},3128,"infiniflow\u002Fragflow","ragflow","RAGFlow is a leading open-source Retrieval-Augmented Generation (RAG) engine that fuses cutting-edge RAG with Agent capabilities to create a superior context layer for LLMs","RAGFlow 是一款领先的开源检索增强生成（RAG）引擎，旨在为大语言模型构建更精准、可靠的上下文层。它巧妙地将前沿的 RAG 技术与智能体（Agent）能力相结合，不仅支持从各类文档中高效提取知识，还能让模型基于这些知识进行逻辑推理和任务执行。\n\n在大模型应用中，幻觉问题和知识滞后是常见痛点。RAGFlow 通过深度解析复杂文档结构（如表格、图表及混合排版），显著提升了信息检索的准确度，从而有效减少模型“胡编乱造”的现象，确保回答既有据可依又具备时效性。其内置的智能体机制更进一步，使系统不仅能回答问题，还能自主规划步骤解决复杂问题。\n\n这款工具特别适合开发者、企业技术团队以及 AI 研究人员使用。无论是希望快速搭建私有知识库问答系统，还是致力于探索大模型在垂直领域落地的创新者，都能从中受益。RAGFlow 提供了可视化的工作流编排界面和灵活的 API 接口，既降低了非算法背景用户的上手门槛，也满足了专业开发者对系统深度定制的需求。作为基于 Apache 2.0 协议开源的项目，它正成为连接通用大模型与行业专有知识之间的重要桥梁。","\u003Cdiv align=\"center\">\n\u003Ca href=\"https:\u002F\u002Fcloud.ragflow.io\u002F\">\n\u003Cimg src=\"web\u002Fsrc\u002Fassets\u002Flogo-with-text.svg\" width=\"520\" alt=\"ragflow logo\">\n\u003C\u002Fa>\n\u003C\u002Fdiv>\n\n\u003Cp align=\"center\">\n  \u003Ca href=\".\u002FREADME.md\">\u003Cimg alt=\"README in English\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FEnglish-DBEDFA\">\u003C\u002Fa>\n  \u003Ca href=\".\u002FREADME_zh.md\">\u003Cimg alt=\"简体中文版自述文件\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F简体中文-DFE0E5\">\u003C\u002Fa>\n  \u003Ca href=\".\u002FREADME_tzh.md\">\u003Cimg alt=\"繁體版中文自述文件\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F繁體中文-DFE0E5\">\u003C\u002Fa>\n  \u003Ca href=\".\u002FREADME_ja.md\">\u003Cimg alt=\"日本語のREADME\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F日本語-DFE0E5\">\u003C\u002Fa>\n  \u003Ca href=\".\u002FREADME_ko.md\">\u003Cimg alt=\"한국어\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F한국어-DFE0E5\">\u003C\u002Fa>\n  \u003Ca href=\".\u002FREADME_id.md\">\u003Cimg alt=\"Bahasa Indonesia\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FBahasa Indonesia-DFE0E5\">\u003C\u002Fa>\n  \u003Ca href=\".\u002FREADME_pt_br.md\">\u003Cimg alt=\"Português(Brasil)\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPortuguês(Brasil)-DFE0E5\">\u003C\u002Fa>\n  \u003Ca href=\".\u002FREADME_fr.md\">\u003Cimg alt=\"README en Français\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FFrançais-DFE0E5\">\u003C\u002Fa>\n  \u003Ca href=\".\u002FREADME_ar.md\">\u003Cimg alt=\"README in Arabic\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FArabic-DFE0E5\">\u003C\u002Fa>\n  \u003Ca href=\".\u002FREADME_tr.md\">\u003Cimg alt=\"Türkçe README\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FTürkçe-DFE0E5\">\u003C\u002Fa>\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n    \u003Ca href=\"https:\u002F\u002Fx.com\u002Fintent\u002Ffollow?screen_name=infiniflowai\" target=\"_blank\">\n        \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Ftwitter\u002Ffollow\u002Finfiniflow?logo=X&color=%20%23f5f5f5\" alt=\"follow on X(Twitter)\">\n    \u003C\u002Fa>\n    \u003Ca href=\"https:\u002F\u002Fcloud.ragflow.io\" target=\"_blank\">\n        \u003Cimg alt=\"Static Badge\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FOnline-Demo-4e6b99\">\n    \u003C\u002Fa>\n    \u003Ca href=\"https:\u002F\u002Fhub.docker.com\u002Fr\u002Finfiniflow\u002Fragflow\" target=\"_blank\">\n        \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fdocker\u002Fpulls\u002Finfiniflow\u002Fragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square\" alt=\"docker pull infiniflow\u002Fragflow:v0.24.0\">\n    \u003C\u002Fa>\n    \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Freleases\u002Flatest\">\n        \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fv\u002Frelease\u002Finfiniflow\u002Fragflow?color=blue&label=Latest%20Release\" alt=\"Latest Release\">\n    \u003C\u002Fa>\n    \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fblob\u002Fmain\u002FLICENSE\">\n        \u003Cimg height=\"21\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-Apache--2.0-ffffff?labelColor=d4eaf7&color=2e6cc4\" alt=\"license\">\n    \u003C\u002Fa>\n    \u003Ca href=\"https:\u002F\u002Fdeepwiki.com\u002Finfiniflow\u002Fragflow\">\n        \u003Cimg alt=\"Ask DeepWiki\" src=\"https:\u002F\u002Fdeepwiki.com\u002Fbadge.svg\">\n    \u003C\u002Fa>\n\u003C\u002Fp>\n\n\u003Ch4 align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fragflow.io\u002Fdocs\u002Fdev\u002F\">Document\u003C\u002Fa> |\n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fissues\u002F12241\">Roadmap\u003C\u002Fa> |\n  \u003Ca href=\"https:\u002F\u002Ftwitter.com\u002Finfiniflowai\">Twitter\u003C\u002Fa> |\n  \u003Ca href=\"https:\u002F\u002Fdiscord.gg\u002FNjYzJD3GM3\">Discord\u003C\u002Fa> |\n  \u003Ca href=\"https:\u002F\u002Fcloud.ragflow.io\">Demo\u003C\u002Fa>\n\u003C\u002Fh4>\n\n\u003Cdiv align=\"center\" style=\"margin-top:20px;margin-bottom:20px;\">\n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Finfiniflow_ragflow_readme_23cfcd2b0d9e.png\" width=\"1200\"\u002F>\n\u003C\u002Fdiv>\n\n\u003Cdiv align=\"center\">\n\u003Ca href=\"https:\u002F\u002Ftrendshift.io\u002Frepositories\u002F9064\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Finfiniflow_ragflow_readme_4a68feb902da.png\" alt=\"infiniflow%2Fragflow | Trendshift\" style=\"width: 250px; height: 55px;\" width=\"250\" height=\"55\"\u002F>\u003C\u002Fa>\n\u003C\u002Fdiv>\n\n\u003Cdetails open>\n\u003Csummary>\u003Cb>📕 Table of Contents\u003C\u002Fb>\u003C\u002Fsummary>\n\n- 💡 [What is RAGFlow?](#-what-is-ragflow)\n- 🎮 [Demo](#-demo)\n- 📌 [Latest Updates](#-latest-updates)\n- 🌟 [Key Features](#-key-features)\n- 🔎 [System Architecture](#-system-architecture)\n- 🎬 [Get Started](#-get-started)\n- 🔧 [Configurations](#-configurations)\n- 🔧 [Build a Docker image](#-build-a-docker-image)\n- 🔨 [Launch service from source for development](#-launch-service-from-source-for-development)\n- 📚 [Documentation](#-documentation)\n- 📜 [Roadmap](#-roadmap)\n- 🏄 [Community](#-community)\n- 🙌 [Contributing](#-contributing)\n\n\u003C\u002Fdetails>\n\n## 💡 What is RAGFlow?\n\n[RAGFlow](https:\u002F\u002Fragflow.io\u002F) is a leading open-source Retrieval-Augmented Generation ([RAG](https:\u002F\u002Fragflow.io\u002Fbasics\u002Fwhat-is-rag)) engine that fuses cutting-edge RAG with Agent capabilities to create a superior context layer for LLMs. It offers a streamlined RAG workflow adaptable to enterprises of any scale. Powered by a converged [context engine](https:\u002F\u002Fragflow.io\u002Fbasics\u002Fwhat-is-agent-context-engine) and pre-built agent templates, RAGFlow enables developers to transform complex data into high-fidelity, production-ready AI systems with exceptional efficiency and precision.\n\n## 🎮 Demo\n\nTry our demo at [https:\u002F\u002Fcloud.ragflow.io](https:\u002F\u002Fcloud.ragflow.io).\n\n\u003Cdiv align=\"center\" style=\"margin-top:20px;margin-bottom:20px;\">\n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Finfiniflow_ragflow_readme_6d909d0a8100.gif\" width=\"1200\"\u002F>\n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Finfiniflow_ragflow_readme_b1f0bc294b34.gif\" width=\"1200\"\u002F>\n\u003C\u002Fdiv>\n\n## 🔥 Latest Updates\n\n- 2026-03-24 [RAGFlow Skill on OpenClaw](https:\u002F\u002Fclawhub.ai\u002Fyingfeng\u002Fragflow-skill) — Provides an official skill to access RAGFlow datasets via OpenClaw.\n- 2025-12-26 Supports 'Memory' for AI agent.\n- 2025-11-19 Supports Gemini 3 Pro.\n- 2025-11-12 Supports data synchronization from Confluence, S3, Notion, Discord, Google Drive.\n- 2025-10-23 Supports MinerU & Docling as document parsing methods.\n- 2025-10-15 Supports orchestrable ingestion pipeline.\n- 2025-08-08 Supports OpenAI's latest GPT-5 series models.\n- 2025-08-01 Supports agentic workflow and MCP.\n- 2025-05-23 Adds a Python\u002FJavaScript code executor component to Agent.\n- 2025-05-05 Supports cross-language query.\n- 2025-03-19 Supports using a multi-modal model to make sense of images within PDF or DOCX files.\n\n## 🎉 Stay Tuned\n\n⭐️ Star our repository to stay up-to-date with exciting new features and improvements! Get instant notifications for new\nreleases! 🌟\n\n\u003Cdiv align=\"center\" style=\"margin-top:20px;margin-bottom:20px;\">\n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Finfiniflow_ragflow_readme_83991a425612.png\" width=\"1200\"\u002F>\n\u003C\u002Fdiv>\n\n## 🌟 Key Features\n\n### 🍭 **\"Quality in, quality out\"**\n\n- [Deep document understanding](.\u002Fdeepdoc\u002FREADME.md)-based knowledge extraction from unstructured data with complicated\n  formats.\n- Finds \"needle in a data haystack\" of literally unlimited tokens.\n\n### 🍱 **Template-based chunking**\n\n- Intelligent and explainable.\n- Plenty of template options to choose from.\n\n### 🌱 **Grounded citations with reduced hallucinations**\n\n- Visualization of text chunking to allow human intervention.\n- Quick view of the key references and traceable citations to support grounded answers.\n\n### 🍔 **Compatibility with heterogeneous data sources**\n\n- Supports Word, slides, excel, txt, images, scanned copies, structured data, web pages, and more.\n\n### 🛀 **Automated and effortless RAG workflow**\n\n- Streamlined RAG orchestration catered to both personal and large businesses.\n- Configurable LLMs as well as embedding models.\n- Multiple recall paired with fused re-ranking.\n- Intuitive APIs for seamless integration with business.\n\n## 🔎 System Architecture\n\n\u003Cdiv align=\"center\" style=\"margin-top:20px;margin-bottom:20px;\">\n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Finfiniflow_ragflow_readme_b47b400b5037.png\" width=\"1000\"\u002F>\n\u003C\u002Fdiv>\n\n## 🎬 Get Started\n\n### 📝 Prerequisites\n\n- CPU >= 4 cores\n- RAM >= 16 GB\n- Disk >= 50 GB\n- Docker >= 24.0.0 & Docker Compose >= v2.26.1\n- [gVisor](https:\u002F\u002Fgvisor.dev\u002Fdocs\u002Fuser_guide\u002Finstall\u002F): Required only if you intend to use the code executor (sandbox) feature of RAGFlow.\n\n> [!TIP]\n> If you have not installed Docker on your local machine (Windows, Mac, or Linux), see [Install Docker Engine](https:\u002F\u002Fdocs.docker.com\u002Fengine\u002Finstall\u002F).\n\n### 🚀 Start up the server\n\n1. Ensure `vm.max_map_count` >= 262144:\n\n   > To check the value of `vm.max_map_count`:\n   >\n   > ```bash\n   > $ sysctl vm.max_map_count\n   > ```\n   >\n   > Reset `vm.max_map_count` to a value at least 262144 if it is not.\n   >\n   > ```bash\n   > # In this case, we set it to 262144:\n   > $ sudo sysctl -w vm.max_map_count=262144\n   > ```\n   >\n   > This change will be reset after a system reboot. To ensure your change remains permanent, add or update the\n   > `vm.max_map_count` value in **\u002Fetc\u002Fsysctl.conf** accordingly:\n   >\n   > ```bash\n   > vm.max_map_count=262144\n   > ```\n   >\n2. Clone the repo:\n\n   ```bash\n   $ git clone https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow.git\n   ```\n3. Start up the server using the pre-built Docker images:\n\n> [!CAUTION]\n> All Docker images are built for x86 platforms. We don't currently offer Docker images for ARM64.\n> If you are on an ARM64 platform, follow [this guide](https:\u002F\u002Fragflow.io\u002Fdocs\u002Fdev\u002Fbuild_docker_image) to build a Docker image compatible with your system.\n\n> The command below downloads the `v0.24.0` edition of the RAGFlow Docker image. See the following table for descriptions of different RAGFlow editions. To download a RAGFlow edition different from `v0.24.0`, update the `RAGFLOW_IMAGE` variable accordingly in **docker\u002F.env** before using `docker compose` to start the server.\n\n```bash\n   $ cd ragflow\u002Fdocker\n\n   # git checkout v0.24.0\n   # Optional: use a stable tag (see releases: https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Freleases)\n   # This step ensures the **entrypoint.sh** file in the code matches the Docker image version.\n\n   # Use CPU for DeepDoc tasks:\n   $ docker compose -f docker-compose.yml up -d\n\n   # To use GPU to accelerate DeepDoc tasks:\n   # sed -i '1i DEVICE=gpu' .env\n   # docker compose -f docker-compose.yml up -d\n```\n\n> Note: Prior to `v0.22.0`, we provided both images with embedding models and slim images without embedding models. Details as follows:\n\n| RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?        |\n|-------------------|-----------------|-----------------------|----------------|\n| v0.21.1           | ≈9       | ✔️                    | Stable release |\n| v0.21.1-slim      | ≈2       | ❌                     | Stable release |\n\n> Starting with `v0.22.0`, we ship only the slim edition and no longer append the **-slim** suffix to the image tag.\n\n4. Check the server status after having the server up and running:\n\n   ```bash\n   $ docker logs -f docker-ragflow-cpu-1\n   ```\n\n   _The following output confirms a successful launch of the system:_\n\n   ```bash\n\n         ____   ___    ______ ______ __\n        \u002F __ \\ \u002F   |  \u002F ____\u002F\u002F ____\u002F\u002F \u002F____  _      __\n       \u002F \u002F_\u002F \u002F\u002F \u002F| | \u002F \u002F __ \u002F \u002F_   \u002F \u002F\u002F __ \\| | \u002F| \u002F \u002F\n      \u002F _, _\u002F\u002F ___ |\u002F \u002F_\u002F \u002F\u002F __\u002F  \u002F \u002F\u002F \u002F_\u002F \u002F| |\u002F |\u002F \u002F\n     \u002F_\u002F |_|\u002F_\u002F  |_|\\____\u002F\u002F_\u002F    \u002F_\u002F \\____\u002F |__\u002F|__\u002F\n\n    * Running on all addresses (0.0.0.0)\n   ```\n\n   > If you skip this confirmation step and directly log in to RAGFlow, your browser may prompt a `network abnormal`\n   > error because, at that moment, your RAGFlow may not be fully initialized.\n   >\n5. In your web browser, enter the IP address of your server and log in to RAGFlow.\n\n   > With the default settings, you only need to enter `http:\u002F\u002FIP_OF_YOUR_MACHINE` (**sans** port number) as the default\n   > HTTP serving port `80` can be omitted when using the default configurations.\n   >\n6. In [service_conf.yaml.template](.