[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-docling-project--docling":3,"tool-docling-project--docling":62},[4,18,26,36,46,54],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":17},4358,"openclaw","openclaw\u002Fopenclaw","OpenClaw 是一款专为个人打造的本地化 AI 助手，旨在让你在自己的设备上拥有完全可控的智能伙伴。它打破了传统 AI 助手局限于特定网页或应用的束缚，能够直接接入你日常使用的各类通讯渠道，包括微信、WhatsApp、Telegram、Discord、iMessage 等数十种平台。无论你在哪个聊天软件中发送消息，OpenClaw 都能即时响应，甚至支持在 macOS、iOS 和 Android 设备上进行语音交互，并提供实时的画布渲染功能供你操控。\n\n这款工具主要解决了用户对数据隐私、响应速度以及“始终在线”体验的需求。通过将 AI 部署在本地，用户无需依赖云端服务即可享受快速、私密的智能辅助，真正实现了“你的数据，你做主”。其独特的技术亮点在于强大的网关架构，将控制平面与核心助手分离，确保跨平台通信的流畅性与扩展性。\n\nOpenClaw 非常适合希望构建个性化工作流的技术爱好者、开发者，以及注重隐私保护且不愿被单一生态绑定的普通用户。只要具备基础的终端操作能力（支持 macOS、Linux 及 Windows WSL2），即可通过简单的命令行引导完成部署。如果你渴望拥有一个懂你",349277,3,"2026-04-06T06:32:30",[13,14,15,16],"Agent","开发框架","图像","数据工具","ready",{"id":19,"name":20,"github_repo":21,"description_zh":22,"stars":23,"difficulty_score":10,"last_commit_at":24,"category_tags":25,"status":17},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,"2026-04-05T11:01:52",[14,15,13],{"id":27,"name":28,"github_repo":29,"description_zh":30,"stars":31,"difficulty_score":32,"last_commit_at":33,"category_tags":34,"status":17},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",158594,2,"2026-04-16T23:34:05",[14,13,35],"语言模型",{"id":37,"name":38,"github_repo":39,"description_zh":40,"stars":41,"difficulty_score":42,"last_commit_at":43,"category_tags":44,"status":17},8272,"opencode","anomalyco\u002Fopencode","OpenCode 是一款开源的 AI 编程助手（Coding Agent），旨在像一位智能搭档一样融入您的开发流程。它不仅仅是一个代码补全插件，而是一个能够理解项目上下文、自主规划任务并执行复杂编码操作的智能体。无论是生成全新功能、重构现有代码，还是排查难以定位的 Bug，OpenCode 都能通过自然语言交互高效完成，显著减少开发者在重复性劳动和上下文切换上的时间消耗。\n\n这款工具专为软件开发者、工程师及技术研究人员设计，特别适合希望利用大模型能力来提升编码效率、加速原型开发或处理遗留代码维护的专业人群。其核心亮点在于完全开源的架构，这意味着用户可以审查代码逻辑、自定义行为策略，甚至私有化部署以保障数据安全，彻底打破了传统闭源 AI 助手的“黑盒”限制。\n\n在技术体验上，OpenCode 提供了灵活的终端界面（Terminal UI）和正在测试中的桌面应用程序，支持 macOS、Windows 及 Linux 全平台。它兼容多种包管理工具，安装便捷，并能无缝集成到现有的开发环境中。无论您是追求极致控制权的资深极客，还是渴望提升产出的独立开发者，OpenCode 都提供了一个透明、可信",144296,1,"2026-04-16T14:50:03",[13,45],"插件",{"id":47,"name":48,"github_repo":49,"description_zh":50,"stars":51,"difficulty_score":32,"last_commit_at":52,"category_tags":53,"status":17},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",108322,"2026-04-10T11:39:34",[14,15,13],{"id":55,"name":56,"github_repo":57,"description_zh":58,"stars":59,"difficulty_score":32,"last_commit_at":60,"category_tags":61,"status":17},6121,"gemini-cli","google-gemini\u002Fgemini-cli","gemini-cli 是一款由谷歌推出的开源 AI 命令行工具，它将强大的 Gemini 大模型能力直接集成到用户的终端环境中。对于习惯在命令行工作的开发者而言，它提供了一条从输入提示词到获取模型响应的最短路径，无需切换窗口即可享受智能辅助。\n\n这款工具主要解决了开发过程中频繁上下文切换的痛点，让用户能在熟悉的终端界面内直接完成代码理解、生成、调试以及自动化运维任务。无论是查询大型代码库、根据草图生成应用，还是执行复杂的 Git 操作，gemini-cli 都能通过自然语言指令高效处理。\n\n它特别适合广大软件工程师、DevOps 人员及技术研究人员使用。其核心亮点包括支持高达 100 万 token 的超长上下文窗口，具备出色的逻辑推理能力；内置 Google 搜索、文件操作及 Shell 命令执行等实用工具；更独特的是，它支持 MCP（模型上下文协议），允许用户灵活扩展自定义集成，连接如图像生成等外部能力。此外，个人谷歌账号即可享受免费的额度支持，且项目基于 Apache 2.0 协议完全开源，是提升终端工作效率的理想助手。",100752,"2026-04-10T01:20:03",[45,13,15,14],{"id":63,"github_repo":64,"name":65,"description_en":66,"description_zh":67,"ai_summary_zh":67,"readme_en":68,"readme_zh":69,"quickstart_zh":70,"use_case_zh":71,"hero_image_url":72,"owner_login":73,"owner_name":74,"owner_avatar_url":75,"owner_bio":76,"owner_company":77,"owner_location":77,"owner_email":77,"owner_twitter":77,"owner_website":77,"owner_url":78,"languages":79,"stars":92,"forks":93,"last_commit_at":94,"license":95,"difficulty_score":32,"env_os":96,"env_gpu":97,"env_ram":98,"env_deps":99,"category_tags":104,"github_topics":105,"view_count":32,"oss_zip_url":77,"oss_zip_packed_at":77,"status":17,"created_at":121,"updated_at":122,"faqs":123,"releases":153},8213,"docling-project\u002Fdocling","docling","Get your documents ready for gen AI","Docling 是一款专为生成式 AI 打造的文档处理工具，旨在将各种复杂格式的文档转化为机器易于理解的结构化数据。它有效解决了传统方法难以精准解析 PDF 布局、表格结构、数学公式及扫描图片等痛点，让非结构化文档能无缝接入大模型应用。\n\n无论是开发者、数据科学家还是研究人员，都能利用 Docling 轻松构建高质量的 RAG（检索增强生成）系统或智能代理。其核心亮点在于强大的多格式支持，不仅涵盖 PDF、Office 文档、图片，甚至能处理音频和专利\u002F财务等专业 XML 标准。Docling 具备先进的版面分析能力，能准确识别阅读顺序与表格逻辑，并提供统一的文档表示格式。此外，它支持本地离线运行以保障数据安全，内置 OCR 功能处理扫描件，并能通过 Visual Language Models 深度理解视觉内容。配合 LangChain、LlamaIndex 等主流框架的即插即用集成，以及便捷的命令行工具，Docling 让文档预处理变得简单高效，是连接真实世界文档与生成式 AI 的理想桥梁。","\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\">\n    \u003Cimg loading=\"lazy\" alt=\"Docling\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdocling-project_docling_readme_5fd3dd08ce51.png\" width=\"100%\"\u002F>\n  \u003C\u002Fa>\n\u003C\u002Fp>\n\n# Docling\n\n\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Ftrendshift.io\u002Frepositories\u002F12132\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdocling-project_docling_readme_4a68feb902da.png\" alt=\"DS4SD%2Fdocling | Trendshift\" style=\"width: 250px; height: 55px;\" width=\"250\" height=\"55\"\u002F>\u003C\u002Fa>\n\u003C\u002Fp>\n\n[![arXiv](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FarXiv-2408.09869-b31b1b.svg)](https:\u002F\u002Farxiv.org\u002Fabs\u002F2408.09869)\n[![Docs](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdocs-live-brightgreen)](https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002F)\n[![PyPI version](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Fdocling)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fdocling\u002F)\n[![PyPI - Python Version](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fpyversions\u002Fdocling)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fdocling\u002F)\n[![uv](https:\u002F\u002Fimg.shields.io\u002Fendpoint?url=https:\u002F\u002Fraw.githubusercontent.com\u002Fastral-sh\u002Fuv\u002Fmain\u002Fassets\u002Fbadge\u002Fv0.json)](https:\u002F\u002Fgithub.com\u002Fastral-sh\u002Fuv)\n[![Ruff](https:\u002F\u002Fimg.shields.io\u002Fendpoint?url=https:\u002F\u002Fraw.githubusercontent.com\u002Fastral-sh\u002Fruff\u002Fmain\u002Fassets\u002Fbadge\u002Fv2.json)](https:\u002F\u002Fgithub.com\u002Fastral-sh\u002Fruff)\n[![Pydantic v2](https:\u002F\u002Fimg.shields.io\u002Fendpoint?url=https:\u002F\u002Fraw.githubusercontent.com\u002Fpydantic\u002Fpydantic\u002Fmain\u002Fdocs\u002Fbadge\u002Fv2.json)](https:\u002F\u002Fpydantic.dev)\n[![pre-commit](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fpre--commit-enabled-brightgreen?logo=pre-commit&logoColor=white)](https:\u002F\u002Fgithub.com\u002Fpre-commit\u002Fpre-commit)\n[![License MIT](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flicense\u002Fdocling-project\u002Fdocling)](https:\u002F\u002Fopensource.org\u002Flicenses\u002FMIT)\n[![PyPI Downloads](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdocling-project_docling_readme_57783af65f7f.png)](https:\u002F\u002Fpepy.tech\u002Fprojects\u002Fdocling)\n[![Docling Actor](https:\u002F\u002Fapify.com\u002Factor-badge?actor=vancura\u002Fdocling&fpr=docling)](https:\u002F\u002Fapify.com\u002Fvancura\u002Fdocling)\n[![Chat with Dosu](https:\u002F\u002Fdosu.dev\u002Fdosu-chat-badge.svg)](https:\u002F\u002Fapp.dosu.dev\u002F097760a8-135e-4789-8234-90c8837d7f1c\u002Fask?utm_source=github)\n[![Discord](https:\u002F\u002Fimg.shields.io\u002Fdiscord\u002F1399788921306746971?color=6A7EC2&logo=discord&logoColor=ffffff)](https:\u002F\u002Fdocling.ai\u002Fdiscord)\n[![OpenSSF Best Practices](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdocling-project_docling_readme_eda489f85045.png)](https:\u002F\u002Fwww.bestpractices.dev\u002Fprojects\u002F10101)\n[![LF AI & Data](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLF%20AI%20%26%20Data-003778?logo=linuxfoundation&logoColor=fff&color=0094ff&labelColor=003778)](https:\u002F\u002Flfaidata.foundation\u002Fprojects\u002F)\n\nDocling simplifies document processing, parsing diverse formats — including advanced PDF understanding — and providing seamless integrations with the gen AI ecosystem.