[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-OpenLAIR--dr-claw":3,"tool-OpenLAIR--dr-claw":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},3128,"ragflow","infiniflow\u002Fragflow","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。",77062,"2026-04-04T04:44:48",[15,14,13,26,54],{"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":79,"owner_website":79,"owner_url":80,"languages":81,"stars":116,"forks":117,"last_commit_at":118,"license":119,"difficulty_score":10,"env_os":120,"env_gpu":121,"env_ram":121,"env_deps":122,"category_tags":128,"github_topics":79,"view_count":23,"oss_zip_url":79,"oss_zip_packed_at":79,"status":16,"created_at":129,"updated_at":130,"faqs":131,"releases":160},1240,"OpenLAIR\u002Fdr-claw","dr-claw","A Super AI Lab with massive AI Doctors as Assistants. Best IDE for Research via AI Power. ","Dr. Claw 是一款面向研究人员和开发者的 AI 研究助手，旨在帮助用户高效完成从创意构思到实验执行、论文撰写和成果发布的全流程研究工作。它提供了一个集成化的研究平台，让用户无需在多个工具之间切换，即可完成整个研究过程。\n\nDr. Claw 解决了传统研究过程中流程繁琐、效率低下以及协作困难的问题。通过自动化任务执行、智能生成研究计划和多代理协作等功能，它显著提升了研究工作的效率与质量。\n\n这款工具特别适合研究人员、学术开发者以及需要频繁进行数据分析和论文写作的用户使用。无论是进行初步的创意生成，还是复杂的实验设计与结果分析，Dr. Claw 都能提供强有力的支持。\n\n其独特的亮点包括结构化的研究实验室界面、一键启动自动研究流程、100 多种研究技能库，以及基于聊天驱动的任务规划功能。此外，它还支持多种 AI 引擎的无缝切换，满足不同场景下的需求。","\u003Cdiv align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FOpenLAIR_dr-claw_readme_5147600a5861.png\" alt=\"Dr. Claw\" width=\"128\" height=\"128\">\n  \u003Ch1>Dr. Claw: Your AI Research Assistant\u003C\u002Fh1>\n  \u003Cp>\u003Cstrong>Full-stack research workspace.\u003C\u002Fstrong>\u003C\u002Fp>\n\u003C\u002Fdiv>\n\n\u003Cp align=\"center\">\n\u003Ca href=\"https:\u002F\u002Fopenlair.github.io\u002Fdr-claw\">\n\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F%F0%9F%8C%90-Homepage-CB2B3E?style=for-the-badge\" alt=\"Homepage\" \u002F>\n\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenLAIR\u002Fdr-claw\">\n\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F%F0%9F%A6%9E-Dr.%20Claw-CB2B3E?style=for-the-badge\" alt=\"Dr. Claw\" \u002F>\n\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenLAIR\u002Fdr-claw\u002Fblob\u002Fmain\u002FLICENSE\">\n\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-GPL--3.0%20%2B%20AGPL--3.0-blue?style=for-the-badge\" alt=\"License: GPL-3.0 + AGPL-3.0\" \u002F>\n\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fjoin.slack.com\u002Ft\u002Fvibe-lab-group\u002Fshared_invite\u002Fzt-3r4bkcx5t-iGyRMI~r09zt7p_ND2eP9A\">\n\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FJoin-Slack-4A154B?style=for-the-badge&logo=slack\" alt=\"Join Slack\" \u002F>\n\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fx.com\u002FVibe2038004\">\n\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FFollow-on%20X-black?style=for-the-badge&logo=x\" alt=\"Follow on X\" \u002F>\n\u003C\u002Fa>\n\u003Ca href=\".\u002Fpublic\u002Fwechat-group-qr.jpg\">\n\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FJoin-WeChat-07C160?style=for-the-badge&logo=wechat&logoColor=white\" alt=\"Join WeChat\" \u002F>\n\u003C\u002Fa>\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  \u003Ca href=\".\u002FREADME.md\">English\u003C\u002Fa> | \u003Ca href=\".\u002FREADME.zh-CN.md\">中文\u003C\u002Fa>\n\u003C\u002Fp>\n\n## Table of Contents\n\n- [Overview](#overview)\n- [Highlights](#highlights)\n- [Quick Start](#quick-start)\n- [Configuration](#configuration)\n- [OpenRouter](#openrouter)\n- [OpenClaw Integration](#openclaw-integration)\n- [Research Lab - Quick Example](#research-lab-quick-example)\n- [Usage Guide](#usage-guide)\n- [Additional Details](#additional-details)\n- [Contributing](#contributing)\n- [FAQ](.