\u002Fdocker\u002Fservice_conf.yaml.template), select the desired LLM factory in `user_default_llm` and update\n   the `API_KEY` field with the corresponding API key.\n\n   > See [llm_api_key_setup](https:\u002F\u002Fragflow.io\u002Fdocs\u002Fdev\u002Fllm_api_key_setup) for more information.\n   >\n\n   _The show is on!_\n\n## 🔧 Configurations\n\nWhen it comes to system configurations, you will need to manage the following files:\n\n- [.env](.\u002Fdocker\u002F.env): Keeps the fundamental setups for the system, such as `SVR_HTTP_PORT`, `MYSQL_PASSWORD`, and\n  `MINIO_PASSWORD`.\n- [service_conf.yaml.template](.\u002Fdocker\u002Fservice_conf.yaml.template): Configures the back-end services. The environment variables in this file will be automatically populated when the Docker container starts. Any environment variables set within the Docker container will be available for use, allowing you to customize service behavior based on the deployment environment.\n- [docker-compose.yml](.\u002Fdocker\u002Fdocker-compose.yml): The system relies on [docker-compose.yml](.\u002Fdocker\u002Fdocker-compose.yml) to start up.\n\n> The [.\u002Fdocker\u002FREADME](.\u002Fdocker\u002FREADME.md) file provides a detailed description of the environment settings and service\n> configurations which can be used as `${ENV_VARS}` in the [service_conf.yaml.template](.\u002Fdocker\u002Fservice_conf.yaml.template) file.\n\nTo update the default HTTP serving port (80), go to [docker-compose.yml](.\u002Fdocker\u002Fdocker-compose.yml) and change `80:80`\nto `\u003CYOUR_SERVING_PORT>:80`.\n\nUpdates to the above configurations require a reboot of all containers to take effect:\n\n> ```bash\n> $ docker compose -f docker-compose.yml up -d\n> ```\n\n### Switch doc engine from Elasticsearch to Infinity\n\nRAGFlow uses Elasticsearch by default for storing full text and vectors. To switch to [Infinity](https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Finfinity\u002F), follow these steps:\n\n1. Stop all running containers:\n\n   ```bash\n   $ docker compose -f docker\u002Fdocker-compose.yml down -v\n   ```\n\n> [!WARNING]\n> `-v` will delete the docker container volumes, and the existing data will be cleared.\n\n2. Set `DOC_ENGINE` in **docker\u002F.env** to `infinity`.\n3. Start the containers:\n\n   ```bash\n   $ docker compose -f docker-compose.yml up -d\n   ```\n\n> [!WARNING]\n> Switching to Infinity on a Linux\u002Farm64 machine is not yet officially supported.\n\n## 🔧 Build a Docker image\n\nThis image is approximately 2 GB in size and relies on external LLM and embedding services.\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow.git\ncd ragflow\u002F\ndocker build --platform linux\u002Famd64 -f Dockerfile -t infiniflow\u002Fragflow:nightly .\n```\n\nOr if you are behind a proxy, you can pass proxy arguments:\n\n```bash\ndocker build --platform linux\u002Famd64 \\\n  --build-arg http_proxy=http:\u002F\u002FYOUR_PROXY:PORT \\\n  --build-arg https_proxy=http:\u002F\u002FYOUR_PROXY:PORT \\\n  -f Dockerfile -t infiniflow\u002Fragflow:nightly .\n```\n\n## 🔨 Launch service from source for development\n\n1. Install `uv` and `pre-commit`, or skip this step if they are already installed:\n\n   ```bash\n   pipx install uv pre-commit\n   ```\n2. Clone the source code and install Python dependencies:\n\n   ```bash\n   git clone https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow.git\n   cd ragflow\u002F\n   uv sync --python 3.12 # install RAGFlow dependent python modules\n   uv run download_deps.py\n   pre-commit install\n   ```\n3. Launch the dependent services (MinIO, Elasticsearch, Redis, and MySQL) using Docker Compose:\n\n   ```bash\n   docker compose -f docker\u002Fdocker-compose-base.yml up -d\n   ```\n\n   Add the following line to `\u002Fetc\u002Fhosts` to resolve all hosts specified in **docker\u002F.env** to `127.0.0.1`:\n\n   ```\n   127.0.0.1       es01 infinity mysql minio redis sandbox-executor-manager\n   ```\n4. If you cannot access HuggingFace, set the `HF_ENDPOINT` environment variable to use a mirror site:\n\n   ```bash\n   export HF_ENDPOINT=https:\u002F\u002Fhf-mirror.com\n   ```\n5. If your operating system does not have jemalloc, please install it as follows:\n\n   ```bash\n   # Ubuntu\n   sudo apt-get install libjemalloc-dev\n   # CentOS\n   sudo yum install jemalloc\n   # OpenSUSE\n   sudo zypper install jemalloc\n   # macOS\n   sudo brew install jemalloc\n   ```\n6. Launch backend service:\n\n   ```bash\n   source .venv\u002Fbin\u002Factivate\n   export PYTHONPATH=$(pwd)\n   bash docker\u002Flaunch_backend_service.sh\n   ```\n7. Install frontend dependencies:\n\n   ```bash\n   cd web\n   npm install\n   ```\n8. Launch frontend service:\n\n   ```bash\n   npm run dev\n   ```\n\n   _The following output confirms a successful launch of the system:_\n\n   ![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Finfiniflow_ragflow_readme_ec85af2f698f.png)\n9. Stop RAGFlow front-end and back-end service after development is complete:\n\n   ```bash\n   pkill -f \"ragflow_server.py|task_executor.py\"\n   ```\n\n## 📚 Documentation\n\n- [Quickstart](https:\u002F\u002Fragflow.io\u002Fdocs\u002Fdev\u002F)\n- [Configuration](https:\u002F\u002Fragflow.io\u002Fdocs\u002Fdev\u002Fconfigurations)\n- [Release notes](https:\u002F\u002Fragflow.io\u002Fdocs\u002Fdev\u002Frelease_notes)\n- [User guides](https:\u002F\u002Fragflow.io\u002Fdocs\u002Fcategory\u002Fuser-guides)\n- [Developer guides](https:\u002F\u002Fragflow.io\u002Fdocs\u002Fcategory\u002Fdeveloper-guides)\n- [References](https:\u002F\u002Fragflow.io\u002Fdocs\u002Fdev\u002Fcategory\u002Freferences)\n- [FAQs](https:\u002F\u002Fragflow.io\u002Fdocs\u002Fdev\u002Ffaq)\n\n## 📜 Roadmap\n\nSee the [RAGFlow Roadmap 2026](https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fissues\u002F12241)\n\n## 🏄 Community\n\n- [Discord](https:\u002F\u002Fdiscord.gg\u002FNjYzJD3GM3)\n- [Twitter](https:\u002F\u002Ftwitter.com\u002Finfiniflowai)\n- [GitHub Discussions](https:\u002F\u002Fgithub.com\u002Forgs\u002Finfiniflow\u002Fdiscussions)\n\n## 🙌 Contributing\n\nRAGFlow flourishes via open-source collaboration. In this spirit, we embrace diverse contributions from the community.\nIf you would like to be a part, review our [Contribution Guidelines](https:\u002F\u002Fragflow.io\u002Fdocs\u002Fdev\u002Fcontributing) first.\n","\u003Cdiv align=\"center\">\n\u003Ca href=\"https:\u002F\u002Fcloud.ragflow.io\u002F\">\n\u003Cimg src=\"web\u002Fsrc\u002Fassets\u002Flogo-with-text.svg\" width=\"520\" alt=\"ragflow logo\">\n\u003C\u002Fa>\n\u003C\u002Fdiv>\n\n\u003Cp align=\"center\">\n  \u003Ca href=\".\u002FREADME.md\">\u003Cimg alt=\"英文自述文件\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FEnglish-DBEDFA\">\u003C\u002Fa>\n  \u003Ca href=\".\u002FREADME_zh.md\">\u003Cimg alt=\"简体中文版自述文件\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F简体中文-DFE0E5\">\u003C\u002Fa>\n  \u003Ca href=\".\u002FREADME_tzh.md\">\u003Cimg alt=\"繁體版中文自述文件\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F繁體中文-DFE0E5\">\u003C\u002Fa>\n  \u003Ca href=\".\u002FREADME_ja.md\">\u003Cimg alt=\"日本語のREADME\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F日本語-DFE0E5\">\u003C\u002Fa>\n  \u003Ca href=\".\u002FREADME_ko.md\">\u003Cimg alt=\"한국어\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F한국어-DFE0E5\">\u003C\u002Fa>\n  \u003Ca href=\".\u002FREADME_id.md\">\u003Cimg alt=\"Bahasa Indonesia\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FBahasa Indonesia-DFE0E5\">\u003C\u002Fa>\n  \u003Ca href=\".\u002FREADME_pt_br.md\">\u003Cimg alt=\"Português(Brasil)\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPortuguês(Brasil)-DFE0E5\">\u003C\u002Fa>\n  \u003Ca href=\".\u002FREADME_fr.md\">\u003Cimg alt=\"README en Français\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FFrançais-DFE0E5\">\u003C\u002Fa>\n  \u003Ca href=\".\u002FREADME_ar.md\">\u003Cimg alt=\"README in Arabic\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FArabic-DFE0E5\">\u003C\u002Fa>\n  \u003Ca href=\".\u002FREADME_tr.md\">\u003Cimg alt=\"Türkçe README\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FTürkçe-DFE0E5\">\u003C\u002Fa>\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n    \u003Ca href=\"https:\u002F\u002Fx.com\u002Fintent\u002Ffollow?screen_name=infiniflowai\" target=\"_blank\">\n        \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Ftwitter\u002Ffollow\u002Finfiniflow?logo=X&color=%20%23f5f5f5\" alt=\"关注X(Twitter)\">\n    \u003C\u002Fa>\n    \u003Ca href=\"https:\u002F\u002Fcloud.ragflow.io\" target=\"_blank\">\n        \u003Cimg alt=\"静态徽章\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F在线演示-4e6b99\">\n    \u003C\u002Fa>\n    \u003Ca href=\"https:\u002F\u002Fhub.docker.com\u002Fr\u002Finfiniflow\u002Fragflow\" target=\"_blank\">\n        \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fdocker\u002Fpulls\u002Finfiniflow\u002Fragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square\" alt=\"docker pull infiniflow\u002Fragflow:v0.24.0\">\n    \u003C\u002Fa>\n    \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Freleases\u002Flatest\">\n        \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fv\u002Frelease\u002Finfiniflow\u002Fragflow?color=blue&label=最新版本\" alt=\"最新版本\">\n    \u003C\u002Fa>\n    \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fblob\u002Fmain\u002FLICENSE\">\n        \u003Cimg height=\"21\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F许可证-Apache--2.0-ffffff?labelColor=d4eaf7&color=2e6cc4\" alt=\"许可证\">\n    \u003C\u002Fa>\n    \u003Ca href=\"https:\u002F\u002Fdeepwiki.com\u002Finfiniflow\u002Fragflow\">\n        \u003Cimg alt=\"Ask DeepWiki\" src=\"https:\u002F\u002Fdeepwiki.com\u002Fbadge.svg\">\n    \u003C\u002Fa>\n\u003C\u002Fp>\n\n\u003Ch4 align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fragflow.io\u002Fdocs\u002Fdev\u002F\">文档\u003C\u002Fa> |\n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fissues\u002F12241\">路线图\u003C\u002Fa> |\n  \u003Ca href=\"https:\u002F\u002Ftwitter.com\u002Finfiniflowai\">Twitter\u003C\u002Fa> |\n  \u003Ca href=\"https:\u002F\u002Fdiscord.gg\u002FNjYzJD3GM3\">Discord\u003C\u002Fa> |\n  \u003Ca href=\"https:\u002F\u002Fcloud.ragflow.io\">演示\u003C\u002Fa>\n\u003C\u002Fh4>\n\n\u003Cdiv align=\"center\" style=\"margin-top:20px;margin-bottom:20px;\">\n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Finfiniflow_ragflow_readme_23cfcd2b0d9e.png\" width=\"1200\"\u002F>\n\u003C\u002Fdiv>\n\n\u003Cdiv align=\"center\">\n\u003Ca href=\"https:\u002F\u002Ftrendshift.io\u002Frepositories\u002F9064\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Finfiniflow_ragflow_readme_4a68feb902da.png\" alt=\"infiniflow%2Fragflow | Trendshift\" style=\"width: 250px; height: 55px;\" width=\"250\" height=\"55\"\u002F>\u003C\u002Fa>\n\u003C\u002Fdiv>\n\n\u003Cdetails open>\n\u003Csummary>\u003Cb>📕 目录\u003C\u002Fb>\u003C\u002Fsummary>\n\n- 💡 [什么是RAGFlow？](#-what-is-ragflow)\n- 🎮 [演示](#-demo)\n- 📌 [最新动态](#-latest-updates)\n- 🌟 [核心功能](#-key-features)\n- 🔎 [系统架构](#-system-architecture)\n- 🎬 [快速入门](#-get-started)\n- 🔧 [配置](#-configurations)\n- 🔧 [构建Docker镜像](#-build-a-docker-image)\n- 🔨 [从源码启动服务进行开发](#-launch-service-from-source-for-development)\n- 📚 [文档](#-documentation)\n- 📜 [路线图](#-roadmap)\n- 🏄 [社区](#-community)\n- 🙌 [贡献](#-contributing)\n\n\u003C\u002Fdetails>\n\n## 💡 什么是RAGFlow？\n\n[RAGFlow](https:\u002F\u002Fragflow.io\u002F) 是一款领先的开源检索增强生成（[RAG](https:\u002F\u002Fragflow.io\u002Fbasics\u002Fwhat-is-rag))引擎，它将前沿的RAG技术与智能体能力相结合，为大语言模型打造卓越的上下文层。RAGFlow提供了一套可扩展的RAG工作流，适用于各种规模的企业。凭借融合的[上下文引擎](https:\u002F\u002Fragflow.io\u002Fbasics\u002Fwhat-is-agent-context-engine)和预构建的智能体模板，RAGFlow使开发者能够以极高的效率和精准度，将复杂数据转化为高保真、可投入生产的AI系统。\n\n## 🎮 演示\n\n请访问 [https:\u002F\u002Fcloud.ragflow.io](https:\u002F\u002Fcloud.ragflow.io) 体验我们的演示。\n\n\u003Cdiv align=\"center\" style=\"margin-top:20px;margin-bottom:20px;\">\n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Finfiniflow_ragflow_readme_6d909d0a8100.gif\" width=\"1200\"\u002F>\n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Finfiniflow_ragflow_readme_b1f0bc294b34.gif\" width=\"1200\"\u002F>\n\u003C\u002Fdiv>\n\n## 🔥 最新动态\n\n- 2026年3月24日 [RAGFlow技能上线OpenClaw](https:\u002F\u002Fclawhub.ai\u002Fyingfeng\u002Fragflow-skill) — 提供官方技能，可通过OpenClaw访问RAGFlow数据集。\n- 2025年12月26日 支持AI智能体的“记忆”功能。\n- 2025年11月19日 支持Gemini 3 Pro。\n- 2025年11月12日 支持从Confluence、S3、Notion、Discord、Google Drive同步数据。\n- 2025年10月23日 支持MinerU和Docling作为文档解析方法。\n- 2025年10月15日 支持可编排的数据摄取管道。\n- 2025年8月8日 支持OpenAI最新的GPT-5系列模型。\n- 2025年8月1日 支持智能体工作流和MCP。\n- 2025年5月23日 为智能体新增Python\u002FJavaScript代码执行组件。\n- 2025年5月5日 支持跨语言查询。\n- 2025年3月19日 支持使用多模态模型理解PDF或DOCX文件中的图像。\n\n## 🎉 敬请期待\n\n⭐️ 请给我们的仓库点个星，以便及时获取激动人心的新功能和改进！您将立即收到新版本发布的通知！🌟\n\n\u003Cdiv align=\"center\" style=\"margin-top:20px;margin-bottom:20px;\">\n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Finfiniflow_ragflow_readme_83991a425612.png\" width=\"1200\"\u002F>\n\u003C\u002Fdiv>\n\n## 🌟 核心功能\n\n### 🍭 **“高质量输入，高质量输出”**\n\n- 基于[深度文档理解](.\u002Fdeepdoc\u002FREADME.md)，从格式复杂的非结构化数据中提取知识。\n- 能够在字数几乎不受限制的数据中精准定位所需信息。\n\n### 🍱 **基于模板的分块处理**\n\n- 智能且可解释。\n- 提供多种模板选项供选择。\n\n### 🌱 **基于证据的引用，减少幻觉**\n\n- 文本分块可视化，便于人工干预。\n- 快速查看关键参考文献，并可追溯引用来源，确保答案有据可依。\n\n### 🍔 **兼容多种异构数据源**\n\n- 支持Word、PPT、Excel、TXT、图片、扫描件、结构化数据、网页等多种格式。\n\n### 🛀 **自动化且无负担的 RAG 工作流**\n\n- 精简高效的 RAG 编排，适用于个人与大型企业。\n- 可配置的语言模型及嵌入模型。\n- 多种召回策略结合融合重排序。\n- 直观的 API，便于与业务系统无缝集成。\n\n## 🔎 系统架构\n\n\u003Cdiv align=\"center\" style=\"margin-top:20px;margin-bottom:20px;\">\n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Finfiniflow_ragflow_readme_b47b400b5037.png\" width=\"1000\"\u002F>\n\u003C\u002Fdiv>\n\n## 🎬 开始使用\n\n### 📝 前置条件\n\n- CPU ≥ 4 核\n- 内存 ≥ 16 GB\n- 磁盘空间 ≥ 50 GB\n- Docker ≥ 24.0.0 & Docker Compose ≥ v2.26.1\n- [gVisor](https:\u002F\u002Fgvisor.dev\u002Fdocs\u002Fuser_guide\u002Finstall\u002F)：仅在您计划使用 RAGFlow 的代码执行器（沙箱）功能时需要安装。\n\n> [!TIP]\n> 如果您的本地机器尚未安装 Docker（Windows、Mac 或 Linux），请参阅 [安装 Docker Engine](https:\u002F\u002Fdocs.docker.com\u002Fengine\u002Finstall\u002F)。