\n\n## Features\n\n* 🗂️ Parsing of [multiple document formats][supported_formats] incl. PDF, DOCX, PPTX, XLSX, HTML, WAV, MP3, WebVTT, images (PNG, TIFF, JPEG, ...), LaTeX, plain text, and more\n* 📑 Advanced PDF understanding incl. page layout, reading order, table structure, code, formulas, image classification, and more\n* 🧬 Unified, expressive [DoclingDocument][docling_document] representation format\n* ↪️ Various [export formats][supported_formats] and options, including Markdown, HTML, WebVTT, [DocTags](https:\u002F\u002Farxiv.org\u002Fabs\u002F2503.11576) and lossless JSON\n* 📜 Support of several application-specifc XML schemas incl. [USPTO](https:\u002F\u002Fwww.uspto.gov\u002Fpatents) patents, [JATS](https:\u002F\u002Fjats.nlm.nih.gov\u002F) articles, and [XBRL](https:\u002F\u002Fwww.xbrl.org\u002F) financial reports.\n* 🔒 Local execution capabilities for sensitive data and air-gapped environments\n* 🤖 Plug-and-play [integrations][integrations] incl. LangChain, LlamaIndex, Crew AI & Haystack for agentic AI\n* 🔍 Extensive OCR support for scanned PDFs and images\n* 👓 Support of several Visual Language Models ([GraniteDocling](https:\u002F\u002Fhuggingface.co\u002Fibm-granite\u002Fgranite-docling-258M))\n* 🎙️ Audio support with Automatic Speech Recognition (ASR) models\n* 🔌 Connect to any agent using the [MCP server](https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Fusage\u002Fmcp\u002F)\n* 💻 Simple and convenient CLI\n\n### What's new\n* 📤 Structured [information extraction][extraction] \\[🧪 beta\\]\n* 📑 New layout model (**Heron**) by default, for faster PDF parsing\n* 🔌 [MCP server](https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Fusage\u002Fmcp\u002F) for agentic applications\n* 💼 Parsing of XBRL (eXtensible Business Reporting Language) documents for financial reports\n* 💬 Parsing of WebVTT (Web Video Text Tracks) files and export to WebVTT format\n* 💬 Parsing of LaTeX files\n* 📝 Parsing of plain-text files (`.txt`, `.text`) and Markdown supersets (`.qmd`, `.Rmd`)\n* 📝 Chart understanding (Barchart, Piechart, LinePlot): converting them into tables, code or adding detailed descriptions\n\n### Coming soon\n\n* 📝 Metadata extraction, including title, authors, references & language\n* 📝 Complex chemistry understanding (Molecular structures)\n\n## Installation\n\nTo use Docling, simply install `docling` from your package manager, e.g. pip:\n```bash\npip install docling\n```\n\n> **Note:** Python 3.9 support was dropped in docling version 2.70.0. Please use Python 3.10 or higher.\n\nWorks on macOS, Linux and Windows environments. Both x86_64 and arm64 architectures.\n\nMore [detailed installation instructions](https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Finstallation\u002F) are available in the docs.\n\n## Getting started\n\nTo convert individual documents with python, use `convert()`, for example:\n\n```python\nfrom docling.document_converter import DocumentConverter\n\nsource = \"https:\u002F\u002Farxiv.org\u002Fpdf\u002F2408.09869\"  # document per local path or URL\nconverter = DocumentConverter()\nresult = converter.convert(source)\nprint(result.document.export_to_markdown())  # output: \"## Docling Technical Report[...]\"\n```\n\nMore [advanced usage options](https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Fusage\u002Fadvanced_options\u002F) are available in\nthe docs.\n\n## CLI\n\nDocling has a built-in CLI to run conversions.\n\n```bash\ndocling https:\u002F\u002Farxiv.org\u002Fpdf\u002F2206.01062\n```\n\nYou can also use 🥚[GraniteDocling](https:\u002F\u002Fhuggingface.co\u002Fibm-granite\u002Fgranite-docling-258M) and other VLMs via Docling CLI:\n```bash\ndocling --pipeline vlm --vlm-model granite_docling https:\u002F\u002Farxiv.org\u002Fpdf\u002F2206.01062\n```\nThis will use MLX acceleration on supported Apple Silicon hardware.\n\nRead more [here](https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Fusage\u002F)\n\n## Documentation\n\nCheck out Docling's [documentation](https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002F), for details on\ninstallation, usage, concepts, recipes, extensions, and more.\n\n## Examples\n\nGo hands-on with our [examples](https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Fexamples\u002F),\ndemonstrating how to address different application use cases with Docling.\n\n## Integrations\n\nTo further accelerate your AI application development, check out Docling's native\n[integrations](https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Fintegrations\u002F) with popular frameworks\nand tools.\n\n## Get help and support\n\nPlease feel free to connect with us using the [discussion section](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fdiscussions).\n\n## Technical report\n\nFor more details on Docling's inner workings, check out the [Docling Technical Report](https:\u002F\u002Farxiv.org\u002Fabs\u002F2408.09869).\n\n## Contributing\n\nPlease read [Contributing to Docling](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fblob\u002Fmain\u002FCONTRIBUTING.md) for details.\n\n## References\n\nIf you use Docling in your projects, please consider citing the following:\n\n```bib\n@techreport{Docling,\n  author = {Deep Search Team},\n  month = {8},\n  title = {Docling Technical Report},\n  url = {https:\u002F\u002Farxiv.org\u002Fabs\u002F2408.09869},\n  eprint = {2408.09869},\n  doi = {10.48550\u002FarXiv.2408.09869},\n  version = {1.0.0},\n  year = {2024}\n}\n```\n\n## License\n\nThe Docling codebase is under MIT license.\nFor individual model usage, please refer to the model licenses found in the original packages.\n\n## LF AI & Data\n\nDocling is hosted as a project in the [LF AI & Data Foundation](https:\u002F\u002Flfaidata.foundation\u002Fprojects\u002F).\n\n### IBM ❤️ Open Source AI\n\nThe project was started by the AI for knowledge team at IBM Research Zurich.