\u002Fdocs\u002Ffaq.md)\n- [License](#license)\n- [Citation](#citation)\n- [Acknowledgments](#acknowledgments)\n- [Support & Community](#support--community)\n\n## Overview\n\nDr. Claw is a general-purpose AI research assistant designed to help researchers and builders execute end-to-end projects across different domains. From shaping an initial idea to running experiments and preparing publication-ready outputs, Dr. Claw keeps the full workflow in one place so teams can focus on research quality and iteration speed.\n\n## Product Screenshot\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FOpenLAIR_dr-claw_readme_b620dae813d2.png\" alt=\"Dr. Claw chat interface\" width=\"1000\">\n\u003C\u002Fp>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>The Philosophy: Leveraged Cognition\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FOpenLAIR_dr-claw_readme_bb46b412f1f1.png\" alt=\"Leveraged Cognition\" width=\"900\">\n\u003C\u002Fp>\n\n**Manual work is too slow. Fully automated AI is too generic. Vibe Researching is the new frontier.** Dr. Claw turns your **Research Taste** into outsized outcomes with **Agentic Execution**--so you can move faster, think bigger, and still hold the line on scientific rigor.\n\n\u003C\u002Fdetails>\n\n## Highlights\n\n- **🔬 Research Lab** — Structured dashboard for end-to-end research: define your brief, generate a pipeline of tasks, track progress across Survey → Ideation → Experiment → Publication → Promotion, and inspect source papers, ideas (rendered with LaTeX math), and cache artifacts — all at a glance\n- **⚡ Auto Research** — Start one-click sequential task execution directly from the Project Dashboard, open the generated session live, and receive an email when the run completes\n- **📚 100+ Research Skills** — A curated library spanning idea generation, code survey, experiment development & analysis, paper writing, review response, and delivery — automatically discovered by agents and applied as task-level assistance\n- **🗂️ Chat-Driven Pipeline** — Describe your research idea in Chat; the agent uses the `inno-pipeline-planner` skill to interactively generate a structured research brief and task list — no manual templates needed\n- **🤖 Multi-Agent Backend** — Seamlessly switch between Claude Code, Gemini CLI, Codex, and OpenRouter as your execution engines\n\n### What the Pipeline Produces\n\n| | Artifact | Location | Description |\n|---|---|---|---|\n| 📚 | Survey reports | `Survey\u002Freports\u002F` | Literature reviews with citations from arXiv, Semantic Scholar, and web sources |\n| 💡 | Research ideas | `Ideation\u002Fideas\u002F` | Brainstorming outputs with multi-persona evaluation scores |\n| 🔬 | Experiment code | `Experiment\u002Fcore_code\u002F` | Implementation from the plan → implement → judge loop |\n| 📊 | Analysis results | `Experiment\u002Fanalysis\u002F` | Statistical analysis, tables, and paper-ready figures |\n| 📝 | Paper draft | `Publication\u002Fpaper\u002F` | Academic manuscript (IEEE\u002FACM format) with citations and LaTeX math |\n| 🎞️ | Presentation | `Promotion\u002Fslides\u002F` | Slide deck, TTS narration audio, and demo video |\n\n> See [docs\u002Fpipeline-outputs.md](docs\u002Fpipeline-outputs.md) for the full artifact list and project directory structure.