\n\n### 🚀 启动服务器\n\n1. 确保 `vm.max_map_count` ≥ 262144：\n\n   > 检查 `vm.max_map_count` 的值：\n   >\n   > ```bash\n   > $ sysctl vm.max_map_count\n   > ```\n   >\n   > 如果该值不足，则将其重置为至少 262144：\n   >\n   > ```bash\n   > # 在此例中，我们将其设置为 262144：\n   > $ sudo sysctl -w vm.max_map_count=262144\n   > ```\n   >\n   > 此更改在系统重启后会失效。为使更改永久生效，请在 **\u002Fetc\u002Fsysctl.conf** 中添加或更新以下内容：\n   >\n   > ```bash\n   > vm.max_map_count=262144\n   > ```\n   >\n2. 克隆仓库：\n\n   ```bash\n   $ git clone https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow.git\n   ```\n3. 使用预构建的 Docker 镜像启动服务器：\n\n> [!CAUTION]\n> 所有 Docker 镜像均基于 x86 平台构建。目前我们尚未提供 ARM64 平台的 Docker 镜像。\n> 如果您使用的是 ARM64 平台，请按照 [此指南](https:\u002F\u002Fragflow.io\u002Fdocs\u002Fdev\u002Fbuild_docker_image) 构建与您的系统兼容的 Docker 镜像。\n\n> 下述命令将下载 RAGFlow 的 `v0.24.0` 版本镜像。不同版本的说明见下表。如需下载非 `v0.24.0` 版本的 RAGFlow，请在使用 `docker compose` 启动服务器前，相应地更新 **docker\u002F.env** 文件中的 `RAGFLOW_IMAGE` 变量。\n\n```bash\n   $ cd ragflow\u002Fdocker\n\n   # 切换至 v0.24.0 分支\n   # （可选：使用稳定标签，详见发布页面：https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Freleases）\n   # 此步骤确保代码中的 `entrypoint.sh` 文件与 Docker 镜像版本一致。\n\n   # 使用 CPU 处理 DeepDoc 任务：\n   $ docker compose -f docker-compose.yml up -d\n\n   # 如需使用 GPU 加速 DeepDoc 任务：\n   # sed -i '1i DEVICE=gpu' .env\n   # docker compose -f docker-compose.yml up -d\n```\n\n> 注意：在 `v0.22.0` 之前，我们同时提供了包含嵌入模型的完整镜像和不含嵌入模型的精简版镜像，具体如下：\n\n| RAGFlow 镜像标签 | 镜像大小 (GB) | 是否包含嵌入模型 | 是否稳定？ |\n|-------------------|-----------------|-----------------------|----------------|\n| v0.21.1           | ≈9       | ✔️                    | 稳定版         |\n| v0.21.1-slim      | ≈2       | ❌                     | 稳定版         |\n\n> 自 `v0.22.0` 起，我们仅提供精简版镜像，并取消了镜像标签中的 **-slim** 后缀。\n\n4. 服务器启动后，检查运行状态：\n\n   ```bash\n   $ docker logs -f docker-ragflow-cpu-1\n   ```\n\n   _以下输出表明系统已成功启动：_\n\n   ```bash\n\n         ____   ___    ______ ______ __\n        \u002F __ \\ \u002F   |  \u002F ____\u002F\u002F ____\u002F\u002F \u002F____  _      __\n       \u002F \u002F_\u002F \u002F\u002F \u002F| | \u002F \u002F __ \u002F \u002F_   \u002F \u002F\u002F __ \\| | \u002F| \u002F \u002F\n      \u002F _, _\u002F\u002F ___ |\u002F \u002F_\u002F \u002F\u002F __\u002F  \u002F \u002F\u002F \u002F_\u002F \u002F| |\u002F |\u002F \u002F\n     \u002F_\u002F |_|\u002F_\u002F  |_|\\____\u002F\u002F_\u002F    \u002F_\u002F \\____\u002F |__\u002F|__\u002F\n\n    * Running on all addresses (0.0.0.0)\n   ```\n\n   > 如果跳过此确认步骤直接登录 RAGFlow，浏览器可能会提示 `network abnormal` 错误，因为此时 RAGFlow 可能尚未完全初始化。\n   >\n5. 在您的浏览器中输入服务器的 IP 地址，登录 RAGFlow。\n\n   > 在默认配置下，您只需输入 `http:\u002F\u002FIP_OF_YOUR_MACHINE` 即可，因为默认的 HTTP 服务端口 `80` 在使用默认配置时可以省略。\n   >\n6. 在 [service_conf.yaml.template](.\u002Fdocker\u002Fservice_conf.yaml.template) 中，在 `user_default_llm` 中选择所需的语言模型工厂，并在 `API_KEY` 字段中填写相应的 API 密钥。\n\n   > 更多信息请参阅 [llm_api_key_setup](https:\u002F\u002Fragflow.io\u002Fdocs\u002Fdev\u002Fllm_api_key_setup)。\n   >\n\n   _演出正式开始！_\n\n## 🔧 配置说明\n\n在系统配置方面，您需要管理以下文件：\n\n- [.env](.\u002Fdocker\u002F.env)：保存系统的基础设置，例如 `SVR_HTTP_PORT`、`MYSQL_PASSWORD` 和 `MINIO_PASSWORD`。\n- [service_conf.yaml.template](.\u002Fdocker\u002Fservice_conf.yaml.template)：用于配置后端服务。该文件中的环境变量将在 Docker 容器启动时自动填充。容器内设置的任何环境变量均可使用，从而允许您根据部署环境自定义服务行为。\n- [docker-compose.yml](.\u002Fdocker\u002Fdocker-compose.yml)：系统依赖于 [docker-compose.yml](.\u002Fdocker\u002Fdocker-compose.yml) 来启动。\n\n> [.\u002Fdocker\u002FREADME](.\u002Fdocker\u002FREADME.md) 文件详细介绍了环境设置和服务配置，这些内容可用作 [service_conf.yaml.template](.\u002Fdocker\u002Fservice_conf.yaml.template) 文件中的 `${ENV_VARS}`。\n\n如需更改默认的 HTTP 服务端口（80），请前往 [docker-compose.yml](.\u002Fdocker\u002Fdocker-compose.yml) 将 `80:80` 修改为 `\u003CYOUR_SERVING_PORT>:80`。\n\n上述配置的更新需要重启所有容器才能生效：\n\n> ```bash\n> $ docker compose -f docker-compose.yml up -d\n> ```\n\n### 将文档引擎从 Elasticsearch 切换至 Infinity\n\nRAGFlow 默认使用 Elasticsearch 存储全文和向量数据。如需切换至 [Infinity](https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Finfinity\u002F)，请按以下步骤操作：\n\n1. 停止所有正在运行的容器：\n\n   ```bash\n   $ docker compose -f docker\u002Fdocker-compose.yml down -v\n   ```\n\n> [!WARNING]\n> `-v` 选项会删除 Docker 容器卷，现有数据将被清空。\n\n2. 将 **docker\u002F.env** 文件中的 `DOC_ENGINE` 设置为 `infinity`。\n3. 启动容器：\n\n   ```bash\n   $ docker compose -f docker-compose.yml up -d\n   ```\n\n> [!WARNING]\n> 目前尚未正式支持在 Linux\u002Farm64 机器上切换至 Infinity。\n\n## 🔧 构建 Docker 镜像\n\n该镜像大小约为 2 GB，依赖外部的 LLM 和嵌入服务。\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow.git\ncd ragflow\u002F\ndocker build --platform linux\u002Famd64 -f Dockerfile -t infiniflow\u002Fragflow:nightly .\n```\n\n或者，如果您在代理后面，可以传递代理参数：\n\n```bash\ndocker build --platform linux\u002Famd64 \\\n  --build-arg http_proxy=http:\u002F\u002FYOUR_PROXY:PORT \\\n  --build-arg https_proxy=http:\u002F\u002FYOUR_PROXY:PORT \\\n  -f Dockerfile -t infiniflow\u002Fragflow:nightly .\n```\n\n## 🔨 从源码启动服务以进行开发\n\n1. 安装 `uv` 和 `pre-commit`，如果已安装则可跳过此步骤：\n\n   ```bash\n   pipx install uv pre-commit\n   ```\n2. 克隆源代码并安装 Python 依赖项：\n\n   ```bash\n   git clone https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow.git\n   cd ragflow\u002F\n   uv sync --python 3.12 # 安装 RAGFlow 所需的 Python 模块\n   uv run download_deps.py\n   pre-commit install\n   ```\n3. 使用 Docker Compose 启动依赖的服务（MinIO、Elasticsearch、Redis 和 MySQL）：\n\n   ```bash\n   docker compose -f docker\u002Fdocker-compose-base.yml up -d\n   ```\n\n   将以下内容添加到 `\u002Fetc\u002Fhosts` 文件中，以便将 **docker\u002F.env** 中指定的所有主机解析为 `127.0.0.1`：\n\n   ```\n   127.0.0.1       es01 infinity mysql minio redis sandbox-executor-manager\n   ```\n4. 如果无法访问 HuggingFace，请设置 `HF_ENDPOINT` 环境变量以使用镜像站点：\n\n   ```bash\n   export HF_ENDPOINT=https:\u002F\u002Fhf-mirror.com\n   ```\n5. 如果您的操作系统没有 jemalloc，请按如下方式安装：\n\n   ```bash\n   # Ubuntu\n   sudo apt-get install libjemalloc-dev\n   # CentOS\n   sudo yum install jemalloc\n   # OpenSUSE\n   sudo zypper install jemalloc\n   # macOS\n   sudo brew install jemalloc\n   ```\n6. 启动后端服务：\n\n   ```bash\n   source .venv\u002Fbin\u002Factivate\n   export PYTHONPATH=$(pwd)\n   bash docker\u002Flaunch_backend_service.sh\n   ```\n7. 安装前端依赖项：\n\n   ```bash\n   cd web\n   npm install\n   ```\n8. 启动前端服务：\n\n   ```bash\n   npm run dev\n   ```\n\n   _以下输出确认系统已成功启动：_\n\n   ![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Finfiniflow_ragflow_readme_ec85af2f698f.png)\n9. 开发完成后停止 RAGFlow 前后端服务：\n\n   ```bash\n   pkill -f \"ragflow_server.py|task_executor.py\"\n   ```\n\n## 📚 文档\n\n- [快速入门](https:\u002F\u002Fragflow.io\u002Fdocs\u002Fdev\u002F)\n- [配置](https:\u002F\u002Fragflow.io\u002Fdocs\u002Fdev\u002Fconfigurations)\n- [发布说明](https:\u002F\u002Fragflow.io\u002Fdocs\u002Fdev\u002Frelease_notes)\n- [用户指南](https:\u002F\u002Fragflow.io\u002Fdocs\u002Fcategory\u002Fuser-guides)\n- [开发者指南](https:\u002F\u002Fragflow.io\u002Fdocs\u002Fcategory\u002Fdeveloper-guides)\n- [参考文档](https:\u002F\u002Fragflow.io\u002Fdocs\u002Fdev\u002Fcategory\u002Freferences)\n- [常见问题解答](https:\u002F\u002Fragflow.io\u002Fdocs\u002Fdev\u002Ffaq)\n\n## 📜 路线图\n\n请参阅 [RAGFlow 2026 年路线图](https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fissues\u002F12241)\n\n## 🏄 社区\n\n- [Discord](https:\u002F\u002Fdiscord.gg\u002FNjYzJD3GM3)\n- [Twitter](https:\u002F\u002Ftwitter.com\u002Finfiniflowai)\n- [GitHub 讨论区](https:\u002F\u002Fgithub.com\u002Forgs\u002Finfiniflow\u002Fdiscussions)\n\n## 🙌 贡献\n\nRAGFlow 的蓬勃发展离不开开源社区的合作。我们欢迎来自社区的各种贡献。如果您希望参与其中，请先阅读我们的 [贡献指南](https:\u002F\u002Fragflow.io\u002Fdocs\u002Fdev\u002Fcontributing)。","# RAGFlow 快速上手指南\n\nRAGFlow 是一款领先的开源检索增强生成（RAG）引擎，深度融合了文档解析能力与 Agent 工作流，旨在为大语言模型提供高质量的上下文层。它支持多种文档格式的深层理解、可视化分块及可追溯的引用，适合构建企业级 AI 应用。\n\n## 环境准备\n\n在开始之前，请确保您的服务器满足以下硬件和软件要求：\n\n### 系统要求\n- **CPU**: 4 核及以上\n- **内存**: 16 GB 及以上\n- **磁盘空间**: 50 GB 及以上\n- **操作系统**: Linux (推荐 Ubuntu\u002FCentOS)，Windows 或 macOS (需安装 Docker Desktop)\n\n### 前置依赖\n- **Docker**: 版本 >= 24.0.0\n- **Docker Compose**: 版本 >= v2.26.1\n- **gVisor** (可选): 仅当您计划使用 RAGFlow 的代码执行器（沙箱）功能时需要安装。\n\n> **提示**: 如果尚未安装 Docker，请访问 [Docker 官方安装文档](https:\u002F\u002Fdocs.docker.com\u002Fengine\u002Finstall\u002F)。国内用户建议使用阿里云或清华大学的 Docker 镜像加速器以提升下载速度。\n\n## 安装步骤\n\n### 1. 配置系统内核参数\nRAGFlow 依赖 Elasticsearch，需要调整 `vm.max_map_count` 参数。\n\n```bash\n# 检查当前值\nsysctl vm.max_map_count\n\n# 如果小于 262144，临时设置（重启失效）\nsudo sysctl -w vm.max_map_count=262144\n\n# 永久生效：编辑 \u002Fetc\u002Fsysctl.conf，添加或修改以下行\necho \"vm.max_map_count=262144\" | sudo tee -a \u002Fetc\u002Fsysctl.conf\nsudo sysctl -p\n```\n\n### 2. 克隆项目代码\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow.git\ncd ragflow\u002Fdocker\n```\n\n> **注意**: 目前官方 Docker 镜像仅支持 x86 架构。如果您使用的是 ARM64 设备（如 Apple Silicon Mac 或鲲鹏服务器），请参考官方文档自行构建镜像。\n\n### 3. 启动服务\n使用 Docker Compose 启动服务。默认使用 CPU 进行文档解析任务。\n\n```bash\n# 使用 CPU 启动\ndocker compose -f docker-compose.yml up -d\n\n# (可选) 如果有 NVIDIA GPU 并希望加速 DeepDoc 任务，先修改配置再启动：\n# sed -i '1i DEVICE=gpu' .env\n# docker compose -f docker-compose.yml up -d\n```\n\n### 4. 验证运行状态\n查看日志确认服务是否完全启动。看到如下 ASCII Logo 即表示成功：\n\n```bash\ndocker logs -f docker-ragflow-cpu-1\n```\n\n预期输出末尾包含：\n```text\n ____   ___    ______ ______ __\n\u002F __ \\ \u002F   |  \u002F ____\u002F\u002F ____\u002F\u002F \u002F____  _      __\n...\n* Running on all addresses (0.0.0.0)\n```\n\n## 基本使用\n\n### 1. 登录系统\n在浏览器中访问您的服务器 IP 地址（默认端口 80，无需手动添加）：\n`http:\u002F\u002F\u003C您的服务器 IP>`\n\n### 2. 配置大模型密钥\n首次登录后，需要配置 LLM 提供商的 API Key 才能正常使用：\n\n1. 点击界面右上角的设置或进入 **Service Configuration**。\n2. 找到 `user_default_llm` 选项，选择您想要使用的模型厂商（如 OpenAI, Azure, Zhipu, Moonshot 等）。\n3. 填入对应的 `API_KEY`。\n   > 详细配置指南可参考：[LLM API Key 设置文档](https:\u002F\u002Fragflow.io\u002Fdocs\u002Fdev\u002Fllm_api_key_setup)\n\n### 3. 创建第一个知识库\n1. **新建知识库**：在首页点击\"+\"号创建一个新的知识库。\n2. **上传文档**：支持 Word, PDF, Excel, TXT, Markdown 等多种格式。RAGFlow 会自动进行深度文档解析和分块。\n3. **解析预览**：在\"Chunk\"标签页查看自动分块结果，支持人工干预和调整模板。\n4. **测试检索**：进入\"Chat\"或\"Test\"页面，输入问题，系统将基于您上传的文档内容进行回答，并展示引用来源。\n\n现在，您已经成功部署并使用了 RAGFlow！您可以进一步探索其 Agent 编排、多路召回重排序等高级功能。","某大型制造企业的技术文档团队正试图构建一个内部智能问答系统，让工程师能快速从数千份复杂的设备维护手册和故障排查报告中获取精准信息。\n\n### 没有 ragflow 时\n- **解析效果差**：传统方案难以处理手册中大量的混合排版、复杂表格和工程图纸，导致关键数据在提取时丢失或错乱。\n- **检索不精准**：简单的关键词匹配无法理解工程师的自然语言提问（如“液压泵异响怎么处理”），经常返回无关文档片段。\n- **回答不可信**：大模型因缺乏准确的上下文依据，容易产生“幻觉”，编造不存在的维修步骤，存在严重安全隐患。\n- **调试黑盒化**：当回答错误时，开发人员无法追溯是文档解析、检索排序还是生成环节出了问题，排查问题如同大海捞针。\n\n### 使用 ragflow 后\n- **深度文档解析**：ragflow 内置的深度文档解析引擎能完美还原手册中的表格结构和图文关系，确保技术参数完整入库。\n- **语义精准匹配**：利用先进的混合检索策略，ragflow 能准确理解工程师的意图，从海量非结构化数据中锁定最相关的知识片段。\n- **可解释的回答**：ragflow 在生成答案时会明确标注引用来源（具体到页码和段落），让每一条维修建议都有据可查，杜绝胡编乱造。\n- **全链路可视化**：通过可视化的调试界面，团队可以清晰看到数据流转的每一步，快速优化检索策略，将系统迭代周期从周缩短至小时。\n\nragflow 通过将深度的文档理解能力与可追溯的检索生成流程相结合，为企业构建了真正可信、可控且高精度的专业知识大脑。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Finfiniflow_ragflow_23cfcd2b.png","infiniflow","InfiniFlow","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Finfiniflow_8380b984.png","",null,"infiniflowai","https:\u002F\u002Fgithub.com\u002Finfiniflow",[83,87,91,95,99,103,107,111,115,118],{"name":84,"color":85,"percentage":86},"Python","#3572A5",46.3,{"name":88,"color":89,"percentage":90},"TypeScript","#3178c6",33.1,{"name":92,"color":93,"percentage":94},"C++","#f34b7d",9.4,{"name":96,"color":97,"percentage":98},"Go","#00ADD8",9.2,{"name":100,"color":101,"percentage":102},"C","#555555",1,{"name":104,"color":105,"percentage":106},"Shell","#89e051",0.3,{"name":108,"color":109,"percentage":110},"Less","#1d365d",0.2,{"name":112,"color":113,"percentage":114},"CSS","#663399",0.1,{"name":116,"color":117,"percentage":114},"Dockerfile","#384d54",{"name":119,"color":120,"percentage":114},"JavaScript","#f1e05a",77062,8658,"2026-04-04T04:44:48","Apache-2.0","Linux, macOS, Windows","非必需（可选用于加速 DeepDoc 任务）。README 未指定具体显卡型号、显存大小或 CUDA 版本，但指出 Docker 镜像仅支持 x86 平台，ARM64 需自行构建。","最低 16GB",{"notes":129,"python":130,"dependencies":131},"1. 必须设置系统参数 vm.max_map_count >= 262144。2. 官方提供的 Docker 镜像仅适用于 x86 架构，ARM64 用户需参考文档自行构建镜像。3. 若需使用代码执行器（沙箱）功能，需安装 gVisor。4. 默认通过 Docker Compose 部署，无需手动安装 Python 依赖库。","未说明",[132,133,134],"Docker >= 24.0.0","Docker Compose >= v2.26.1","gVisor (可选，用于代码执行器沙箱)",[15,14,13,26,54],[137,138,139,140,141,142,143,144,145,146,147,148,149,150,151,152,153,154,155,156],"document-understanding","llm","rag","retrieval-augmented-generation","agent","graphrag","deepseek","deepseek-r1","ollama","ai","agentic-ai","mcp","openai","agentic","deep-research","agentic-workflow","context-engineering","context-retrieval","harness","openclaw",40,"2026-03-27T02:49:30.150509","2026-04-06T00:58:36.152950",[161,166,171,176,180,185],{"id":162,"question_zh":163,"answer_zh":164,"source_url":165},14424,"RAGFlow 中知识图谱索引是如何构建的？","知识图谱索引的构建过程包括以下步骤：\n1. **知识图谱构建步骤**：在数据提取和索引之间增加构建步骤，从现有块中创建额外块以增强多跳问答能力。\n2. **实体类型配置**：定义要提取的实体类型（默认包括组织、个人、事件和类别）。\n3. **构建方法选择**：可选择 GraphRAG 或 LightRAG（默认，资源消耗较少）来提取实体和关系。\n4. **实体解析**：可选启用实体去重，合并相似实体以构建更有效的图谱。\n5. **社区报告生成**：可选生成由关系链接的实体集群报告。\n这些步骤会在新文件上传时自动更新并构建统一的知识图谱。","https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fissues\u002F6985",{"id":167,"question_zh":168,"answer_zh":169,"source_url":170},14425,"如何在 Docker 部署的 RAGFlow 中连接本地 Windows 上运行的 Ollama 服务？","如果 Ollama 运行在本地 Windows 系统（非 Docker 环境），需要在 RAGFlow 的 Ollama 设置中将基础 URL 填写为 `http:\u002F\u002Fhost.docker.internal:11434\u002F`。\n\n同时，确保在 `docker-compose.yml` 或 `docker-compose-gpu.yml` 中添加以下配置以允许容器访问宿主机网络：\n```yaml\nextra_hosts:\n  - \"host.docker.internal:host-gateway\"\n```\n无需修改环境变量 `OLLAMA_HOST` 或 Ragflow 容器的其他构建配置。","https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fissues\u002F333",{"id":172,"question_zh":173,"answer_zh":174,"source_url":175},14426,"启动 RAGFlow 时遇到 'dependency failed to start: container ragflow-mysql is unhealthy' 错误如何解决？","该问题通常由端口冲突或 MySQL 启动超时引起，可尝试以下解决方案：\n\n1. **修改端口配置**：\n   - 在 `ragflow\u002Fdocker\u002F.env` 文件中，将 `MYSQL_PORT=5545` 改为未被占用的端口（如 `3307`）。\n   - 在 `ragflow\u002Fdocker\u002Fdocker-compose-base.yml` 文件中，将 `${MYSQL_PORT}:3306` 替换为具体的端口号（如 `3307:3306`）。\n\n2. **优化 MySQL 配置与健康检查**：\n   - 将镜像版本指定为 `mysql:8.0`。