\n\n[supported_formats]: https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Fusage\u002Fsupported_formats\u002F\n[docling_document]: https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Fconcepts\u002Fdocling_document\u002F\n[integrations]: https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Fintegrations\u002F\n[extraction]: https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Fexamples\u002Fextraction\u002F\n","\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\">\n    \u003Cimg loading=\"lazy\" alt=\"Docling\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdocling-project_docling_readme_5fd3dd08ce51.png\" width=\"100%\"\u002F>\n  \u003C\u002Fa>\n\u003C\u002Fp>\n\n# Docling\n\n\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Ftrendshift.io\u002Frepositories\u002F12132\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdocling-project_docling_readme_4a68feb902da.png\" alt=\"DS4SD%2Fdocling | Trendshift\" style=\"width: 250px; height: 55px;\" width=\"250\" height=\"55\"\u002F>\u003C\u002Fa>\n\u003C\u002Fp>\n\n[![arXiv](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FarXiv-2408.09869-b31b1b.svg)](https:\u002F\u002Farxiv.org\u002Fabs\u002F2408.09869)\n[![Docs](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdocs-live-brightgreen)](https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002F)\n[![PyPI version](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Fdocling)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fdocling\u002F)\n[![PyPI - Python Version](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fpyversions\u002Fdocling)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fdocling\u002F)\n[![uv](https:\u002F\u002Fimg.shields.io\u002Fendpoint?url=https:\u002F\u002Fraw.githubusercontent.com\u002Fastral-sh\u002Fuv\u002Fmain\u002Fassets\u002Fbadge\u002Fv0.json)](https:\u002F\u002Fgithub.com\u002Fastral-sh\u002Fuv)\n[![Ruff](https:\u002F\u002Fimg.shields.io\u002Fendpoint?url=https:\u002F\u002Fraw.githubusercontent.com\u002Fastral-sh\u002Fruff\u002Fmain\u002Fassets\u002Fbadge\u002Fv2.json)](https:\u002F\u002Fgithub.com\u002Fastral-sh\u002Fruff)\n[![Pydantic v2](https:\u002F\u002Fimg.shields.io\u002Fendpoint?url=https:\u002F\u002Fraw.githubusercontent.com\u002Fpydantic\u002Fpydantic\u002Fmain\u002Fdocs\u002Fbadge\u002Fv2.json)](https:\u002F\u002Fpydantic.dev)\n[![pre-commit](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fpre--commit-enabled-brightgreen?logo=pre-commit&logoColor=white)](https:\u002F\u002Fgithub.com\u002Fpre-commit\u002Fpre-commit)\n[![License MIT](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flicense\u002Fdocling-project\u002Fdocling)](https:\u002F\u002Fopensource.org\u002Flicenses\u002FMIT)\n[![PyPI Downloads](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdocling-project_docling_readme_57783af65f7f.png)](https:\u002F\u002Fpepy.tech\u002Fprojects\u002Fdocling)\n[![Docling Actor](https:\u002F\u002Fapify.com\u002Factor-badge?actor=vancura\u002Fdocling&fpr=docling)](https:\u002F\u002Fapify.com\u002Fvancura\u002Fdocling)\n[![Chat with Dosu](https:\u002F\u002Fdosu.dev\u002Fdosu-chat-badge.svg)](https:\u002F\u002Fapp.dosu.dev\u002F097760a8-135e-4789-8234-90c8837d7f1c\u002Fask?utm_source=github)\n[![Discord](https:\u002F\u002Fimg.shields.io\u002Fdiscord\u002F1399788921306746971?color=6A7EC2&logo=discord&logoColor=ffffff)](https:\u002F\u002Fdocling.ai\u002Fdiscord)\n[![OpenSSF Best Practices](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdocling-project_docling_readme_eda489f85045.png)](https:\u002F\u002Fwww.bestpractices.dev\u002Fprojects\u002F10101)\n[![LF AI & Data](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLF%20AI%20%26%20Data-003778?logo=linuxfoundation&logoColor=fff&color=0094ff&labelColor=003778)](https:\u002F\u002Flfaidata.foundation\u002Fprojects\u002F)\n\nDocling 简化了文档处理流程，能够解析多种格式的文件——包括高级 PDF 解析——并提供与生成式 AI 生态系统的无缝集成。\n\n## 功能特性\n\n* 🗂️ 支持多种文档格式的解析 [supported_formats]，包括 PDF、DOCX、PPTX、XLSX、HTML、WAV、MP3、WebVTT、图像（PNG、TIFF、JPEG 等）、LaTeX、纯文本等\n* 📑 高级 PDF 解析能力，涵盖页面布局、阅读顺序、表格结构、代码、公式、图像分类等\n* 🧬 统一且表达力强的 [DoclingDocument][docling_document] 表示格式\n* ↪️ 多种导出格式及选项，包括 Markdown、HTML、WebVTT、[DocTags](https:\u002F\u002Farxiv.org\u002Fabs\u002F2503.11576) 和无损 JSON\n* 📜 支持多种特定于应用的 XML 模式，包括 [USPTO](https:\u002F\u002Fwww.uspto.gov\u002Fpatents) 专利、[JATS](https:\u002F\u002Fjats.nlm.nih.gov\u002F) 文章以及 [XBRL](https:\u002F\u002Fwww.xbrl.org\u002F) 财务报告。\n* 🔒 支持本地执行，适用于敏感数据和气隙环境\n* 🤖 即插即用的 [integrations] 集成，包括 LangChain、LlamaIndex、Crew AI 和 Haystack，用于代理型 AI 应用\n* 🔍 对扫描 PDF 和图像的广泛 OCR 支持\n* 👓 支持多种视觉语言模型（[GraniteDocling](https:\u002F\u002Fhuggingface.co\u002Fibm-granite\u002Fgranite-docling-258M)）\n* 🎙️ 支持音频处理，配备自动语音识别（ASR）模型\n* 🔌 可通过 [MCP 服务器](https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Fusage\u002Fmcp\u002F) 连接到任何代理\n* 💻 简单便捷的命令行界面\n\n### 最新功能\n* 📤 结构化 [信息提取][extraction] \\[🧪 测试版\\]\n* 📑 默认采用新的布局模型（**Heron**），以提升 PDF 解析速度\n* 🔌 [MCP 服务器](https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Fusage\u002Fmcp\u002F)，专为代理型应用设计\n* 💼 支持解析 XBRL（可扩展商业报告语言）文档，用于财务报告\n* 💬 支持解析 WebVTT（Web 视频文本轨道）文件，并导出为 WebVTT 格式\n* 💬 支持解析 LaTeX 文件\n* 📝 支持解析纯文本文件（`.txt`、`.text`）以及 Markdown 的扩展格式（`.qmd`、`.Rmd`）\n* 📝 图表理解（柱状图、饼图、折线图）：可将其转换为表格、代码，或添加详细描述\n\n### 即将推出\n* 📝 元数据提取，包括标题、作者、参考文献及语言\n* 📝 复杂化学结构解析（分子结构）\n\n## 安装\n要使用 Docling，只需从你的包管理器中安装 `docling`，例如 pip：\n```bash\npip install docling\n```\n\n> **注意：** Docling 2.70.0 版本已不再支持 Python 3.9。请使用 Python 3.10 或更高版本。\n\n支持 macOS、Linux 和 Windows 环境，兼容 x86_64 和 arm64 架构。\n\n更多详细的安装说明，请参阅文档：[安装指南](https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Finstallation\u002F)。\n\n## 快速入门\n要使用 Python 转换单个文档，可以使用 `convert()` 方法，例如：\n\n```python\nfrom docling.document_converter import DocumentConverter\n\nsource = \"https:\u002F\u002Farxiv.org\u002Fpdf\u002F2408.09869\"  # 文档路径或 URL\nconverter = DocumentConverter()\nresult = converter.convert(source)\nprint(result.document.export_to_markdown())  # 输出： \"## Docling Technical Report[...]\"\n```\n\n更多高级用法，请参阅文档：[高级用法](https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Fusage\u002Fadvanced_options\u002F)。\n\n## 命令行界面\nDocling 内置了命令行工具，可用于执行转换操作。\n\n```bash\ndocling https:\u002F\u002Farxiv.org\u002Fpdf\u002F2206.01062\n```\n\n你还可以通过 Docling 命令行界面使用 🥚[GraniteDocling](https:\u002F\u002Fhuggingface.co\u002Fibm-granite\u002Fgranite-docling-258M) 和其他 VLM 模型：\n```bash\ndocling --pipeline vlm --vlm-model granite_docling https:\u002F\u002Farxiv.org\u002Fpdf\u002F2206.01062\n```\n此命令将在支持 MLX 加速的 Apple Silicon 硬件上运行。\n\n更多信息请参阅：[命令行使用指南](https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Fusage\u002F)。\n\n## 文档\n请查阅 Docling 的[文档](https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002F)，了解安装、使用、概念、配方、扩展等内容的详细信息。\n\n## 示例\n通过我们的[示例](https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Fexamples\u002F)动手实践，了解如何使用 Docling 解决不同的应用场景。\n\n## 集成\n为进一步加速你的 AI 应用开发，请查看 Docling 与流行框架和工具的原生[集成](https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Fintegrations\u002F)。\n\n## 获取帮助和支持\n欢迎随时通过 [讨论区](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fdiscussions) 与我们联系。\n\n## 技术报告\n\n如需了解更多关于 Docling 内部工作机制的详细信息，请参阅 [Docling 技术报告](https:\u002F\u002Farxiv.org\u002Fabs\u002F2408.09869)。\n\n## 贡献\n\n有关详细信息，请阅读 [贡献指南](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fblob\u002Fmain\u002FCONTRIBUTING.md)。