\n\n\u003Cdetails>\n\u003Csummary>\u003Cspan style=\"font-size: 1.17em; font-weight: 600;\">More Features\u003C\u002Fspan>\u003C\u002Fsummary>\n\n- **💬 Interactive Chat + Shell** — Chat with your agent or drop into a full terminal — side by side with your research context\n- **📁 File & Git Explorer** — Browse files with syntax highlighting, live-edit, stage changes, commit, and switch branches without leaving the UI\n- **📱 Responsive & PWA-Ready** — Desktop, tablet, and mobile layouts with bottom tab bar, swipe gestures, and Add-to-Home-Screen support\n- **🔄 Session Management** — Resume conversations, manage multiple sessions, and track full history across projects\n\n\u003C\u002Fdetails>\n\n### Feature Gallery\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>Expand screenshots\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n\u003Cp>\u003Cstrong>Project Dashboard\u003C\u002Fstrong> — Start from the project overview, review status, and launch end-to-end automation.\u003C\u002Fp>\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FOpenLAIR_dr-claw_readme_a6f713435c62.png\" alt=\"Project dashboard interface\" width=\"1000\">\n\u003C\u002Fp>\n\n\u003Cp>\u003Cstrong>Skill Library\u003C\u002Fstrong> — Browse reusable research skills across ideation, experimentation, and writing.\u003C\u002Fp>\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FOpenLAIR_dr-claw_readme_66752d79fcbb.png\" alt=\"Skill library interface\" width=\"1000\">\n\u003C\u002Fp>\n\n\u003Cp>\u003Cstrong>News Dashboard\u003C\u002Fstrong> — Follow research-relevant updates without leaving the workspace.\u003C\u002Fp>\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FOpenLAIR_dr-claw_readme_3c254c2a7a78.png\" alt=\"News dashboard interface\" width=\"1000\">\n\u003C\u002Fp>\n\n\u003C\u002Fdetails>\n\n\n## Quick Start\n\n### Prerequisites\n\n- [Node.js](https:\u002F\u002Fnodejs.org\u002F) v20 or higher (**v22 LTS recommended**, see `.nvmrc`)\n- At least one of the following CLI tools installed and configured:\n  - [Claude Code CLI](https:\u002F\u002Fdocs.anthropic.com\u002Fen\u002Fdocs\u002Fclaude-code)\n  - [Gemini CLI](https:\u002F\u002Fgeminicli.com\u002Fdocs\u002Fget-started\u002Finstallation\u002F)\n  - [Codex CLI](https:\u002F\u002Fdevelopers.openai.com\u002Fcodex\u002Fcli\u002F)\n- Some systems need native build tools for dependencies like `node-pty` and `better-sqlite3`. If `npm install` fails, see [FAQ](docs\u002Ffaq.md).\n\nCursor agent support is in progress and coming soon.\n\n### Installation \n\n1. **Clone the repository:**\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FOpenLAIR\u002Fdr-claw.git\ncd dr-claw\n```\n\n2. **Install dependencies:**\n```bash\nnpm install\n```\n\n3. **Configure environment:**\n```bash\ncp .env.example .env\n# Edit .env with your preferred settings (port, etc.)\n```\n\nNeed custom ports, auth, or workspace settings? See [docs\u002Fconfiguration.md](docs\u002Fconfiguration.md).\n\n4. **Start the application:**\n\n```bash\n# Development mode (with hot reload)\nnpm run dev\n```\n\nThen create your account via the bowser `http:\u002F\u002Flocalhost:5173`.\n\n5. **Use the application**\n\nThere are two ways to interact with Dr. Claw: the **frontend UI** workflow or the **terminal-only**. The UI provides richer visualization but may encounter occasional bugs; the terminal approach is more stable and lightweight.\n\n#### Option A: Frontend UI\n\nOpen your browser at `http:\u002F\u002Flocalhost:5173` (or the port you configured in `.env`).