\n   - 增加健康检查的重试次数和间隔时间，例如：\n     ```yaml\n     healthcheck:\n       interval: 20s\n       timeout: 20s\n       retries: 20\n     ```\n   - 添加启动参数以确保字符集和认证插件兼容：\n     ```yaml\n     command:\n       --max_connections=1000\n       --character-set-server=utf8mb4\n       --collation-server=utf8mb4_general_ci\n       --default-authentication-plugin=caching_sha2_password\n     ```","https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fissues\u002F329",{"id":177,"question_zh":178,"answer_zh":179,"source_url":165},14427,"用户提问后，RAGFlow 系统内部是如何处理问题并生成答案的？","虽然官方未提供全自动的可视化流程图解，但根据源码分析，处理流程主要包含：\n1. **关键词向量化搜索**：在进行关键词搜索时，系统通过 `get_relevant_ents_by_keywords` 函数调用 `get_vector` 方法，将关键词转换为向量形式进行检索。\n2. **知识图谱增强**：如果启用了知识图谱，系统会利用构建好的图谱进行多跳推理和实体关联，以补充传统向量检索的不足。\n3. **结果整合**：最终结合向量检索结果和图谱推理结果，由大模型生成最终回答。",{"id":181,"question_zh":182,"answer_zh":183,"source_url":184},14428,"如何切换 RAGFlow 默认的 MinIO 对象存储为阿里云 OSS？","目前直接删除 MinIO 配置并添加 OSS 信息会导致启动报错（提示缺少 MinIO 连接配置）。这表明当前版本可能尚未完全支持通过简单修改 YAML 来无缝切换至 OSS，或者需要保留部分 MinIO 兼容配置项。建议关注官方后续更新或提交 Bug 标签的新 Issue 以获取官方支持的具体配置指南。","https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fissues\u002F10979",{"id":186,"question_zh":187,"answer_zh":188,"source_url":189},14429,"使用本地 Ollama 模型进行对话时出现异常（如缺少思考结束标记），而在线模型正常，原因是什么？","此问题可能与特定本地模型的输出格式解析有关。有用户反馈在使用本地部署的 Deepseek-R1 (sglang) 时，仅显示思考开始标记而无结束标记，而在线模型或其他本地量化版本表现不同。这通常是因为本地推理后端（如 sglang, ollama）与大模型本身的输出协议（如 Thinking Token 的处理）存在兼容性差异。建议检查模型版本是否最新，或尝试更换本地推理后端（如从 Ollama 切换到其他支持完整思维链输出的框架），并关注官方代码路径的修复进展。","https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fissues\u002F11232",[191,196,201,206,211,216,221,226,231,236,241,246,250,255,260,265,270,275,280,285],{"id":192,"version":193,"summary_zh":194,"released_at":195},81244,"nightly","于2026-04-03 21:32:35+08:00 从 60ec5880e5d3864fe4038f58ca6ec9c35f9a6cd3 构建的夜间版本\n","2025-12-01T13:17:42",{"id":197,"version":198,"summary_zh":199,"released_at":200},81245,"v0.24.0","# 概要\n\n## 新特性\n\n- 内存\n  - 引入了面向开发者的 API 和 SDK，便于集成。\n  - 在控制台中新增了内存提取日志展示功能，以提升调试和追踪效率。\n- 数据集\n  - 增加了对元数据批量管理的支持。\n  - 将“目录（ToC）”更名为“页码索引（PageIndex）”。\n- 代理\n  - 推出了全新的类聊天代理对话管理界面，支持会话及对话历史的保留。\n  - 引入了多沙箱机制，目前支持本地 gVisor 和阿里云，并兼容主流沙箱 API（可在管理页面中进行配置）。\n- 聊天\n  - 新增“思考”模式，移除了原有的“推理”配置选项。\n  - 针对深度研究场景优化了检索策略，提升了召回准确率。\n- 管理端\n  - 增加了对多管理员账号配置的支持。\n- 模型配置中心\n  - 在添加新模型时，新增了模型连接测试功能。\n- 生态系统\n  - 新增对 OceanBase 的支持，作为 MySQL 的数据库替代方案。\n  - 新增对 PaddleOCR-VL 的支持。\n- 模型\n  - 新增对 Kimi 2.5、Stepfun 3 和 doubao-embedding-vision 等模型的支持。\n- 数据源\n  - 新增了对 Zendesk、Bitbucket 等平台的数据源集成。\n\n## 变更内容\n* 修复：KG 搜索问题。由 @KevinHuSh 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12364 中完成。\n* 功能：新增 Zendesk 数据源集成，支持配置与同步功能。由 @buua436 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12344 中完成。\n* 文档重构：RAGFlow 管理端文档重构。由 @JinHai-CN 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12361 中完成。\n* 文档：新增 v0.23.1 版本发布说明。由 @writinwaters 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12371 中完成。\n* 修复基础文档中的 URL 错误。由 @yingfeng 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12372 中完成。\n* 文档：更新 v0.23.0 版本发布说明。由 @writinwaters 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12374 中完成。\n* 功能：Bitbucket 连接器。由 @Magicbook1108 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12332 中完成。\n* 修复：元数据同步问题；在首页新增内存标签页。由 @dcc123456 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12368 中完成。\n* 修复：WebDAV 连接器问题。由 @Magicbook1108 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12380 中完成。\n* 功能：使用 shadcn 重构文档页面。#10427。由 @cike8899 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12376 中完成。\n* 修复：Dockerfile.deps 中 Tika 版本不匹配问题（3.0.0 → 3.2.3）。由 @mmtmr 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12267 中完成。\n* 更新 README 文件。由 @yingfeng 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12386 中完成。\n* 管理端 CLI：支持授予\u002F撤销用户管理员权限。由 @JinHai-CN 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12381 中完成。\n* 更新管理端 CLI 的 HELP 命令。由 @JinHai-CN 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12387 中完成。\n* 功能：读取 Web API 测试用例。由 @Lynn-Inf 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12383 中完成。\n* 修复：修复了上传 DSL 对话框 w","2026-02-10T09:27:14",{"id":202,"version":203,"summary_zh":204,"released_at":205},81246,"v0.23.1","## v0.23.1\n\n发布于2025年12月31日。\n\n### 改进\n\n- 内存：在选择所有内存类型时，增强了内存提取的稳定性。\n- RAG：优化了针对图片和表格的上下文窗口提取策略。\n\n\n### 修复的问题\n\n- 内存：\n  - 如果存在空的内存对象，RAGFlow 服务器将无法启动。\n  - 无法删除新创建的空内存。\n- RAG：不支持 MDX 文件解析。\n\n### 数据源\n\n- GitHub\n- GitLab\n- Asana\n- IMAP\n\n## 变更内容\n* 由 @JinHai-CN 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12269 中修复文档中的错误\n* 文档：由 @RinCodeForge927 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12250 中添加关于 .env 文件中默认密码的安全警告\n* 由 @yiminghub2024 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12275 中更新 deploy_local_llm.mdx 中 vLLM 指南的图片\n* 由 @yingfeng 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12276 中更新 deploy_local_llm.mdx\n* 修复：数据源 S3 页面样式，由 @dcc123456 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12255 中完成\n* 由 @JinHai-CN 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12281 中修复 IDE 警告\n* 修复：从 ES 初始化内存大小，由 @Lynn-Inf 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12282 中完成\n* 重构：保持时间戳一致性，由 @Woody-Hu 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12279 中完成\n* 由 @JinHai-CN 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12288 中修复 ts 和 datetime 的一致性问题\n* 修复：当文档解析器配置发生变更时，删除块图像。由 @lys1313013 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12262 中完成\n* 功能：增加 MDX 文件支持，由 @lys1313013 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12261 中实现\n* 修复：如果聊天页面出现错误信息，后续的消息引用将无法正确显示。#12252，由 @cike8899 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12283 中完成\n* 修复（SDK）：确保 rm_chunk API 中定义的变量，由 @maoxiaowang 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12274 中完成\n* 安全：在视觉运算符中用 ast.literal_eval 替代不安全的 eval，由 @RinCodeForge927 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12236 中完成\n* 功能：添加 Asana 数据源集成及配置选项，由 @buua436 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12239 中完成\n* Helm：改进，由 @LyleLaii 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10976 中完成\n* 重构：docling 解析器将关闭字节流，由 @Woody-Hu 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12280 中完成\n* 功能：GitHub 连接器，由 @Magicbook1108 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12292 中完成\n* 功能：GitLab 连接器，由 @Magicbook1108 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12248 中完成\n* 撤销“功能：GitHub 连接器”，由 @KevinHuSh 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12296 中完成\n* 修复：选择要删除的消息，由 @Lynn-Inf 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12295 中完成\n* 修复：S3 参数错误，由 @dcc123456 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F12290 中完成\n* 文档：为 Do 添加可选的代理参数","2025-12-31T06:04:03",{"id":207,"version":208,"summary_zh":209,"released_at":210},81247,"v0.23.0","## v0.23.0\n\n于2025年12月27日发布。\n\n### 新特性\n\n- 内存\n  - 实现了用于管理内存的**Memory**接口。\n  - 支持通过**Retrieval**或**Message**组件配置上下文。\n- 代理\n  - 通过重构底层架构，提升了**Agent**组件的性能。\n  - **Agent**组件现在可以输出结构化数据，供下游组件使用。\n  - 支持使用Webhook触发代理执行。\n  - 支持语音输入\u002F输出。\n  - 支持为每个**Agent**组件配置多个**Retrieval**组件。\n- 数据摄入管道\n  - 在**Transformer**组件中支持提取目录，以提升长上下文RAG的性能。\n- 数据集\n  - 支持为图片和表格配置上下文窗口。\n  - 引入父子分块策略。\n  - 支持在文件解析过程中自动生成元数据。\n- 聊天：支持语音输入。\n\n### 改进\n\n- 将RAGFlow的文档引擎[Infinity]([GitHub - infiniflow\u002Finfinity: 专为LLM应用打造的原生AI数据库，提供惊人的](https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Finfinity))升级至v0.6.15（向后兼容）。\n\n### 数据源\n\n- Google Cloud Storage\n- Gmail\n- Dropbox\n- WebDAV\n- Airtable\n\n### 模型支持\n\n- GPT-5.2\n- GPT-5.2 Pro\n- GPT-5.1\n- GPT-5.1 Instant\n- Claude Opus 4.5\n- MiniMax M2\n- GLM-4.7。\n- A MinerU配置界面。\n- AI Badgr（模型提供商）。\n\n### API变更\n\n#### HTTP API\n\n- [与代理对话](.\u002Freferences\u002F[http_api_reference.md](http:\u002F\u002Fhttp_api_reference.md\u002F)#converse-with-agent)返回完整的执行轨迹日志。\n- [创建聊天完成](.\u002Freferences\u002F[http_api_reference.md](http:\u002F\u002Fhttp_api_reference.md\u002F)#create-chat-completion)支持基于元数据的过滤。\n- [与聊天助手对话](.\u002Freferences\u002F[http_api_reference.md](http:\u002F\u002Fhttp_api_reference.md\u002F)#converse-with-chat-assistant)支持基于元数据的过滤。\n\n## 变更内容\n* 修复cohere rerank的base_url默认值，由@HeyPhiS在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11353中完成。\n* 功能：添加OceanBase文档引擎，由@whhe在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11228中完成。\n* 修复(llm)：处理total_token_count_from_response中的None响应，由@zhangzhefang-github在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10941中完成。\n* 功能：为PPTX和电子表格文档类型添加TCADP解析器。由@aidansu在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11041中完成。\n* 本地化：更新俄语语言包，由@rjohny55在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11393中完成。\n* 本地化en：添加俄语语言选项，由@rjohny55在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11392中完成。\n* 功能（画布）：添加思维标签支持，由@huang-aoqin在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11359中完成。\n* 在docker-compose-base.yml中对命令使用数组语法，由@panzi在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11391中完成。\n* 将infinity升级至0.6.6，由@yuzhichang在https:\u002F\u002Fgithub.com\u002Fin","2025-12-27T13:25:18",{"id":212,"version":213,"summary_zh":214,"released_at":215},81248,"v0.22.1","版本 v0.22.1 由 cfdccebb178bc2c49eacf321e5da483ed1e48511 构建，发布时间为 2025-11-19 20:13:52+08:00\n\n## 概述\n\n### 改进\n\n- Agent：\n  - 支持将 Agent 输出导出为 Word 或 Markdown 格式。\n  - 新增 **列表操作** 组件。\n  - 新增 **变量聚合器** 组件。\n- 数据源：\n  - 支持兼容 S3 的数据源，例如 MinIO。\n  - 新增与 JIRA 的数据同步功能。\n- 继续对 **个人资料** 页面布局进行重新设计。\n- 将 Flask Web 框架从同步模式升级为异步模式，以提高并发能力，并避免在请求上游 LLM 服务时出现阻塞问题。\n\n### 修复的问题\n\n- v0.22.0 中的一个问题：用户无法解析已上传的文件，也无法在包含解析后文件的数据集中切换嵌入模型，尤其是在使用 `-full` RAGFlow 版本中的内置模型时。\n- Word 文档中图片拼接问题。[#11310](https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11310)\n- 聊天记录中图片与文本混排时显示不正确。\n\n### 新增支持的模型\n\n- Gemini 3 Pro 预览版\n\n## 变更内容\n* @JinHai-CN 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11210 中重构函数名称\n* @buua436 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11222 中修复：重置 Agent 组件的输出\n* @dcc123456 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11221 中修复：支持个人资料头像裁剪 #10703\n* @Magicbook1108 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11212 中修复：更新 README 文件\n* @yngvarhuang 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11231 中新增：文档上传 API 支持父路径 20251112\n* @KevinHuSh 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11238 中重构：移除无用代码\n* @buua436 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11229 中新增：支持生成知识图谱和 RAPTOR 的 API\n* @asiroliu 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11244 中测试：通过添加“(1)”来处理重复名称\n* @Magicbook1108 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11246 中修复：法律解析器\n* @JinHai-CN 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11249 中进行了小幅调整\n* @yongtenglei 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11206 中新增：为 RAPTOR 添加容错机制\n* @buua436 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11254 中新增：更新 check_embedding API\n* @writinwaters 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11260 中补充文档：默认管理员凭据\n* @JinHai-CN 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11267 中重构管理服务器中的服务配置获取逻辑\n* @dcc123456 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11269 中修复：添加会话变量类型并修改配置\n* @redredrrred 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11248 中移除代码中残留的账号和密码信息\n* @Magicbook1108 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F11266 中修复：重排序算法\n* @Lynn-Inf 在 https:\u002F\u002Fgithub.com\u002F 中…","2025-11-19T12:14:02",{"id":217,"version":218,"summary_zh":219,"released_at":220},81249,"v0.22.0","发布于2025年11月12日。\n\n## 重大变更\n\n- 从本版本起，我们仅提供精简版（不嵌入模型）的 Docker 镜像，并且不再在镜像标签后附加 `-slim` 后缀。\n\n## 新特性\n\n- 数据集：\n  - 支持从五个在线数据源同步数据（AWS S3、Google Drive、Notion、Confluence 和 Discord）。\n  - RAPTOR 可以在整个数据集上构建，也可以针对单个文档进行构建。\n- 数据摄取管道：在 **Parser** 组件中支持 [Docling 文档解析](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling)。\n- 推出全新的管理 Web UI 控制面板，用于图形化用户管理和服务状态监控。\n- Agent：\n  - 支持结构化输出。\n  - 在 **Retrieval** 组件中支持元数据过滤。\n  - 引入了具有数据操作和会话变量定义功能的 **Variable aggregator** 组件。\n\n## 优化改进\n\n- Agent：在 **Await Response** 组件中支持可视化前序组件的输出。\n- 重构了模型提供商页面。\n- 将 RAGFlow 的文档引擎 Infinity 升级至 v0.6.5。\n\n## 新增模型\n- Kimi-K2-Thinking\n\n## 新的 Agent 模板\n\n- 交互式 Agent，结合实时用户反馈动态优化 Agent 输出。\n\n\n## 变更内容\n* 功能：新增 Docling 解析器，由 @buua436 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10759 中实现。\n* 移动测试文件，由 @JinHai-CN 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10765 中完成。\n* 不再发布完整版镜像，由 @yuzhichang 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10654 中提出。\n* 修复：ERROR 20 方法 rag.nlp.search.Dealer.search() 参数 highlight=\"None\" 违反类型提示 bool | list，因为 \u003Cclass \"builtins.NoneType\"> \"None\" 既不是列表也不是布尔值。由 @Woody-Hu 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10743 中修复。\n* 修复：英文优先使用同义词匹配而非 WordNet，由 @Magicbook1108 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10762 中完成。\n* 修复：修正了数据集日志头像显示错误的问题 #9869，由 @cike8899 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10764 中修复。\n* 补充文档中未完善的参数说明，由 @jiasu-hezhip 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10758 中完成。\n* 修复：Opendal 缺失租户 ID，由 @Woody-Hu 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10774 中修复。\n* 功能：调整画布节点样式 #10703，由 @cike8899 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10795 中实现。\n* 从 EditTag 中移除 value={field}，由 @adrianad 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10767 中完成。\n* 修复了传递空数组不会更新的 bug，由 @jiasu-hezhip 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10755 中修复。\n* 增加关于 RAGFlow 镜像的更多文档，由 @yuzhichang 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10800 中完成。\n* 修复：Gemini CV 模型聊天问题，由 @KevinHuSh 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10799 中修复。\n* 文档：更新 README 文件，由 @KevinHuSh 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10801 中完成。\n* 修复：检索 DOCX 图片时出现错误 (docx.image.exceptions.UnrecognizedIm","2025-11-12T06:57:46",{"id":222,"version":223,"summary_zh":224,"released_at":225},81250,"v0.21.1","版本 v0.21.1 由 de24e74b4c90ec154965c1bc7dc7a0fe79e669a1 构建，发布时间：2025-10-23 19:08:09+08:00\n\n## 新特性\n\n- 实验性功能：新增对使用 MinerU 解析 PDF 文档的支持。详情请参阅 [这里](.\u002Ffaq.mdx#how-to-use-mineru-to-parse-pdf-documents)。\n\n## 改进\n\n- 优化了数据集和个人中心页面的 UI\u002FUX。\n- 将 RAGFlow 的文档引擎 [Infinity](https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Finfinity) 升级至 v0.6.1。