\n\n## 参考文献\n\n如果您在项目中使用了 Docling，请考虑引用以下内容：\n\n```bibtex\n@techreport{Docling,\n  author = {Deep Search Team},\n  month = {8},\n  title = {Docling Technical Report},\n  url = {https:\u002F\u002Farxiv.org\u002Fabs\u002F2408.09869},\n  eprint = {2408.09869},\n  doi = {10.48550\u002FarXiv.2408.09869},\n  version = {1.0.0},\n  year = {2024}\n}\n```\n\n## 许可证\n\nDocling 代码库采用 MIT 许可证。对于各个模型的使用，请参阅原始软件包中提供的模型许可证。\n\n## LF AI & Data\n\nDocling 是 [LF AI & Data 基金会](https:\u002F\u002Flfaidata.foundation\u002Fprojects\u002F) 的一个托管项目。\n\n### IBM ❤️ 开源人工智能\n\n该项目由 IBM 苏黎世研究院的知识型人工智能团队发起。\n\n[supported_formats]: https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Fusage\u002Fsupported_formats\u002F\n[docling_document]: https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Fconcepts\u002Fdocling_document\u002F\n[integrations]: https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Fintegrations\u002F\n[extraction]: https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Fexamples\u002Fextraction\u002F","# Docling 快速上手指南\n\nDocling 是一款强大的文档处理工具，支持解析 PDF、DOCX、PPTX、图片等多种格式，具备先进的 PDF 布局理解、表格识别及 OCR 能力，并可无缝集成到 Gen AI 生态中。\n\n## 环境准备\n\n*   **操作系统**：支持 macOS、Linux 和 Windows。\n*   **架构支持**：兼容 x86_64 和 arm64（Apple Silicon）。\n*   **Python 版本**：要求 **Python 3.10** 或更高版本（v2.70.0 起已停止支持 Python 3.9）。\n*   **前置依赖**：建议确保 `pip` 为最新版本。国内用户可配置清华源或阿里源以加速下载。\n\n## 安装步骤\n\n使用 pip 直接安装核心包：\n\n```bash\npip install docling\n```\n\n> **提示**：若需更快的下载速度，可使用国内镜像源：\n> ```bash\n> pip install docling -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n> ```\n\n如需更详细的安装选项（如特定硬件加速支持），请参考官方文档。\n\n## 基本使用\n\n### 1. Python 代码调用\n\n最简单的使用方式是通过 `DocumentConverter` 转换单个文档（支持本地路径或 URL）：\n\n```python\nfrom docling.document_converter import DocumentConverter\n\nsource = \"https:\u002F\u002Farxiv.org\u002Fpdf\u002F2408.09869\"  # 可以是本地文件路径或 URL\nconverter = DocumentConverter()\nresult = converter.convert(source)\n\n# 导出为 Markdown 格式\nprint(result.document.export_to_markdown())\n```\n\n### 2. 命令行工具 (CLI)\n\nDocling 内置了便捷的命令行工具，可直接在终端运行：\n\n**基础转换：**\n```bash\ndocling https:\u002F\u002Farxiv.org\u002Fpdf\u002F2206.01062\n```\n\n**使用视觉语言模型 (VLM)：**\n支持调用 GraniteDocling 等模型进行更高级的视觉理解（在 Apple Silicon 上自动启用 MLX 加速）：\n\n```bash\ndocling --pipeline vlm --vlm-model granite_docling https:\u002F\u002Farxiv.org\u002Fpdf\u002F2206.01062\n```\n\n转换完成后，文件将默认导出为 Markdown 或其他指定格式。更多高级配置选项请查阅官方文档。","某金融风控团队需要每天从数百份扫描版财报（PDF）和会议纪要（音频）中提取关键数据，以训练内部的风险预测大模型。\n\n### 没有 docling 时\n- 扫描版 PDF 中的表格被识别为乱码或纯文本，行列结构完全丢失，导致财务数据无法对齐。\n- 处理音频会议纪要需单独搭建语音识别流程，再人工将文字与文档内容拼接，耗时且易出错。\n- 复杂的页面布局（如双栏排版、页眉页脚）干扰阅读顺序，大模型常因上下文错乱而产生幻觉。\n- 不同格式（DOCX, PPTX, 图片）需要编写多套解析脚本，维护成本极高且难以统一输出标准。\n- 敏感财务数据必须上传至第三方云 API 处理，存在严重的数据合规与泄露风险。\n\n### 使用 docling 后\n- 利用高级 OCR 与布局分析能力，精准还原财报中的复杂表格结构，确保数值与表头准确对应。\n- 内置 ASR 模型直接处理音频文件，自动将会议语音转为文本并与相关文档内容统一整合。\n- 智能识别页面阅读顺序并过滤噪声，为大模型提供逻辑连贯的上下文，显著降低幻觉率。\n- 通过统一的 DoclingDocument 格式一站式解析多种文件类型，直接输出高质量的 Markdown 或 JSON。\n- 支持本地化部署，所有敏感数据均在内部服务器处理，完美满足金融行业的隐私合规要求。\n\ndocling 将杂乱无章的多模态文档转化为大模型可直接理解的结构化知识，让金融数据分析从“人工清洗”迈向“自动化智能”。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdocling-project_docling_14f758bc.png","docling-project","Docling Project","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fdocling-project_5c3ab177.png","",null,"https:\u002F\u002Fgithub.com\u002Fdocling-project",[80,84,88],{"name":81,"color":82,"percentage":83},"Python","#3572A5",99.3,{"name":85,"color":86,"percentage":87},"Shell","#89e051",0.6,{"name":89,"color":90,"percentage":91},"Dockerfile","#384d54",0.2,57957,3972,"2026-04-16T10:54:42","MIT","Linux, macOS, Windows","非必需。支持 Apple Silicon (M1\u002FM2\u002FM3) 通过 MLX 加速；若使用 NVIDIA GPU，具体型号和 CUDA 版本未在 README 中明确说明，但需支持 PyTorch 后端。","未说明",{"notes":100,"python":101,"dependencies":102},"Python 3.9 支持已在 v2.70.0 中移除。默认使用新的 Heron 布局模型以加快解析速度。支持本地离线运行以处理敏感数据。可选集成 GraniteDocling 等视觉语言模型 (VLM) 和自动语音识别 (ASR) 模型，这些额外模型可能需要更高的硬件资源。","3.10+",[65,103],"Pydantic v2",[15,13,14],[106,107,108,109,110,111,112,113,114,115,116,117,118,119,120],"ai","convert","documents","pdf","tables","document-parser","document-parsing","docx","html","markdown","pdf-converter","pdf-to-json","pdf-to-text","pptx","xlsx","2026-03-27T02:49:30.150509","2026-04-17T08:24:13.871945",[124,129,134,139,144,149],{"id":125,"question_zh":126,"answer_zh":127,"source_url":128},36751,"如何完全禁用 OCR 并阻止 Docling 自动下载模型？","要在禁用 OCR 的同时防止模型自动下载，需要先将所有必要模型下载到本地，然后在代码中指定本地路径。具体步骤如下：\n1. 使用命令行下载所有模型：\n   docling-tools models download --all -o .\u002Fmodels\n2. 在 Python 代码中配置 artifacts_path 指向本地模型目录：\n   from pathlib import Path\n   from docling.datamodel.base_models import InputFormat\n   from docling.datamodel.pipeline_options import PdfPipelineOptions\n   from docling.document_converter import DocumentConverter, PdfFormatOption\n   \n   local_models = Path(\".\u002Fmodels\")\n   pdf_options = PdfPipelineOptions(do_ocr=False, generate_parsed_pages=False)\n   \n   converter = DocumentConverter(\n       format_options={\n           InputFormat.PDF: PdfFormatOption(\n               pipeline_options=pdf_options,\n               artifacts_path=local_models\n           )\n       }\n   )","https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2312",{"id":130,"question_zh":131,"answer_zh":132,"source_url":133},36752,"运行示例时遇到 'MultiScaleDeformableAttention' 扩展构建错误怎么办？","该错误通常是因为缺少 CUDA 开发文件或依赖版本不匹配。解决方案包括：\n1. 确保安装了正确的开发包（如 python3.x-dev）和 CUDA 工具包。\n2. 推荐使用包含完整开发环境的 Docker 镜像，例如：nvidia\u002Fcuda:12.4.1-cudnn-devel-ubuntu22.04（注意需手动安装 Python）。\n3. 验证以下版本组合通常可解决问题：\n   - Python 3.12 + python3.12-dev\n   - torch 2.5.1 + torchvision 0.20.1\n   - 最新版本的 Docling\n   - NVIDIA CUDA 12.4.1 (或更高兼容版本)\n如果问题依旧，请检查 transformers 库的相关 issue 以获取更详细的编译环境要求。","https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F603",{"id":135,"question_zh":136,"answer_zh":137,"source_url":138},36753,"如何在导出 Markdown 时标记页码或分页符？","目前可以通过在导出时使用 page_break_placeholder 参数，然后通过脚本替换占位符来实现页码标记。示例代码如下：\n\nmarkdown_output = doc.export_to_markdown(page_break_placeholder=\"\u003C-- Page Break -->\")\n\nmarkdown_with_pages = markdown_output\npage_index = 1\nwhile \"\u003C-- Page Break -->\" in markdown_with_pages:\n    markdown_with_pages = markdown_with_pages.replace(\n        \"\u003C-- Page Break -->\", \n        f\"\\n\\n--- Page {page_index} ---\\n\\n\", \n        1\n    )\n    page_index += 1\n# 处理最后一页\nmarkdown_with_pages += f\"\\n\\n--- Page {page_index} ---\\n\\n\"\n\n未来版本可能会官方集成此功能，目前建议使用上述变通方法。","https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F309",{"id":140,"question_zh":141,"answer_zh":142,"source_url":143},36754,"Docling 对 NumPy 1.x 和 2.x 的支持情况如何？","项目目前的策略是：对于 Python 3.9 到 3.12 版本，暂时锁定使用 NumPy \u003C 2.0；对于 Python 3.13 及以上版本，支持 NumPy >= 2.0。如果在 Python 3.9-3.12 环境下安装导致 NumPy 2.x 被降级，这是预期行为，因为某些依赖项在旧版 Python 上尚未完全兼容 NumPy 2.x。建议暂时使用 NumPy \u003C 2.3.0 以确保稳定性，直到后续版本解决兼容性问题。","https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F488",{"id":145,"question_zh":146,"answer_zh":147,"source_url":148},36755,"为什么设置了 do_ocr=False 仍然提示需要安装 VLM 模型？","即使禁用了 OCR (do_ocr=False)，Docling 的某些管道组件可能仍默认尝试加载视觉语言模型 (VLM) 用于其他功能（如图片描述）。