\n\n\n\n#### Option B: Terminal Only\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FOpenLAIR_dr-claw_readme_937babfcfd12.png\" alt=\"Terminal workflow example\" width=\"800\">\n\u003C\u002Fp>\n\nOpen a **second terminal** (keep `npm run dev` running in the first) and install the `drclaw` CLI harness:\n\n```bash\npip install -e .\u002Fagent-harness\n```\n\nThen log in with the credentials you created during setup:\n\n```bash\ndrclaw auth login --username YOUR_USERNAME --password YOUR_PASSWORD\n```\n\nInstall at least one agent CLI (if you haven't already):\n\n| Agent | Install | Auth |\n|-------|---------|------|\n| Claude Code | `npm install -g @anthropic-ai\u002Fclaude-code` | `claude` → follow OAuth prompt |\n| Gemini CLI | `npm install -g @google\u002Fgemini-cli` | `gemini` → Google sign-in, or `export GOOGLE_API_KEY=...` |\n| Codex CLI | `npm install -g @openai\u002Fcodex` | `codex login`, or `export OPENAI_API_KEY=...` |\n| **OpenRouter** | No CLI needed | `export OPENROUTER_API_KEY=sk-or-...` (get a key at [openrouter.ai\u002Fkeys](https:\u002F\u002Fopenrouter.ai\u002Fkeys)) |\n\n> **OpenRouter** lets you use *any* model (GPT-5, Claude, Gemini, DeepSeek, Llama, Mistral, Qwen, etc.) through a single API key. Select your model in the UI or set `OPENROUTER_MODEL` in `.env`.\n\nNavigate to the project directory you want to work in and launch any of the agents:\n\n```bash\ncd \u002Fpath\u002Fto\u002Fyour\u002Fproject\nclaude    # or: gemini | codex\n```\n\nSkills from `dr-claw\u002Fskills\u002F` are automatically symlinked into each project's `.claude\u002Fskills\u002F` directory when the project is created, so the agent discovers them without extra configuration. You can also reference any skill manually inside a session:\n\n```\n> Read .claude\u002Fskills\u002Finno-experiment-analysis\u002FSKILL.md and follow it to analyze my results.\n```\n\n\n\n#### Option C: OpenRouter Terminal Chat\n\nFor a lightweight terminal-only experience using any [OpenRouter](https:\u002F\u002Fopenrouter.ai\u002F) model, use the built-in `dr-claw chat` command. No browser or UI required — just an interactive agentic session with full tool-calling capabilities (file I\u002FO, shell, grep, glob, web search\u002Ffetch).\n\n```bash\n# Make sure OPENROUTER_API_KEY is set (or pass --key)\nexport OPENROUTER_API_KEY=sk-or-...\n\n# Launch a chat session with any model\nnode server\u002Fcli.js chat --model moonshotai\u002Fkimi-k2.5\n```\n\nYou can also pass the API key inline:\n\n```bash\nnode server\u002Fcli.js chat --model anthropic\u002Fclaude-sonnet-4 --key sk-or-your-key\n```\n\n| Flag | Description |\n|------|-------------|\n| `--model \u003Cslug>` | OpenRouter model slug (e.g., `moonshotai\u002Fkimi-k2.5`, `anthropic\u002Fclaude-sonnet-4`, `deepseek\u002Fdeepseek-r1`) |\n| `--key \u003Ckey>` | OpenRouter API key (defaults to `OPENROUTER_API_KEY` env var) |\n\nBrowse all available models at [openrouter.ai\u002Fmodels](https:\u002F\u002Fopenrouter.ai\u002Fmodels).\n\n\n\nIf agent web search does not work later, see [Troubleshooting Web Search](#troubleshooting-web-search) below.\n\n## OpenClaw Integration\n\nThis section is written for **new users integrating OpenClaw with Dr. Claw for the first time**. The goal is not to expose every detail. The goal is to get a reliable first integration running quickly:\n\n- OpenClaw can see which Dr. Claw projects exist\n- OpenClaw can find sessions waiting for user input\n- OpenClaw can reply into a chosen session so Dr. Claw keeps going\n- OpenClaw can summarize project \u002F portfolio progress and recommend what to focus on next\n\nThe clean mental model is:\n- **Dr. Claw**: owns the real