\n\n## 修复的问题\n\n- 视频解析相关问题。\n\n## 变更内容\n* 修复 Google Cloud 理性推理模型（如 Gemini）的问题 - @glu000 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10502 中对 #10474 问题的进一步修复\n* 修复：支持打印表格时自动调整宽度 - @Lynn-Inf 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10575 中完成\n* 文档：更新发布标签 - @Lynn-Inf 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10578 中完成\n* 修复 (_handle_entity_relation_summary)：正确计算 descriptions_list - @ffffffff0 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10534 中完成\n* 修复 (i18n)：更新 RAPTOR 功能的中英文描述 #9869 - @dcc123456 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10581 中完成\n* 重构管理员客户端的消息提示 - @JinHai-CN 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10583 中完成\n* 文档：为 v0.21.0 更新数据集配置、KG 构建及 RAPTOR 构建 - @writinwaters 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10584 中完成\n* 不对 Infinity 进行重排序 - @yuzhichang 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10579 中完成\n* 文档：添加长上下文 RAG 指南 - @writinwaters 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10591 中完成\n* 使用相对路径导入同一模块 - @JinHai-CN 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10587 中完成\n* 改进文件管理功能 - @yongtenglei 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10577 中完成\n* 更新管理员客户端 README 和文档 - @JinHai-CN 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10594 中完成\n* 功能：支持指定 OpenRouter 模型提供商 - @buua436 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10550 中完成\n* 构建 (依赖项)：将 \u002Fweb 目录下的 axios 从 1.7.2 升级至 1.12.0 - @dependabot[bot] 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10393 中完成\n* 修复 (数据集设置)：移除配置中的 TagItems 介绍及使用 #9869 - @dcc123456 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10595 中完成\n* 修复：添加 toc_kwd 字段并更新 page_num_int 类型 - @asiroliu 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10596 中完成\n* 更新架构图及 ragflow-cli 版本 - @JinHai-CN 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10605 中完成\n* 功能：新增忘记密码重置功能，解决 #8547 问题 - @Magicbook1108 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10586 中完成\n* 修复：添加 RetryingPooledPostgresqlDatabase 以处理 max_retries 参数 - @arogan178 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10524 中完成\n* 功能：折叠工具节点和检索节点的多余部分 #9869 - @cike8899 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F","2025-10-23T11:08:10",{"id":227,"version":228,"summary_zh":229,"released_at":230},81251,"v0.21.0","发布于2025年10月15日。\n\n### 新特性\n\n- 可编排的摄取管道：支持自定义的数据摄取和清洗工作流，使用户能够灵活设计数据流转流程，或直接在画布上应用官方提供的数据流转模板。\n- GraphRAG与RAPTOR写入流程优化：将自动增量构建流程替换为手动批量构建，显著降低构建开销。\n- 长上下文RAG：自动生成功能文档级别的目录（TOC）结构，以缓解因分块不准确或过细导致的上下文丢失问题，大幅提升检索质量。该功能现已通过TOC提取模板提供。\n- 管理CLI：引入全新的命令行工具用于系统管理，允许用户通过命令行管理和监控RAGFlow的服务状态。\n\n### 改进\n\n- 重新设计了RAGFlow的登录和注册页面。\n- 将RAGFlow的文档引擎Infinity升级至v0.6.0。\n\n### 新增模型\n\n- 通义千问3系列\n- Claude Sonnet 4.5\n- 美团LongCat-Flash-Thinking\n\n## 新代理模板\n\n- 公司研究报告深度剖析代理：专为金融机构设计，帮助分析师快速整理信息、生成研究报告并作出投资决策。\n- 可编排的摄取管道模板：用户可在画布上应用此模板，快速建立标准化的数据摄取和清洗流程。\n\n\n## 变更内容\n* 功能：数据流转支持电子表格和文字处理文档，由@yongtenglei在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9996中实现。\n* 功能：数据流转支持Markdown格式，由@Lynn-Inf在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10003中实现。\n* 重构：改进图像转Base64时BytesIO的使用，由@Woody-Hu在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9997中完成。\n* 修复：修复了新添加的工具操作符在编辑表单后消失的问题 #10013，由@cike8899在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10016中解决。\n* 功能：在创建代理对话框中新增类型卡片 #9869，由@cike8899在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10025中实现。\n* 功能（配置）：更新服务配置模板的新默认值，由@asiroliu在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10029中完成。\n* 功能：从代理列表页导入DSL #9869，由@cike8899在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10033中实现。\n* 功能：增加了与数据流知识库相关的UI功能 #3221，由@dcc123456在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10038中实现。\n* 重构：改进GeminiCV中BytesIO的使用，由@Woody-Hu在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10042中完成。\n* 功能：翻译解析器操作符 #9869，由@cike8899在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10037中实现。\n* 修复：在ragflow\u002Frag\u002Fapp\u002Fnaive.py中，若一行内包含多张图片，其他图片将会丢失，由@Woody-Hu在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9968中修复。\n* 修复：令牌数量超出限制的问题，由@Lynn-Inf在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F10046中解决。\n* ","2025-10-15T05:35:24",{"id":232,"version":233,"summary_zh":234,"released_at":235},81252,"v0.20.5","## 摘要\n\n### 改进\n- 优化了智能体性能：提升了简单任务的规划和反思速度；针对可并行化的场景优化了并发工具调用，显著减少了整体响应时间。\n- 公开了智能体提示框架：开发者现在可以在系统提示部分自定义和覆盖框架级别的提示，从而增强灵活性和控制能力。\n- 增强了执行 SQL 组件：将原有的变量引用组件替换为文本输入框，支持自由编写 SQL 并保留变量功能。\n- 聊天功能：重新启用了推理模式和跨语言搜索功能。\n- 检索 API 优化：在 Retrieve chunks 方法中新增了元数据过滤支持。\n\n### 新增模型\n- 美团 LongCat\n- Kimi：kimi-k2-turbo-preview 和 kimi-k2-0905-preview\n- 通义千问：qwen3-max-preview\n- SiliconFlow：DeepSeek V3.1\n\n### 修复的问题\n- 数据集：已删除的文件仍可被搜索到。\n- 聊天功能：无法与 Ollama 模型进行对话。\n- 智能体：解决了引用切换失败、任务模式需要对话触发、多轮对话中重复回答以及并行执行结果重复总结等问题。\n\n## 变更内容\n* 修复 @JinHai-CN 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9766 中的错别字\n* 功能\u002F工作流 英文转中文 @YyBoom233 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9742 中完成\n* 修复：移除第二轮值。@KevinHuSh 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9756 中完成\n* 重构：改进 upload_to_minio 的尝试逻辑 @Woody-Hu 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9735 中完成\n* 在 translation table index.tsx 中添加俄语 @rjohny55 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9773 中完成\n* 在 README 中澄清 pre-commit 与 uv 的安装方法 @aopstudio 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9749 中完成\n* 功能：添加 AvatarUpload 组件 #3221 @cike8899 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9777 中完成\n* 添加俄语版本 Update app.tsx @rjohny55 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9772 中完成\n* 智能体在任务模式下直接输出结果 #9745 @cike8899 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9746 中完成\n* 改进模型标签渲染方式，将逗号分隔的字符串拆分为带有样式的 \u003CTag> 组件 @aopstudio 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9762 中完成\n* 修复：智能体模板错误。@KevinHuSh 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9784 中完成\n* 功能：使用 AvatarUpload 替换数据集和搜索页面上的头像设置 #3221 @cike8899 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9785 中完成\n* 修复：优化了变量节点显示及智能体模板多语言支持 #3221 @dcc123456 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9787 中完成\n* 重构：统一智能体补全与 OpenAI 兼容补全 API 的引用格式 @yongtenglei 在 https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9792 中完成\n* 修复：解决了聊天页面上思考模式无法关闭的问题 #9789 @cike8899 在 gith 中完成","2025-09-10T03:23:40",{"id":237,"version":238,"summary_zh":239,"released_at":240},81253,"v0.20.4","## 摘要\n\n### 改进\n\n- Agent组件：完成Agent组件的中文本地化。\n- 引入`ENABLE_TIMEOUT_ASSERTION`环境变量，用于启用或禁用文件解析任务的超时断言。\n- 数据集：\n  - 改进了Markdown文件解析，支持AST以避免意外分块。\n  - 增强了HTML解析功能，支持基于bs4的HTML标签遍历。\n\n### 新增模型\n\n智谱GLM-4.5\n\n### 新增Agent模板\n\n电商客服工作流：该模板旨在利用内部知识库处理关于产品功能和多产品比较的咨询，并管理安装预约的安排。\n\n### 修复的问题\n\n- 数据集：\n  - 无法与团队共享资源。\n  - 对上传文件的数量和大小限制不合理。\n- 聊天：\n  - 无法在回复中预览引用的文件。\n  - 文件上传后无法发送消息。\n- 一起OAuth2认证失败问题。\n- 数据集中多条件元数据搜索中的逻辑错误。\n- 在多轮对话中，引用数量无限增加。\n\n\n## 变更内容\n* 功能：通过URL使旧页面可访问 #3221，由@cike8899在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9589中提出\n* 功能：0.20.x版本之后也应列出参考文献，由@huang-aoqin在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9582中提出\n* 还原“0.20.x版本之后也应列出参考文献”功能，由@KevinHuSh在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9592中提出\n* 修复（Web）：更新分段控件样式并添加金属质感渐变，由@dcc123456在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9591中提出\n* 修复：KeyError: 'globals' KeyError： 'globals'，由@Woody-Hu在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9571中提出\n* 修复：更新助手时传递空数据库数组，由@jiasu-hezhip在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9570中提出\n* 修复：修复聊天页面底部保存按钮在小屏幕上无法显示的问题 #3221，由@cike8899在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9596中提出\n* 修复（Web）：优化文本显示效果 #3221，由@dcc123456在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9594中提出\n* 功能：根据设计稿更新部分颜色 #3221，由@cike8899在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9599中提出\n* 修复：扩大Raptor超时限制，由@KevinHuSh在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9600中提出\n* 功能：扩展MCP服务器的功能，由@yongtenglei在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8707中提出\n* 功能（Web）：优化数据集页面和分段组件 #3221，由@dcc123456在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9605中提出\n* 功能：高级Markdown解析，由@yongtenglei在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9607中提出\n* 修复：移除上传控件的文件大小和数量限制 #9613 #9598，由@cike8899在https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9618中提出\n* 修复：角色重复……，由@KevinHuS提出","2025-08-27T10:56:25",{"id":242,"version":243,"summary_zh":244,"released_at":245},81254,"v0.20.3","# Summary\r\n\r\n## Improvements\r\n- Revamps the user interface for the Datasets, Chat, and Search pages.\r\n- Search and Chat: Introduces document-level metadata filtering, allowing automatic or manual filtering during chats or searches.\r\n- Search: Supports creating search apps tailored to various business scenarios\r\n- Chat: Supports comparing answer performance of up to three chat model settings on a single Chat page.\r\n- Agent: \r\n  - Implements a toggle in the Agent component to enable or disable citation. \r\n  - Introduces a drag-and-drop method for creating components. \r\n- Documentation: Corrects inaccuracies in the API reference.\r\n\r\n## New Agent templates\r\n- Report Agent: A template for generating summary reports in internal question-answering scenarios, supporting the display of tables and formulae. [#9427](https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9427) (The template is now in active use at a research institution.)\r\n\r\n## Fixed issues\r\n- The timeout mechanism introduced in v0.20.0 caused tasks like GraphRAG to halt.\r\n- Predefined opening greeting in the Agent component was missing during conversations. \r\n- An automatic line break issue in the prompt editor. \r\n- A memory leak issue caused by PyPDF. [#9469](https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9469)\r\n\r\n## What's Changed\r\n* Doc: updated supported model name by @writinwaters in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9343\r\n* Feat: New search page components and features  by @dcc123456 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9344\r\n* Feat: Modify the agent list return field name #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9373\r\n* Support Russian in UI by @rjohny55 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9362\r\n* Support Russian in UI Update config.ts by @rjohny55 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9361\r\n* Feat: Run eslint when the project is running to standardize everyone's code #9377 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9379\r\n* Fix: Replace StrEnum with strenum in code_exec.py by @asiroliu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9376\r\n* Agent plans tasks by referring to its own prompt. by @TeslaZY in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9315\r\n* Fix broken data stream when writing image file by @mustangxu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9354\r\n* Fix \"no `tc` element at grid_offset\" by @mustangxu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9375\r\n* Feat: add full list of supported AWS Bedrock regions by @yongtenglei in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9378\r\n* Fix: Remove default dataset_ids from Chat class initialization by @asiroliu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9381\r\n* Fix: Update test assertions and add GraphRAG config in dataset tests by @asiroliu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9386\r\n* Fix: handle missing dataset_ids when creating chat assistant by @jiasu-hezhip in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9324\r\n* Support the case of one cell split by multiple columns. by @HaiyangPeng in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9225\r\n* docs(sandbox): update \u002Fetc\u002Fhosts entry to include required services by @BDanial in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9144\r\n* Feat: added functionality to choose all datasets if no id is provided by @Drasek in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9184\r\n* Feat:  Allow chat to use meta data #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9393\r\n* Python SDK: add meta_fields to Document class by @shasha79 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9387\r\n* Fix: Update test assertions and simplify test cases by @asiroliu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9400\r\n* Refa: list canvas by @yongtenglei in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9341\r\n* Refa: migrate chat models to LiteLLM by @yongtenglei in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9394\r\n* Feat: add full list of supported AWS Bedrock regions by @rjohny55 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9395\r\n* feat(next-search): Added AI summary functionality #3221 by @dcc123456 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9402\r\n* Fix:HTTP request component failed to retrieve the corresponding value by @Woody-Hu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9399\r\n* Add fallback to use 'calamine' parse engine in excel_parser.py by @mustangxu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9374\r\n* Fix: Before executing the SQL, remove tags in the format [ID: number] to avoid execution errors. by @goodboyasia in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9326\r\n* Feat: add meta data filter. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9405\r\n* Fix:TypeError: OllamaCV.chat() got an unexpected keyword argument 'stop' by @Woody-Hu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9363\r\n* Fix \"File contains no valid workbook part\" by @mustangxu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9360\r\n* Fix: Patch LiteLLM by @yongtenglei in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9416\r\n* F","2025-08-20T03:08:52",{"id":247,"version":248,"summary_zh":79,"released_at":249},81255,"v0.20.2","2025-08-19T11:33:43",{"id":251,"version":252,"summary_zh":253,"released_at":254},81256,"v0.20.1","Release v0.20.1 created from 9b026fc5b6f234a3abdf954f359bc2f9e99d9ba8 at 2025-08-08 18:51:10+08:00\r\n\r\n# Summary\r\n\r\n## New Features\r\n- The Retrieval components now support the dynamic specification of knowledge base names using variables.