要彻底避免模型下载报错，必须像禁用 OCR 一样，预先下载所有模型并通过 artifacts_path 参数指定本地路径，或者在初始化 PipelineOptions 时明确关闭所有需要模型的选项（如 generate_parsed_pages=False 等），并确保后端配置正确。","https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2515",{"id":150,"question_zh":151,"answer_zh":152,"source_url":128},36756,"遇到模型下载超时错误（Operation timed out）如何解决？","模型下载超时通常是由于网络连接问题导致的。推荐的解决方法是：\n1. 手动下载模型文件：在有稳定网络环境的机器上使用 `docling-tools models download --all -o .\u002Fmodels` 命令下载。\n2. 将下载好的模型文件夹复制到目标机器。\n3. 在代码中通过 `artifacts_path` 参数指向该本地文件夹，从而跳过在线下载步骤。\n这样可以避免因网络波动导致的构建或运行失败。",[154,159,164,169,174,179,184,189,194,199,204,209,214,219,224,229,234,239,244,249],{"id":155,"version":156,"summary_zh":157,"released_at":158},297153,"v2.89.0","### 功能\n\n* LaTeX 后端中显式处理 TikZ 环境 ([#3187](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3187)) ([`a15c16e`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fa15c16e19fc9531e68916d15a1976ba76414c545))\n\n### 修复\n\n* **ocr:** 将 RapidOCR 英文资源与 3.8 版本的移动端模型对齐 ([#3291](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3291)) ([`251c8b2`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F251c8b217a72453205242993e03ca8004cb2877e))\n* **docx:** 在表格单元格中隔离列表状态 ([#3294](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3294)) ([`740c386`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F740c386730db1e846abc83c16e8519cd776e3ca6))\n* **pipeline:** 防止在图表提取过程中因管道选项被修改而导致缓存未命中 ([#3300](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3300)) ([`5b84911`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F5b84911a4cbed35e75e80134188be3ff4f962df0))\n\n### 文档\n\n* 在序列化笔记本中添加带索引的图片占位符示例 ([#3293](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3293)) ([`cd2e5b6`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fcd2e5b633d41733053bd5490f9de496c0b2d5d15))\n\n### 性能\n\n* **markdown:** 避免在 Markdown 后端调试日志中过早进行字符串格式化 ([#3301](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3301)) ([`a64c378`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fa64c3784d049d1b36013f26e31c302fafc8dd239))","2026-04-16T08:08:38",{"id":160,"version":161,"summary_zh":162,"released_at":163},297154,"v2.88.0","### 功能\n\n* **服务:** 为 Docling 服务建立客户端 SDK ([#3264](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3264)) ([`42157a3`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F42157a3e100ae306f74938310018be3909cabf8c))\n\n### 修复\n\n* **OCR:** 支持 RapidOCR 3.8 移动端模型命名 ([#3277](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3277)) ([`6b257ec`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F6b257ece330db9c39b8834b2b5a87b9c1eecb1fa))\n\n### 文档\n\n* 为编码助手添加代理技能包（SKILL.md、流水线、转换\u002F评估）([#3174](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3174)) ([`c23622f`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fc23622f6f53f3e286009a92545a2208cb62d148c))","2026-04-13T14:05:29",{"id":165,"version":166,"summary_zh":167,"released_at":168},297155,"v2.87.0","### 功能\n\n* **vlm:** 添加 Nanonets OCR2 入门指南 ([#3274](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3274)) ([`9970d1e`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F9970d1ef94c5e826080834d0f8858cfd8f9e7edb))\n\n### 修复\n\n* Transformers v5 与 AUTOMODEL_CAUSALLM VLM 的兼容性 ([#3276](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3276)) ([`d431224`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fd43122447f9b5b9dcad1f88819b8cb2a59f62b33))\n* **vlm:** 为 OCR 预设添加显式 MLX 支持 ([#3272](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3272)) ([`27d3cf4`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F27d3cf490ffcd3cb3c48fde8644844618b8a9d2f))\n* **markdown:** 规范化重复的首行破折号标记 ([#3286](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3286)) ([`a6aeddf`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fa6aeddf9e2fac7e3e3cfc73e558d1acf8299df61))\n* **docx:** 保留内联 SDT 引用 ([#3280](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3280)) ([`6cb1bc0`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F6cb1bc0c0297d11a2fabd7115880acd3fcea46e0))\n* **pptx:** 转换时尊重 page_range 参数 ([#3282](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3282)) ([`e4fd937`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fe4fd93742e5c5f473354bdb5f8853d3da438e9a7))\n* **vlm:** 支持工具调用 API 响应 ([#3271](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3271)) ([`9c3ab93`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F9c3ab934d6d0abad6bbdda4474d18eb73f1dd661))\n* **pdf:** 扩展连字映射，加入荷兰语 IJ 和 PUA 字形 U+F0A0 ([#3254](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3254)) ([`ab5254d`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fab5254df7c313ca258bdaa34f4bf64e0007b409f))\n\n### 文档\n\n* 添加 AG2 多智能体文档分析示例 ([#3261](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3261)) ([`1fed840`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F1fed840506ed3e6b1b0e29a3f9810b7b32d2268b))","2026-04-13T07:37:15",{"id":170,"version":171,"summary_zh":172,"released_at":173},297156,"v2.86.0","### 功能\n\n* 支持 GraniteVision v4 ([#3217](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3217)) ([`fd83420`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Ffd834204fadcb15190f3f2c289841143773b5f9d))\n* 为 DC 文档添加签名\u002F印章 HTML 块 ([#3251](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3251)) ([`9b4b67b`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F9b4b67b23e77d6d9063ee141196707412bde1673))\n* **vlm:** 为 VLM 管道页面添加 PARTIAL_SUCCESS 状态 ([#3215](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3215)) ([`6699642`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F6699642fa081a9cb50869c4d1206f9d7c89b782d))\n\n### 修复\n\n* **latex:** 忽略已过滤的间距命令的参数 ([#3245](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3245)) ([`6180925`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F61809252eccc09d6af05530fd8921a4cdb46edcc))\n\n### 文档\n\n* 在 README 中添加图表理解说明 ([#3253](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3253)) ([`d5af473`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fd5af473a70345e7f5ba405e8570ef350c65700ea))","2026-04-10T14:15:04",{"id":175,"version":176,"summary_zh":177,"released_at":178},297157,"v2.85.0","### 功能\n\n* 增加对 Falcon-OCR 的支持 ([#3237](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3237)) ([`d0e19be`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fd0e19be14ff3dbe8d44b5bf8bfe4cf53b58249f6))\n* 增加对 LightOnOCR-2-1B 的支持 ([#3213](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3213)) ([`f2affd7`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Ff2affd76149aa7c1ed84df1e84ef537f3905559b))\n\n### 修复\n\n* **latex:** 展开自定义宏参数 ([#3223](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3223)) ([`77a2505`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F77a2505bc2da4b8eede604071978cebf33addaa5))","2026-04-07T14:22:03",{"id":180,"version":181,"summary_zh":182,"released_at":183},297158,"v2.84.0","### 功能\n\n* Glm ocr ([#3146](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3146)) ([`a9265d8`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fa9265d854a195993d2e63bfc8c4bb2f76be7f9d9))\n* 切换到最新版本的 DocumentFigureClassifier 模型 v2.5 ([#3171](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3171)) ([`d046390`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fd046390bf4bff2c538cb33eebb03dce56d122d37))\n* 