projects, sessions, pipelines, artifacts, and execution\n- **`drclaw` CLI**: exposes that state through a stable local control surface exposed primarily as `drclaw`\n- **OpenClaw**: acts as the user-facing secretary on mobile, chat, or voice\n\n### Fastest path for new users\n\nIf you only want the shortest successful path, do these 5 things:\n\n1. Start Dr. Claw\n2. Install the `drclaw` CLI\n3. Give OpenClaw local shell \u002F `exec` access\n4. Install the provided OpenClaw skill\n5. Make `chat waiting` and `digest portfolio` work end-to-end\n\nOnce those two commands work, OpenClaw already behaves like a usable research secretary.\n\n### Step 0: confirm the prerequisites\n\nBefore integrating, make sure:\n- you can already run Dr. Claw locally\n- you already have at least one project, or can create one under `~\u002Fvibelab\u002F...`\n- you have configured at least one execution backend such as Claude Code, Gemini CLI, or Codex\n- your OpenClaw instance is allowed to run local tools\n\nIf those are not true yet, get Dr. Claw itself working first.\n\n### Step 1: start and verify the Dr. Claw server\n\nFrom the repo root:\n\n```bash\nnpm install\nnpm run dev\n```\n\nIn another terminal:\n\n```bash\ndrclaw --json auth status\ndrclaw server status\n```\n\nUse `auth status` as the primary reachability check. If it returns JSON, the server is reachable.\n\n`drclaw server status` only reports the daemon launched by `drclaw server on`. If you started the app manually with `npm run dev`, it may still show `STOPPED` even though `http:\u002F\u002Flocalhost:3001` is working.\n\nIf you want Dr. Claw to manage the background process for you:\n\n```bash\ndrclaw server on\n```\n\n### Step 2: install and verify the `drclaw` CLI\n\nFrom the repo root:\n\n```bash\npip install -e .\u002Fagent-harness\n```\n\nThen verify:\n\n```bash\ndrclaw --help\ndrclaw --json auth status\ndrclaw --json projects list\n```\n\nA good sequence is:\n- `drclaw --json auth status` should return JSON if the server is reachable\n- `drclaw --json projects list` will work only after login or if you already have a saved token\n\nIf `projects list` returns `Not logged in`, authenticate first:\n\n```bash\ndrclaw auth login --username \u003Cusername> --password \u003Cpassword>\n```\n\n### Step 3: give OpenClaw local CLI execution capability\n\nThe key integration is not a deep API bridge. The key is that **OpenClaw can directly execute `drclaw ...` locally**.\n\nAt minimum, OpenClaw should be able to run commands such as:\n\n```bash\ndrclaw --json chat waiting\ndrclaw --json digest portfolio\ndrclaw --json chat reply --project \u003Cproject> --session \u003Csession-id> -m \"\u003Cmessage>\"\ndrclaw --json workflow continue --project \u003Cproject> --session \u003Csession-id> -m \"\u003Cinstruction>\"\n```\n\nRecommended approach:\n- enable local `exec` \u002F shell for OpenClaw\n- prefer direct local CLI calls\n- avoid building an extra proxy layer at the beginning\n\nThe thinner this layer is, the easier it is to debug and keep reliable.