\r\n- The user interface now includes a French language option.\r\n## Added Models\r\n- GPT 5\r\n- Claude 4.1\r\n## New agent Templates (both workflow and agentic)\r\n- SQL Assistant Workflow: Empowers non-technical teams (e.g., operations, product) to independently query business data. \r\n- Choose Your Knowledge Base Workflow: Lets users select a knowledge base to query during conversations. [#9325](https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9325)\r\n- Choose Your Knowledge Base Agent: Delivers higher-quality responses with extended reasoning time, suited for complex queries. [#9325](https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9325)\r\n## Fixed Issues\r\n- The Agent component was unable to invoke models installed via vLLM.\r\n- Agents could not be shared with the team.\r\n- Embedding an Agent into a webpage was not functioning properly.\r\n\r\n## What's Changed\r\n* Feat: Add the migration script and its doc, added `backup` as default… by @Treedy2020 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8245\r\n* Fix: list tags api by using tenant id instead of user id by @whhe in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9103\r\n* Fix:Error when parsing files using Gemini: **ERROR**: GENERIC_ERROR - Unknown field for GenerationConfig: max_tokens by @Woody-Hu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9195\r\n* Fix: enlarge the timeout limits. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9201\r\n* Fix:Repeated images when parsing markdown files with images by @Woody-Hu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9196\r\n* Refactor: support config strong test by @Woody-Hu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9198\r\n* Fix: migrate deprecated Langfuse API from v2 to v3 by @yongtenglei in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9204\r\n* docs(agent): Correct several spelling errors, such as: Ouline -> Outline by @Kennytian in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9188\r\n* fix #8424 NPE in dify_retrieval.py, add log exception by @mustangxu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9212\r\n* fix \"TypeError: '\u003C' not supported between instances of 'Emu' and 'Non… by @mustangxu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9209\r\n* fix \"out of memory\" if slide.get_thumbnail() to a huge image by @mustangxu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9211\r\n* Feat: New Agent startup parameters add knowledge base parameter #9194 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9210\r\n* Fix: Web Server Accepts Invalid Data That Could Cause Problems in uv.lock by @kira-offgrid in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8966\r\n* Feat: list documents supports range filtering by @yongtenglei in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9214\r\n* Fix: Fixed the issue where the prompt word edit box had no scroll bar #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9215\r\n* Fix: Fixed the loss of Await Response function on the share page and other style issues #3221 by @dcc123456 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9216\r\n* v0.20.0 release notes by @writinwaters in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9218\r\n* Fix: Fixed the issue where the agent's chat box could not automatically scroll to the bottom #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9219\r\n* Fix: correct single base64 image handling in image prompt by @yongtenglei in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9220\r\n* Fix: Add prompt text to the form in the MCP module by @dcc123456 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9222\r\n* Perf Impr: Decouple reasoning and extraction for faster, more precise logic by @es-zx in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9191\r\n* Fix: self-deployed LLM error, by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9217\r\n* Refa: ensure Redis stream queue could be created properly by @yongtenglei in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9223\r\n* Docs: Updated tips for max rounds by @writinwaters in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9235\r\n* Fix:local variable 'response' referenced before assignment by @Woody-Hu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9230\r\n* Fix: Set the maximum number of rounds for the agent to 1 #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9238\r\n* Fix: fix mismatch of assitant message and its reference by @sinopec in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9233\r\n* Refactor:Improve the chat stream logic for NvidiaCV by @Woody-Hu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9242\r\n* Fix: update broken agent OpenAI-Compatible completion due to v0.20.0 changes by @yongtenglei in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9241\r\n* Feat: Render dialog list #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F9249\r\n* Fix: PlainParser using fix in presentation by @yzztin in https:","2025-08-08T10:51:11",{"id":256,"version":257,"summary_zh":258,"released_at":259},81257,"v0.20.0","Released on August 1, 2025.\r\n\r\n## Features and improvements\r\n\r\n- Unified orchestration of both Agents and Workflows.\r\n- A comprehensive refactor of the Agent, greatly enhancing its capabilities and usability, with support for Multi-Agent configurations, planning and reflection, and visual functionalities.\r\n- Fully realized MCP functionality, allowing for MCP Server import, Agents functioning as MCP Clients, and RAGFlow itself operating as an MCP Server.\r\n- Access to runtime logs for Agents.\r\n- Chat histories with Agents available through the management panel.\r\n- Integration of a new, more robust version of Infinity, enabling the auto-tagging functionality with Infinity as the underlying document engine.\r\n- An OpenAI-compatible API that supports file reference information.\r\n- Support for new models, including Kimi K2, Grok 4, and Voyage.\r\n\r\n## 10 new Agent templates are now available:\r\n\r\n- Multi-Agent based Deep Research: Collaborative Agent teamwork led by a Lead Agent with multiple Subagents, distinct from traditional workflow orchestration.\r\n- An intelligent Q&A chatbot leveraging internal knowledge bases, designed for customer service and training scenarios.\r\n- A resume analysis template used by the RAGFlow team to screen, analyze, and record candidate information.\r\n- A blog generation workflow that transforms raw ideas into SEO-friendly blog content.\r\n- An intelligent customer service workflow.\r\n- A user feedback analysis template that directs user feedback to appropriate teams through semantic analysis.\r\n- Trip Planner: Uses web search and MCP maps to assist with travel planning.\r\n- Image Lingo: Translates content from uploaded photos.\r\n- An information search assistant that retrieves answers from both internal knowledge bases and the web.\r\n\r\n## Breaking changes\r\nFrom v0.20.0 onwards, Agents are no longer compatible with earlier versions, and all existing Agents from previous versions must be rebuilt following the upgrade.\r\n\r\n## What's Changed\r\n* Feat: Add Tavily operator #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8400\r\n* Docs: Added v0.19.1 release notes by @writinwaters in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8398\r\n* Fix: Document parse via API will alot problen by @Woody-Hu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8407\r\n* Fix: rank feature score should be greater than 0. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8416\r\n* Feat: Solved the conflict between the Handle click and drag events of the canvas node #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8413\r\n* Fix: doc_aggs issue. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8418\r\n* Fix: Database Query Vulnerable to Injection Attacks in rag\u002Futils\u002Fopendal_conn.py by @kira-offgrid in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8408\r\n* Feat: The delete button is displayed only when the cursor is hovered over the connection line #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8422\r\n* Feat: Add HTTPS setup instructions and configuration for Nginx by @yesidc in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8401\r\n* Fix: AzureChat model code by @asiroliu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8426\r\n* Update .env ,Defaults to the v0.19.1-slim edition by @jason-this in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8412\r\n* Refa: upgrade MCP SDK to v1.9.4 by @yongtenglei in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8421\r\n* Docs: Miscellaneous editorial updates by @writinwaters in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8430\r\n* Feat: Scratch MCP tool calling support. by @notsyncing in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8263\r\n* Feat: Use the message_id returned by the interface as the id of the reply message #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8434\r\n* fix a bug when using huggingface embedding api by @Rainman5042 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8432\r\n* Feat: Delete the agent and tool nodes downstream of the agent node #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8450\r\n* Refactor:Improve insert file logic by @Woody-Hu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8445\r\n* Test: Add document app tests by @asiroliu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8456\r\n* FIX:Saving an RGBA image directly as JPEG will cause an error. If the… by @WuWeiFlow in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8399\r\n* Feat: Add IterationNode component #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8461\r\n* Fix the bug of long type value overflow by @HaiyangPeng in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8313\r\n* Docs: exporting created knowledge graphs is not supported by @writinwaters in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8465\r\n* Feat: add MCP dashboard operations by @yongtenglei in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8460\r\n* Fix: some cases Task return but not set progress by @Woody-Hu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8469\r\n* Feat: add MCP treamable-http transpo","2025-08-01T13:59:25",{"id":261,"version":262,"summary_zh":263,"released_at":264},81258,"v0.19.1","## v0.19.1\r\n\r\nReleased on June 23, 2025.\r\n\r\n### Fixed issues\r\n\r\n- A memory leak issue during high-concurrency requests.\r\n- Fixed a freezing issue when parsing large files with GraphRAG entity resolution enabled. [#8223](https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8223)\r\n- A context error occurring when using Sandbox in standalone mode. [#8340](https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8340)\r\n- An excessive CPU usage issue caused by Ollama. [#8216](https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8216)\r\n- A bug in the Code Component. [#7949](https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7949)\r\n\r\n### Improvements\r\n\r\n- Added support for Qwen 3 Embedding. [#8184](https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8184) \r\n- Added support for Voyage Multimodal 3. [#7987](https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7987)\r\n- Enabled role-based authentication for S3 bucket access. [#8149](https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8149) \r\n- Added support for models installed via Ollama or VLLM when creating a knowledge base through the API. [#8069](https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F8069)\r\n\r\n## What's Changed\r\n* Feat: Migrate the code operator to the new agent. #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7731\r\n* Feat: Render the agent list page by page #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7736\r\n* Feat: Rename agent #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7740\r\n* 0519 pdfparser by @writinwaters in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7747\r\n* Feat: delete useless image blobs when task executor meet edge cases by @Woody-Hu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7727\r\n* Feat: Fixed the issue where the page would refresh continuously when opening the sheet on the right side of the canvas #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7756\r\n* Feat: Switching the programming language of the code operator will switch the corresponding language template #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7770\r\n* Feat: Add return value widget  to CodeForm #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7776\r\n* Fix: improve task cancel lag by @Woody-Hu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7765\r\n* Fix: resolve regex library warnings by @emmanuel-ferdman in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7782\r\n* fix：When creating a new assistant, an avatar was uploaded, but when selecting the assistant to start a new chat, the default avatar still appears in the chat window instead of the one uploaded during creation by @kunger309 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7769\r\n* Feat: Refactor BeginForm with shadcn #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7792\r\n* Feat: Verify the parameters of the begin operator #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7794\r\n* Docs: RAGFlow does not suppport batch metadata setting by @writinwaters in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7795\r\n* Fix baidu request error by @gegef2009 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7799\r\n* Feat: Synchronize BeginForm's query data to the canvas #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7798\r\n* Fix backquotes in text2sql causing execution errors by @Sol-Ghf in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7793\r\n* Feat: Reconstruct the QueryTable of BeginForm using shandcn #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7807\r\n* Docs: Added instructions on cross-language search by @writinwaters in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7812\r\n* Feat: Translate the begin operator #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7811\r\n* Fix: Improve First Chunk Size by @Woody-Hu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7806\r\n* Feat: add code_executor_manager by @yongtenglei in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7814\r\n* Update rag\u002Fnlp\u002Fquery.py by @Sol-Ghf in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7816\r\n* Feat: Refactor the MessageForm with shadcn #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7820\r\n* Fix: Fixed the issue that the script text of the code operator is not displayed after refreshing the page after saving the script text of the code operator #4977 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7825\r\n* Delete Corresponding Minio Bucket When Deleting a Knowledge Base by @Hank-MIA in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7841\r\n* Fix: list out of boundary by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7843\r\n* Fix:Setting the message_history_window_size to 0 does not take effect by @Woody-Hu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7842\r\n* Fix: restore task limiter. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7844\r\n* Feat: Upgrade react-hook-form to the latest version to solve the problem that appending a useFieldArray entry cannot trigger the watch callback function #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002F","2025-06-20T10:00:06",{"id":266,"version":267,"summary_zh":268,"released_at":269},81259,"v0.19.0","Released on May 26, 2025.\r\n\r\n### New features\r\n\r\n- Cross-language search is supported in the Knowledge and Chat modules, enhancing search accuracy and user experience in multilingual environments, such as in Chinese-English knowledge bases.\r\n- Agent component: A new Code component supports Python and JavaScript scripts, enabling developers to handle more complex tasks like dynamic data processing.\r\n- Enhanced image display: Images in Chat and Search now render directly within responses, rather than as external references. Knowledge retrieval testing can retrieve images directly, instead of texts extracted from images.\r\n - Claude 4 and ChatGPT o3: Developers can now use the newly released, most advanced Claude model alongside OpenAI’s latest ChatGPT o3 inference model.\r\n> The following features are contributed by our community contributors:\r\n\r\n- Agent component: Enables tool calling within the Generate Component. Kudos to [notsyncing](https:\u002F\u002Fgithub.com\u002Fnotsyncing).\r\n- Markdown rendering: Image references in a markdown file can be displayed after chunking. Kudos to [Woody-Hu](https:\u002F\u002Fgithub.com\u002FWoody-Hu).\r\n- Document engine support: OpenSearch can now be used as RAGFlow's document engine. Kudos to [pyyuhao](https:\u002F\u002Fgithub.com\u002Fpyyuhao).\r\n\r\n## What's Changed\r\n* Feat: Create a folder #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7228\r\n* Feat: Deleting files in batches. #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7234\r\n* Fix\u002Fmcp doc by @yxkong in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7239\r\n* Fix: improve retrieval API by @Woody-Hu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6744\r\n* Fix: mcp server start by @treywelsh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7251\r\n* chore: adds pre-commit by @wanpdsantos in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7242\r\n* \u003Cthink> tag is missing. by @White-Bear1122 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7256\r\n* fix: task_executor bug fix by @benni82 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7253\r\n* Feat: Display document parsing status #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7241\r\n* Feat: Delete and rename files in the knowledge base #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7268\r\n* Feat: Adds OpenSearch2.19.1 as the vector_database support by @pyyuhao in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7140\r\n* Docs: How to disable user registration by @writinwaters in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7265\r\n* Feat: Replace the logo of novita by @neal-davis in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7287\r\n* Fix: Qwen-vl-plus url error by @yongtenglei in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7281\r\n* Doc: Updated sharing behavior in the open-source editions. by @writinwaters in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7293\r\n* Feat: Display the document configuration dialog with shadcn #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7302\r\n* Fix pymysql interface error by @Yue-Lyu123 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7295\r\n* Fix: LLM generated tag issue. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7301\r\n* Feat: Save the configuration information of the knowledge base document #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7317\r\n* Feat: Markdown add image by @Woody-Hu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7124\r\n* Feat: Save document metadata #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7323\r\n* Fix: Creating Knowledge Base Support Enter Key by @Woody-Hu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7258\r\n* DOC: Miscellaneous UI and editorial updates by @writinwaters in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7324\r\n* Feat: Filter document by running status and file type. #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7340\r\n* Feat: Create empty document. #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7343\r\n* [BREAKING CHANGE] GET to POST: enhance document list capability by @yongtenglei in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7349\r\n* Updated RAPTOR-specific UI by @writinwaters in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7348\r\n* Feat: Batch operations on documents in a dataset #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7352\r\n* Update Groq AI Model Config by @wemod123 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7335\r\n* Fix: tenant_id spelling error. by @xiaosl-cell in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7331\r\n* Fix mcp server startup failure by @yiminghub2024 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7329\r\n* Feat: Preview the file #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7355\r\n* perf: Optimize GraphRAG’s LOOP_PROMPT by @liuzhenghua in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7356\r\n* feat: replace models of novita by @neal-davis in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7360\r\n* Fix: Add title_tks for Pictures by @Woody-Hu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F7","2025-05-23T12:17:18",{"id":271,"version":272,"summary_zh":273,"released_at":274},81260,"v0.18.0","Release v0.18.0 created from 94181a990b957ed302952b4de17583d2b44f3099 at 2025-04-23 14:11:59+08:00\r\n\r\n## New features\r\n\r\n- MCP server: enables access to RAGFlow's knowledge bases via MCP.\r\n- DeepDoc supports adopting VLM model as a processing pipeline during document layout recognition, enabling in-depth analysis of images in PDFs.\r\n- Agent version control: all updates are continuously logged and can be rolled back to a previous version via export.\r\n- Team collaboration: Agents can be shared with team members.\r\n- OpenAI-compatible APIs: Agents can be called via OpenAI-compatible APIs.\r\n- User registration control: administrators can enable or disable user registration through an environment variable.\r\n\r\n## Improvements\r\n\r\n- Enhanced answer referencing: Citation accuracy in generated responses is improved.\r\n- Enhanced question-answering experience: users can now manually stop streaming output during a conversation.\r\n\r\n## Compatibility changes\r\n\r\n- From this release on wards, built-in re-rank models have been removed because they have minimal impact on retrieval rates but significantly increase retrieval time.\r\n\r\n## What's Changed\r\n* Test: Added test cases for Download Documents HTTP API by @asiroliu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6032\r\n* Fix: optimize OCR garbage identification to reduce unnecessary filtering by @yongtenglei in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6027\r\n* Fix: chat_completion answer data incorrect by @alen08266 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6041\r\n* Improve API Documentation, Standardize Error Handling, and Enhance Comments by @PStarH in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5990\r\n* Fix: Reduce excessive IO operations by loading LLM factory configurations by @utopia2077 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6047\r\n* Feat: Add deepseek to llm_factories by @c-goosen in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6051\r\n* Fix 6030 by @yuzhichang in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6070\r\n* Fix: empty doc id validation. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6064\r\n* Feat: Change “Document parser” to \"PDF parser\" #6072 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6073\r\n* Feat: add token comsumption & speed to little lamp. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6077\r\n* Refa: remove max toekns for image2txt models. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6078\r\n* Fix: remove NaN output of components. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6079\r\n* Fix executor name by @yuzhichang in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6080\r\n* Fix: Prevent password boxes other than login passwords from displaying passwords saved in the browser's password manager by default. #6033 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6084\r\n* Doc: update launch from source. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6074\r\n* Fix: check the file name length. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6083\r\n* Refa: enlarge default max request body size. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6088\r\n* Fix: none parse_config updating. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6092\r\n* Fix: Knowledge base page cannot upload folders #6062 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6096\r\n* Fix: fix document concurrent upload issue by @yongtenglei in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6095\r\n* Refa: enlarge http body size. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6100\r\n* Fix: limit the depth of DFS by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6101\r\n* Regards kb_id at ElasticSearch insert, update, delete. by @yuzhichang in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6105\r\n* Fix: empty tag field issue. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6103\r\n* Fix: Fixed the issue that events cannot be triggered after the shadcn-ui dialog is closed #3221. by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6108\r\n* Introduced task priority by @yuzhichang in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6118\r\n* Fix: update link of deploy_local_llm.mdx by @jingfelix in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6110\r\n* Test: Added test cases for Update Documents HTTP API by @asiroliu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6106\r\n* Doc: Update readme document by @zmxccxy in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6052\r\n* Test: update test cases per pr #6095 to fix issue #6039 by @asiroliu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6143\r\n* Feat: Limit view with more knowledge when list knowledge so many by @isthaison in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6093\r\n* Feat: add OCR's muti-gpus and parallel processing support by @debugdoctor in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5972\r\n* Fix: validation of readonly fields. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6144\r\n* Fix: retrieval debug mode is","2025-04-23T06:12:59",{"id":276,"version":277,"summary_zh":278,"released_at":279},81261,"v0.17.2","### Improvements\r\n\r\n- Adds OpenAI-compatible APIs.\r\n- Introduces a German user interface.\r\n- Accelerates knowledge graph extraction.\r\n- Enables Tavily-based web search in the **Retrieval** agent component.\r\n- Adds Tongyi-Qianwen QwQ models (OpenAI-compatible).\r\n- Supports CSV files in the **General** chunk method.\r\n\r\n### Fixed issues\r\n\r\n- Unable to add models via Ollama\u002FXinference, an issue introduced in v0.17.1.\r\n\r\n### Related APIs\r\n\r\n#### HTTP APIs\r\n\r\n[Create chat completion](.\u002Freferences\u002Fhttp_api_reference.md#openai-compatible-api)\r\n\r\n#### Python APIs\r\n\r\n[Create chat completion](.\u002Freferences\u002Fpython_api_reference.md#openai-compatible-api)\r\n\r\n## What's Changed\r\n* Docs: updates issue templates. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5913\r\n* Reverted some of the version changes by @writinwaters in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5908\r\n* Docs: reformat. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5914\r\n* Feat: Add german translations by @HashtagAssist in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5866\r\n* Docs: updates. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5921\r\n* Optimized graphrag again by @yuzhichang in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5927\r\n* TEST: Added test cases for Update Dataset HTTP API by @asiroliu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5924\r\n* Feat: Add Breadcrumb component #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5929\r\n* Fix: When calling the Create chat completion API, the response data… by @alen08266 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5928\r\n* Feat: apply LLM to optimize citations. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5935\r\n* Fix:signal.SIGUSR1 and signal.SIGUSR2 can't use in window. so don't bind signal.SIGUSR1 and signal.SIGUSR2 in the windows env by @donblack01 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5941\r\n* Fix: HTTP API Updates Read-Only Dataset Fields During Modification #5923  by @RaghavPatidar18 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5937\r\n* Fix: ollama model add error. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5947\r\n* Fix: empty value updating. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5949\r\n* Refa: Optimization of the Deep Research Module Code Structure by @liwenju0 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5959\r\n* Fixed a Docusaurus display issue. by @writinwaters in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5969\r\n* Fix: Add Knowledge Base Document Parsing Status Check by @liwenju0 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5966\r\n* TEST: Remove unstable assertion introduced in PR #5924 by @asiroliu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5968\r\n* Feat: `Retrieval` supports internet search. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5974\r\n* For an Agent with an Input Begin value, on the first call the return … by @isthaison in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5957\r\n* Feat: Why can't Retrieval component support internet web search. #5973 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5978\r\n* Feat: Add qwq model support to Tongyi-Qianwen factory by @kuro5989 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5981\r\n* Feat: Alter Item to TransferListItemType #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5986\r\n* Feat: add use KG to retrieval component. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5988\r\n* Feat: add CSV file parsing support by @yongtenglei in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5989\r\n* TEST: Added test cases for Upload Documents HTTP API by @asiroliu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5991\r\n* Refa: refine the context window size warning. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5993\r\n* Fix: encode detect error. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6006\r\n* Test: update test cases per issue #5920 #5923 by @asiroliu in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6007\r\n* Optimize graphrag cache get entity by @yuzhichang in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6018\r\n* 0.17.1 release notes by @writinwaters in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6021\r\n* Feat: Set the default value of Chunk token number to 512 #6016 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6017\r\n* Fix: empty api-key causes problems. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6022\r\n* DOCS: for release. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6023\r\n* Added 0.17.2 release notes by @writinwaters in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F6028\r\n\r\n## New Contributors\r\n* @HashtagAssist made their first contribution in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5866\r\n* @alen08266 made their first contribution in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5928\r\n* @kuro5989 made their first contribution in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5981\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fcompare\u002Fv0.17.1.","2025-03-13T08:12:47",{"id":281,"version":282,"summary_zh":283,"released_at":284},81262,"v0.17.1","## Improvements\r\n\r\n- Improves English tokenization quality.\r\n- Improves the table extraction logic in Markdown document parsing.\r\n- Updates SiliconFlow's model list.\r\n- Supports parsing XLS files (Excel97~2003) with improved corresponding error handling.\r\n- Supports Huggingface rerank models.\r\n- Enables relative time expressions (\"now\", \"yesterday\", \"last week\", \"next year\", and more) in the **Rewrite** agent component.\r\n\r\n## Fixed issues\r\n\r\n- A repetitive knowledge graph extraction issue.\r\n- Issues with API calling.\r\n- Options in the **Document parser** dropdown are missing.\r\n- A Tavily web search issue.\r\n- Unable to preview diagrams or images in an AI chat.\r\n\r\n## What's Changed\r\n* Fix: possible memory leaks close #5277 by @yihong0618 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5500\r\n* Fix email validation regex by @felipehertzer in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5533\r\n* Fix: in case running KG repeatly. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5538\r\n* Optimize prompt. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5541\r\n* Feat: Add NextLLMSelect with shadcn-ui. #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5542\r\n* Fix: bad case for tokenizer. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5543\r\n* Restructured guides by @writinwaters in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5549\r\n* Feat: Combine Select and LlmSettingFieldItems into LLMSelect. #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5548\r\n* Restructured guides by @writinwaters in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5555\r\n* Feat: Render RewriteQuestionForm with shadcn-ui #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5551\r\n* Feat: add 'delete' for agent's sessions api and unify apis of agent sdk by @debugdoctor in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5525\r\n* Feat: Render QWeatherForm with shadcn-ui. #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5558\r\n* Feat: add toc to api doc by @hjlarry in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5552\r\n* Made task_executor async to speedup parsing by @yuzhichang in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5530\r\n* Feat: Render WikipediaForm and BaiduForm with shadcn-ui. #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5564\r\n* Minor: Fixed broken links by @writinwaters in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5565\r\n* Fix: LLM with ___ return cannot be deleted #5585 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5587\r\n* Perf: update novita.ai LLM library by @debugdoctor in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5574\r\n* Add separate API service resource to Helm chart by @sd109 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5572\r\n* Fix: fix may lose part of information of last stream chunck by @yongtenglei in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5584\r\n* Fix: remove KB id restriction while creating chat. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5588\r\n* Introduced jemalloc by @yuzhichang in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5590\r\n* Fix nursery.start_soon. Close #5575 by @yuzhichang in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5591\r\n* Fix: better start experience PYTHONPATH in shell by @yihong0618 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5593\r\n* Docs: about meta files in API reference. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5594\r\n* Feat: Render MessageForm with shadcn-ui. #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5596\r\n* Added 0.17.0 release notes by @writinwaters in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5608\r\n* Feat: Render DynamicCategorize with shadcn-ui. #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5610\r\n* Fix：Fix the bug of incorrectly gets the APIToken. by @kunyi-yang in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5597\r\n* use to_df replace to_pl when get infinity Result by @dwxiayi in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5604\r\n* fix: issue #5600 by @yihong0618 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5620\r\n* Fix: image file can't preview by @hjlarry in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5626\r\n* Refactor: Optimize error handling and support parsing of XLS(EXCEL97—2003) files. by @hy89 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5633\r\n* Minor: improve doc and rm unused file by @hjlarry in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5634\r\n* Fix: parameter error. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5641\r\n* Refactored DocumentService.update_progress by @yuzhichang in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5642\r\n* Feat: Upload file UI\u002FUX enhancements by @Omarlsg98 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5359\r\n* Feat: The parsing method is paper and needs to display Document parser. #5467 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5652\r\n* fix: issue #5600 by @GuoDapeng in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F5645\r\n* Fix: tavily search error. by @KevinHuSh i","2025-03-11T07:04:59",{"id":286,"version":287,"summary_zh":288,"released_at":289},81263,"v0.17.0","## Summary\r\n- AI chat: Implements Deep Research for agentic reasoning. To activate this, enable the **Reasoning** toggle under the **Prompt Engine** tab of your chat assistant dialogue.\r\n- AI chat: Leverages Tavily-based web search to enhance contexts in agentic reasoning. To activate this, enter the correct Tavily API key under the **Assistant Setting** tab of your chat assistant dialogue.\r\n- AI chat: Supports initiating a chat without specifying knowledge bases.\r\n- AI chat: HTML files can also be previewed and referenced, in addition to PDF files.\r\n- Dataset: Adds a **Document parser** dropdown menu to dataset configurations. This includes a DeepDoc model option, which is time-consuming, a much faster **naive** option (plain text), which skips DLR (Document Layout Recognition), OCR (Optimal Character Recognition), and TSR (Table Structure Recognition) tasks, and several currently *experimental* large model options.\r\n- Agent component: **(x)** or a forward slash `\u002F` can be used to insert available keys (variables) in the system prompt field of the **Generate** or **Template** component.\r\n- Object storage: Supports using Aliyun OSS (Object Storage Service) as a file storage option.\r\n- Models: Updates the supported model list for Tongyi-Qianwen, adding DeepSeek-specific models; adds ModelScope as a model provider.\r\n- APIs: Document metadata can be updated through an API.\r\n\r\n\r\n## What's Changed\r\n* More models for siliconflow. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4756\r\n* Config chat share by @isthaison in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4700\r\n* Fix Gemini chat issue. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4757\r\n* optimize TenantLLMService.increase_usage for \"can't update token usag… by @DavidSche in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4755\r\n* Refresh Gemini model list. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4780\r\n* Add a comment to valkey. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4783\r\n* Fix divided by zero issue. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4784\r\n* Increase ES update script length. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4785\r\n* Fix: The requested interface timeout will cause the page to crash #4787 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4788\r\n* Inner prompt parameter setting. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4806\r\n* Feat: Add VariablePickerMenuPlugin to select variables in the prompt text box by menu #4764 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4765\r\n* Feat: Fixed the issue where the prompt always displayed the initial value when switching between different generate operators #4764 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4808\r\n* Feat: Remove begin's width from agent templates #4764 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4809\r\n* Fix KG circle. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4823\r\n* Fix: Fail to open console with Firefox #4816 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4838\r\n* Reformat error message. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4829\r\n* Feat: Modify the Preset configurations item style to distinguish it from other fields #4844 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4845\r\n* Added a release notes by @writinwaters in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4848\r\n* Feat: Add  ChatCard #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4852\r\n* Make the update script shorter. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4854\r\n* Feat: Add Sessions component #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4865\r\n* fix categorize agent input content not format error by @WANGRUI-ZB in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4842\r\n* Feat: Add ChatBasicSetting component #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4876\r\n* Feat: Add hatPromptEngine component #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4881\r\n* Fix: Knowledge base page crashes when network connection is lost. #4894 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4895\r\n* Changed elasticsearch image url by @yuzhichang in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4897\r\n* fix: big xls file error by @SkyfireWXY in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4859\r\n* Remove use of eval() from operators.py by @panzi in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4888\r\n* Fix xinference chat role order issue. by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4898\r\n* Remove use of eval() from search.py by @panzi in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4887\r\n* Display thinking for deepseek r1 by @KevinHuSh in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4904\r\n* Feat: Add LlmSettingFieldItems component #3221 by @cike8899 in https:\u002F\u002Fgithub.com\u002Finfiniflow\u002Fragflow\u002Fpull\u002F4906\r\n* Fix: After deleting all conversati","2025-03-02T10:48:10"]