移除提取功能的弃用标记 ([#3220](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3220)) ([`e9a39e8`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fe9a39e872048f31b57402926ae3a40c05b7d24d0))\n","2026-04-01T18:35:27",{"id":185,"version":186,"summary_zh":187,"released_at":188},297159,"v2.83.0","### 功能\n\n* 升级至 transformers v5 ([#3200](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3200)) ([`d2c6357`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fd2c6357982d79629440919188d73bda18bc678c8))\n* 用于远程 KServe v2 API 的 OCR 模型 ([#3189](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3189)) ([`8522b00`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F8522b00146a2217760ad1944934926ed0e9f5d39))\n\n### 修复\n\n* **pdf:** 将超链接传播到 DoclingDocument 文本项中 ([#3131](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3131)) ([`524edcc`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F524edcce73869a87b6ccf73bc16324742bd36648))\n* **xlsx:** 在 Excel 表格扫描中保护最后一行的边界 ([#3197](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3197)) ([`85ac377`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F85ac3775148494e2767bbe17ce8d7a28a8baf6b6))\n* 解析多列\u002F多行表格单元格中的 LaTeX 宏 ([#3204](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3204)) ([`89c68f8`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F89c68f8ec373c6012c963a39ea70f5c122e0e779))\n* 处理空 CSV 文件而不崩溃 ([#3196](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3196)) ([`f283484`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Ff2834848aeaa63ac51f4968e1665b6b8e77b90e4))\n\n### 文档\n\n* 添加基于行的分块器文档和示例 ([#3210](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3210)) ([`3a64f41`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F3a64f41af86c90af71d6befe619f9f5a12a26e5f))","2026-03-31T09:32:59",{"id":190,"version":191,"summary_zh":192,"released_at":193},297160,"v2.82.0","### 功能\n\n* 实现基于无头浏览器的 HTML 后端（[#2969](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2969)）（[`1c74a9b`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F1c74a9b9c7c2019b85abef8f0f94381a83b721df)）\n\n### 修复\n\n* **omml:** 修正分数、数学运算符和函数的 LaTeX 输出（[#3122](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3122)）（[`e36125b`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fe36125ba2ddfbe584fc752e6dc7ca0f0f8f58d87)）\n* 管理 PDFium 后端资源的生命周期，以避免 SIGSEGV\u002FSIGTRAP 崩溃（[#3180](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3180)）（[`a0fc3c9`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fa0fc3c9d731c29f896680b17fa6df5549e2dfc5d)）\n* **docx:** 将多个 OMML 公式拆分为单独的公式项（[#3123](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3123)）（[`90d6dd4`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F90d6dd4e87d96167aced588249dcb2e0f47cd68f)）\n* 允许用户参数在 API VLM 引擎中覆盖引擎默认值（[#3116](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3116)）（[`fdf5e20`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Ffdf5e20ccd8ae85ea73effa6c743910ed295564d)）\n* **vlm:** 在 API 响应中处理 content_filter 的结束原因（[#3051](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3051)）（[`f0e3d1d`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Ff0e3d1df2a086710d5c9629426595f5d54ed65aa)）\n* **cli:** 避免为非图像导出生成图片（[#3127](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3127)）（[`5473e07`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F5473e074505e0bd46985683800fa8f929fd53492)）\n* 尊重图片描述的批处理和缩放选项（[#3132](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3132)）（[`9abf0fd`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F9abf0fd3851429183debfb90e2a9f975c9654beb)）\n\n### 文档\n\n* 修复导致空响应或错误响应的 vLLM VLM 流水线引擎选项参数错误问题（[#3167](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3167)）（[`fffd445`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Ffffd4457892002f5668e3a37b3c7a79e36936405)）","2026-03-25T09:40:08",{"id":195,"version":196,"summary_zh":197,"released_at":198},297161,"v2.81.0","### 功能\n\n* 将纯文本文件和 Quarto\u002FR Markdown 文件路由到 Markdown 后端（[#3161](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3161)）（[`96d7c7e`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F96d7c7ec79992d8dddedfafaaedb7f9bf6e14f40)）\n\n### 修复\n\n* **docx:** 编号标题后缺少列表项 (#2665)（[#2678](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2678)）（[`2f7c09e`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F2f7c09e0d8f07a5fa0aaf4f33bdfb1f71d3f3063)）\n* 避免 pypdfium 后端线程不安全的关闭操作（[#3160](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3160)）（[`afb4bb6`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fafb4bb68023c5d8fb8dc5e39413a27678e642293)）\n* 处理 MsWordDocumentBackend 中的外部图片关系（[#3114](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3114)）（[`8ae0974`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F8ae0974a9d86a447f78e4950bc0a45d5eba31e98)）\n* 在 Windows CLI 上处理目录输入时的 PermissionError 异常（[#3149](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3149)）（[`a39317a`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fa39317a147859c68bf8aef635276a23585725529)）\n* 避免管道选项的原地修改破坏缓存键（[#3115](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3115)）（[`412af62`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F412af62135869978b7d22e1dd4ee2725623fad44)）\n* 在 get_engine_config 中保留 torch_dtype，并将其添加到 CodeFormulaV2 中（[#3117](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3117)）（[`53a5f80`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F53a5f80a43849d853d4e0598d3875e6aac2f88e0)）\n* 在提取帧后释放图像后端资源（[#3134](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3134)）（[`1e841eb`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F1e841ebcbd048fbfc11d63b4086539b7cd88bb77)）","2026-03-20T21:33:00",{"id":200,"version":201,"summary_zh":202,"released_at":203},297162,"v2.80.0","### 功能\n\n* 添加 VllmCudaGraphMode ([#3125](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3125)) ([`f950679`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Ff950679f60ab6b1a9b057e7131fc8c8334e6e62e))\n","2026-03-14T05:57:49",{"id":205,"version":206,"summary_zh":207,"released_at":208},297163,"v2.79.0","### Feature\n\n* Add fact metadata and linkbase relationships for XBRL ([#3084](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3084)) ([`7952efe`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F7952efee2fcbae2a9c516d75acd8995c004fc949))\n\n### Fix\n\n* Use OCR cells with TableFormer v2 ([#3107](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3107)) ([`93f6fee`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F93f6feeabcef81b1f71a189458b0166af9db176c))\n* Add self-consistency check in the table-structure model ([#3105](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3105)) ([`2a0e11f`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F2a0e11f762fc06e16597c5d3662bc47a500efefa))\n* Correct typos in log messages and add missing error log ([#3097](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3097)) ([`198d0af`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F198d0af19b20424e118301d47d155e4b021e50a7))\n* Don't force cast to float32 in API Kserve v2 inputs ([#3101](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3101)) ([`fef01f8`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Ffef01f8c88ed827e6443f4f6fc25fa94571dcd41))\n","2026-03-12T07:40:53",{"id":210,"version":211,"summary_zh":212,"released_at":213},297164,"v2.78.0","### Feature\n\n* Add support for TableFormer v2 ([#3013](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3013)) ([`4ccd1d4`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F4ccd1d465deb8d521c09e2da61b537a9236d6560))\n* Add gRPC transport for KServe v2 API engine ([#3074](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3074)) ([`3d90778`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F3d90778e3e5762b16758e1c121f42890e32f0560))\n\n### Fix\n\n* **html:** Fix broken document tree and quadratic complexity in rich table cells ([#3025](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3025)) ([`80f75b8`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F80f75b8896a6b15c5422c56e9a423e4d2e6673cd))\n* Loosen dependency for pandas3 ([#3095](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3095)) ([`5188180`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F5188180ea31dd90567140affc564ce2729b6e4a1))\n* Add parse timeout to legacy LaTeX documents ([#3019](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3019)) ([`1192714`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F1192714b536ebb8117785b06ed85e7d203e0996d))\n* **msword:** Skip GroupItem targets without comments attribute ([#3080](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3080)) ([`ee16285`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fee16285651e5c2f963e051b1ee32b50a043191e2))\n\n### Documentation\n\n* Fix code in rag langchain chunker tokenizer ([#2993](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2993)) ([`d113e61`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fd113e611c445db6793fd94b3fee9c4109513d04a))\n* Update code snippet to use modern pipeline options syntax ([#3087](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3087)) ([`95b759e`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F95b759e5199f1142fb66dc2088c0c36177c5c284))\n* Set HuggingFaceEndpoint task for Mixtral examples ([#2945](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2945)) ([`5d3ac38`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F5d3ac38a65000cd39766f87557c685668224ad7f))\n","2026-03-10T14:55:09",{"id":215,"version":216,"summary_zh":217,"released_at":218},297165,"v2.77.0","### Feature\n\n* Track vlm_inference time for mlx_model pipeline ([#3060](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3060)) ([`38c4bb2`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F38c4bb26e8e3a7797d1caec3f690a7c8d5d9a735))\n* Add configurable graph_optimization_level for ONNX Runtime engines ([#3071](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3071)) ([`cfc6636`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fcfc6636a2a0e6b149dd51714d20e9b93f3f6463b))\n\n### Fix\n\n* **docx:** Preserve URL fragments and query params in hyperlinks ([#3050](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3050)) ([`cd9dd10`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fcd9dd10ccfe2a112af10ad135f8293d3bf845e1a))\n* Detect Office Open XML formats from ZIP contents when filename has no extension ([#3073](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3073)) ([`56f06fe`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F56f06fe372e3bfda29c14d66de0a066afb4c79c0))\n* **readingorder:** Assign FURNITURE content_layer to footer\u002Fheader in container groups ([#3044](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3044)) ([`f7cb304`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Ff7cb304daa7b7bfe49ba23b81d53fb16da4024af))\n* **docx:** Handle list items immediately after numbered headings ([#3070](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3070)) ([`56eb127`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F56eb12782c804b7ec36145bf52c1e005839c816b))\n* **rapidocr:** ORT thread configuration for RapidOCR backend ([#3062](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3062)) ([`68336c2`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F68336c2bda2b79f10759ad1587626c47500f4fb4))\n\n### Documentation\n\n* Add examples and fix docstring bug in DocumentConverter ([#3064](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3064)) ([`653940e`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F653940e0251e1bc5f311aded31690c64f42d9819))\n* Add docstrings to PipelineOptions classes ([#3065](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3065)) ([`8b99085`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F8b990856cd48fec12c68d940e665d8187d349753))\n","2026-03-06T13:45:29",{"id":220,"version":221,"summary_zh":222,"released_at":223},297166,"v2.76.0","### Feature\n\n* Export to WebVTT format ([#3036](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3036)) ([`d276e60`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fd276e6056106b6aa04fee65def96d3e10557d632))\n\n### Fix\n\n* **xlsx:** Handle OneCellAnchor images in Excel backend ([#3045](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3045)) ([`859c302`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F859c302310289c5bab45a6e160e7cc3b9c538343))\n* Normalize Unicode ligatures in PDF text extraction ([#3057](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3057)) ([`6198e69`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F6198e69dec33d9c14b3be279b19924d73e5eb3fb))\n* **ocr:** Update RapidOCR torch GPU config key ([#3049](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3049)) ([`477359b`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F477359b772039b9c9c0d31c9dabcd755abdeb560))\n* Convert PIL images to RGB before picture description ([#3014](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3014)) ([`90ce93d`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F90ce93d8a095ea17040bd6a91ded0b463998bea9))\n* **msword:** Use outlineLvl for heading levels and clamp to minimum 1 ([#2916](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2916)) ([`a3d2b4b`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fa3d2b4bcc07fc00fff3039ae2046ee69b7587ab2))\n\n### Documentation\n\n* Add metaxy integration ([#3058](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3058)) ([`7aacc6c`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F7aacc6c18da3e856babb0f06afd7c985774f118e))\n* Removes merge conflict artifacts ([#3055](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3055)) ([`672125c`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F672125cd1bb5e22bb7a677f48157a55ca93f9ff6))\n* Add audio & video processing guide ([#3038](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3038)) ([`1321b39`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F1321b39cd8203d5e1cd60191cc9e979c5b939f98))\n* Add XBRL conversion example notebook and update feature listings ([#3039](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3039)) ([`1eb5c21`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F1eb5c21dabfed02bfe71cb7fc502d124562f1ba8))\n","2026-03-02T14:43:14",{"id":225,"version":226,"summary_zh":227,"released_at":228},297167,"v2.75.0","### Feature\n\n* Create a backend parser for XBRL instance reports ([#3017](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3017)) ([`334ba6e`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F334ba6e51fa7feb5f2ae15fce4612c7b3fad67d6))\n* Unified model-family inference engines (including image-classification) and KServe v2 API support ([#2979](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2979)) ([`0353293`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F03532938b52fb1513e2ea3afffc6da6a7ded7cc7))\n\n### Fix\n\n* Skip ASR segments when length is zero ([#2998](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2998)) ([`6b824f8`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F6b824f844b698eb015c28b69addfbaca169ec8d4))\n* **docx:** Guard against None hyperlink address in _get_paragraph_elements (#2367) ([#3022](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3022)) ([`236216e`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F236216ed4e2b7c4b627a81b6b77dd8bac01428a5))\n","2026-02-24T20:16:57",{"id":230,"version":231,"summary_zh":232,"released_at":233},297168,"v2.74.0","### Feature\n\n* Introduce docling-parse v5 and deprecate old docling-parse backends ([#2872](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2872)) ([`bf417e6`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fbf417e6d264ebaf93bda7f53534e2cc50ccb2284))\n\n### Fix\n\n* Security vulnerabilities with XML External Entity and related attacks ([#3009](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3009)) ([`576bada`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F576bada7b7d542ea308778a053bc3c4d49086f20))\n* **csv:** Set default delimiter by default ([#3005](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3005)) ([`a1b0e3f`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fa1b0e3fd6bde26466399ea477ae5624d72d24781))\n* Improved deserialization of engine_options ([#3008](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F3008)) ([`dbba6ea`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fdbba6ea27fd92e2dfcf79e136a96cea5784edf8a))\n","2026-02-17T21:16:44",{"id":235,"version":236,"summary_zh":237,"released_at":238},297169,"v2.73.1","### Fix\n\n* **asciidoc:** Handle commas in image alt text ([#2983](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2983)) ([`86b6912`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F86b691204d2e4c2a54c99d80063e2dd5b5428168))\n* Use timezone-aware datetime ([#2947](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2947)) ([`e2870f9`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fe2870f94ed78caeb6db9d735b5a73fa80e5e2104))\n* Add failed pages to DoclingDocument for page break consistency ([#2939](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2939)) ([`1f91482`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F1f914826bb07c32766e7db37f86baec3ea772a11))\n","2026-02-13T15:34:54",{"id":240,"version":241,"summary_zh":242,"released_at":243},297170,"v2.73.0","### Feature\n\n* Inference engines abstraction for object detection model family with HF Transformers and ONNX runtime ([#2959](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2959)) ([`14e474c`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F14e474c95555f04e5c4ac55351ad802d372858fc))\n* Added support for parsing LaTeX (.tex) documents ([#2890](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2890)) ([`e6ccb8b`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fe6ccb8b2c1d99fa6e2660d7c4bb866af7960bc2d))\n* Introduce pluggable VLM runtime system with preset-based configuration ([#2919](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2919)) ([`d4c8713`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fd4c87133f3f4dcfc8c7619d533bac31cc297350d))\n\n### Fix\n\n* Restore expected behavior for artifacts_path and accelerator_options in VLM engines ([#2961](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2961)) ([`9721321`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F9721321c4604da9334e84f7942b41222b580ae96))\n* Allow offline chart extraction model artifacts ([#2957](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2957)) ([`ae4fdbb`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fae4fdbbb09fd377bb271e9b2efe541873eeb2990))\n\n### Documentation\n\n* Add LaTeX and WebVTT as supported types ([#2974](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2974)) ([`704ef0a`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F704ef0afbaca782d35454b66b26e3cb931c79653))\n","2026-02-11T09:54:00",{"id":245,"version":246,"summary_zh":247,"released_at":248},297171,"v2.72.0","### Feature\n\n* Add chart extraction models ([#2848](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2848)) ([`fe45c71`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Ffe45c71fe7ad137088e3719dc99e337860120d33))\n\n### Fix\n\n* **backend:** Improve Excel table bounds detection and flatten merged cells ([#2778](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2778)) ([`3110c43`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F3110c439da48fe215379492a29a310e64e9d67e7))\n* **pptx:** Handle picture shapes with external image references ([#2914](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2914)) ([`5e452a2`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F5e452a2e8fcb5ea43a8a7666320998604279c152))\n\n### Documentation\n\n* Add granite vision for charts ([#2946](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2946)) ([`a5ad8f2`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fa5ad8f24ff0fe9bcb0b915a26b4132b6dbb65f93))\n","2026-02-03T15:08:34",{"id":250,"version":251,"summary_zh":252,"released_at":253},297172,"v2.71.0","### Feature\n\n* Webvtt and source tracker ([#2787](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2787)) ([`0602a7c`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F0602a7cdab17b0e42057e1ef502048e95bd589f4))\n* Add support for Word document comments extraction ([#2834](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2834)) ([`b6ca094`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002Fb6ca09451963c606b5d280b74e559278717bb911))\n\n### Fix\n\n* Allow newer typer versions ([#2930](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2930)) ([`6f205ae`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F6f205ae2119fe694abaf200df5662837b3854f53))\n* **rapidocr:** Use new model links for RapidOCR ([#2928](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2928)) ([`82b7982`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F82b7982e1b23f46fb664fb3229f3eac054957077))\n* Presets for ollama ([#2926](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fissues\u002F2926)) ([`4a269de`](https:\u002F\u002Fgithub.com\u002Fdocling-project\u002Fdocling\u002Fcommit\u002F4a269de91aeadd1a1c48b814dc4e5a2c28efe6d8))\n","2026-01-30T17:11:21"]