\n\n### Step 4: link OpenClaw in one command\n\nRun:\n\n```bash\ndrclaw install --server-url http:\u002F\u002Flocalhost:3001\n```\n\nThis will automatically:\n- copy the Dr. Claw skill into `~\u002F.openclaw\u002Fworkspace\u002Fskills\u002Fdrclaw`\n- install the helper scripts OpenClaw uses for serialized local turns\n- save the Dr. Claw server URL for future CLI \u002F OpenClaw use\n- remember the local `drclaw` executable path\n\nIf you also want to save the default push channel during setup:\n\n```bash\ndrclaw install --server-url http:\u002F\u002Flocalhost:3001 --push-channel feishu:\u003Cchat_id>\n```\n\nThe compatibility form is still available:\n\n```bash\ndrclaw openclaw install --server-url http:\u002F\u002Flocalhost:3001\n```\n\n### Step 5: make two core commands work first\n\nFor a new user, do not start with everything. Start with these two:\n\n1. **Find which sessions are waiting for user input**\n```bash\ndrclaw --json chat waiting\n```\n\n2. **Get portfolio-wide progress and recommendations**\n```bash\ndrclaw --json digest portfolio\n```\n\nIf OpenClaw can call both commands and summarize the result back to the user, your minimum viable integration is already working.\n\n### Step 6: add the reply loop\n\nThe next common user action is: see a waiting session, then ask OpenClaw to answer it.\n\nThe fixed pattern is:\n\n1. Find waiting sessions:\n```bash\ndrclaw --json chat waiting\n```\n\n2. Let the user choose a project and session\n\n3. Send the reply:\n```bash\ndrclaw --json chat reply --project \u003Cproject> --session \u003Csession-id> -m \"\u003Cmessage>\"\n```\n\n4. Immediately re-check whether it is still waiting:\n```bash\ndrclaw --json chat waiting --project \u003Cproject>\n```\n\nIf the user wants to keep discussing the same project\u002Fsession, switch to:\n\n```bash\ndrclaw --json chat project --project \u003Cproject> --session \u003Csession-id> -m \"\u003Cinstruction>\"\n```\n\nThat is the better pattern for multi-turn project-scoped discussion.\n\n### Step 7: recommended fixed operating patterns\n\nRecommended OpenClaw flows:\n\n1. **User asks: what needs my attention right now?**\n```bash\ndrclaw --json digest portfolio\n```\n\n2. **User asks: which sessions are waiting for me?**\n```bash\ndrclaw --json chat waiting\n```\n\n3. **User asks: what is the latest state of this project?**\n```bash\ndrclaw --json projects latest \u003Cproject>\ndrclaw --json projects progress \u003Cproject>\n```\n\n4. **User says: reply to this session and keep it moving**\n```bash\ndrclaw --json chat reply --project \u003Cproject> --session \u003Csession-id> -m \"\u003Cmessage>\"\n```\n\n5. **User says: I just had a new idea, create a project and help me shape it**\n```bash\ndrclaw --json projects idea \u002Fabsolute\u002Fpath\u002Fto\u002Fproject --name \"\u003Cdisplay-name>\" --idea \"\u003Cidea text>\"\n```\n\n### Step 8: prefer serialized local turns\n\nWhen OpenClaw repeatedly runs `openclaw agent --local`, use the serialized wrapper to avoid session-lock collisions:\n\n```bash\nagent-harness\u002Fskills\u002Fdr-claw\u002Fscripts\u002Fopenclaw_drclaw_turn.sh\n```\n\nExample:\n\n```bash\nopenclaw_drclaw_turn.sh --json -m \"Use your exec tool to run `drclaw --json digest portfolio`. Return only raw stdout.\"\n```\n\nIn practice: **when OpenClaw calls Dr. Claw locally, stable serial turns are better than risky parallel turns.**\n\n### Step 9: how to know the integration is successful\n\nA new user can consider the integration complete once all 4 are true:\n- OpenClaw can list Dr. Claw projects\n- OpenClaw can identify waiting sessions\n- OpenClaw can successfully send one reply into a chosen session\n- OpenClaw can produce one `digest portfolio` style summary with recommendations\n\nAt that point, OpenClaw is no longer just a chat surface. It becomes Dr. Claw's mobile secretary.\n\n### Step 10: what end users can say afterwards\n\nAfter setup, users should be able to talk to OpenClaw naturally:\n- “Check which Dr. Claw projects are waiting for my reply.”\n- “Summarize the last message and current progress of this project.”\n- “Reply to this session: continue with option B and report back when finished.”\n- “Summarize recent experiment progress across projects and recommend what I should focus on today.”\n- “I just had a new idea. Create a Dr. Claw project, discuss it with me, refine it, and start execution planning.”\n\nThe goal is not to replace Dr. Claw. The goal is to make Dr. Claw **callable, reportable, steerable, and remotely manageable** through OpenClaw.\n\n## Configuration\n\nDr. Claw reads local settings from `.env`. For most users, the only required step is copying `.env.example` to `.env`, but these are the settings you are most likely to adjust early:\n\n- `PORT`: backend server port\n- `VITE_PORT`: frontend dev server port\n- `HOST`: bind address for the frontend and backend\n- `JWT_SECRET`: required before exposing Dr. Claw beyond localhost\n- `WORKSPACES_ROOT`: default root for new project workspaces\n\nFor the full environment reference and deployment notes, see [docs\u002Fconfiguration.md](docs\u002Fconfiguration.md).\n\nAuto Research email notifications are configured inside the app at **Settings → Email**. The v1 flow supports Claude Code, Codex, Gemini, and OpenRouter engines for unattended task execution, and interrupted runs are automatically reconciled so they do not remain stuck in `running`.\n\n## OpenRouter\n\n[OpenRouter](https:\u002F\u002Fopenrouter.ai\u002F) is integrated as a first-class provider, giving you access to **hundreds of models** (GPT-5, Claude, Gemini, DeepSeek, Llama, Mistral, Qwen, Kimi, and more) through a single API key.\n\n### Setup\n\n1. Get an API key at [openrouter.ai\u002Fkeys](https:\u002F\u002Fopenrouter.ai\u002Fkeys).\n2. Set the key in one of three ways:\n   - **Environment variable:** `export OPENROUTER_API_KEY=sk-or-...`\n   - **`.env` file:** add `OPENROUTER_API_KEY=sk-or-...` to your project `.env`\n   - **UI:** go to **Settings → OpenRouter** and paste your key\n\n### Using OpenRouter in the UI\n\n1. Open a project and go to **Chat**.\n2. Under **Choose Your AI Assistant**, click **OpenRouter**.\n3. Search for a model in the dropdown (it fetches the full list from OpenRouter) or type a custom model slug.\n4. Start chatting — the agent has the same tool-calling capabilities as Claude, Gemini, and Codex (file read\u002Fwrite, shell, grep, glob, web search\u002Ffetch, todo).\n\nOpenRouter is also available in **Auto Research** on the Project Dashboard — select it as the provider and pick any model.\n\n### Using OpenRouter in the Terminal\n\nNo browser needed. The `dr-claw chat` CLI gives you a fully agentic terminal session:\n\n```bash\n# Basic usage\nnode server\u002Fcli.js chat --model moonshotai\u002Fkimi-k2.5\n\n# With an explicit API key\nnode server\u002Fcli.js chat --model deepseek\u002Fdeepseek-r1 --key sk-or-your-key\n```\n\nThe CLI supports the same tools as the UI (file I\u002FO, shell, grep, glob, web search, web fetch, todo). Type your message and the agent will execute multi-step research tasks autonomously.\n\n### Default Model\n\nSet `OPENROUTER_MODEL` in `.env` to change the default model used when none is specified:\n\n```env\nOPENROUTER_MODEL=moonshotai\u002Fkimi-k2.5\n```\n\nIf unset, the default is `anthropic\u002Fclaude-sonnet-4`.\n\n\u003Ca id=\"research-lab-quick-example\">\u003C\u002Fa>\n\n## Research Lab — Quick Example\n\nThe core feature of Dr. Claw is the **Research Lab**.\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>Research Lab Screenshot\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FOpenLAIR_dr-claw_readme_2f54d80ef79c.png\" alt=\"Research Lab workflow\" width=\"1000\">\n\u003C\u002Fp>\n\n\u003C\u002Fdetails>\n\nThe typical flow is:\n\n1. Configure one supported agent in **Settings**.\n2. Configure notification settings in **Settings → Email** if you want completion email notifications.\n3. Describe your research idea in **Chat**.\n4. Let the agent generate `.pipeline\u002Fdocs\u002Fresearch_brief.json` and `.pipeline\u002Ftasks\u002Ftasks.json`.\n5. Review the pipeline in **Research Lab** and either send tasks back to **Chat** manually or click **Auto Research** on the Project Dashboard to run them sequentially.\n\nFor full step-by-step operations, see **Usage Guide** below.\n\n## Usage Guide\n\nAfter starting Dr. Claw, open your browser and follow the steps below.\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>Step 1 — Create or Open a Project\u003C\u002Fstrong>\u003C\u002Fsummary>\n\nWhen you first open Dr. Claw you will see the **Projects** sidebar. You have two options:\n\n- **Open an existing project** — Dr. Claw auto-discovers registered projects and linked sessions from Claude Code, Codex, and Gemini.\n- **Create a new project** — Click the **\"+\"** button, choose a directory on your machine, and Dr. Claw will set up the workspace: agent folders such as `.claude\u002F`, `.agents\u002F`, `.gemini\u002F`, standard workspace metadata, linked `skills\u002F` directories, preset research dirs (`Survey\u002Freferences`, `Survey\u002Freports`, `Ideation\u002Fideas`, `Ideation\u002Freferences`, `Experiment\u002Fcode_references`, `Experiment\u002Fdatasets`, `Experiment\u002Fcore_code`, `Experiment\u002Fanalysis`, `Publication\u002Fpaper`, `Promotion\u002Fhomepage`, `Promotion\u002Fslides`, `Promotion\u002Faudio`, `Promotion\u002Fvideo`), and **instance.json** at the project root with absolute paths for those directories. Cursor agent support is coming soon.\n\n> **Default project storage path:** New projects are stored under `~\u002Fdr-claw` by default. You can change this in **Settings → Appearance → Default Project Path**, or set the `WORKSPACES_ROOT` environment variable. The setting is persisted in `~\u002F.claude\u002Fproject-config.json`.\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>Step 2 — Generate Your Research Pipeline via Chat\u003C\u002Fstrong>\u003C\u002Fsummary>\n\nAfter creating or opening a project, Dr. Claw opens **Chat** by default. If no research pipeline exists yet, an onboarding banner appears with a **Use in Chat** button that injects a starter prompt.\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>Chat Screenshot\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FOpenLAIR_dr-claw_readme_7dae94904768.png\" alt=\"Chat interface\" width=\"1000\">\n\u003C\u002Fp>\n\n\u003C\u002Fdetails>\n\nDescribe your research idea — even a rough one is fine. The agent uses the `inno-pipeline-planner` skill to ask clarifying questions and then generates:\n- `.pipeline\u002Fdocs\u002Fresearch_brief.json` (your structured research brief)\n- `.pipeline\u002Ftasks\u002Ftasks.json` (the task pipeline)\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>Step 3 — Review in Research Lab and Execute Tasks\u003C\u002Fstrong>\u003C\u002Fsummary>\n\nSwitch to **Research Lab** to review the generated tasks, progress metrics, and artifacts. Then execute tasks:\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>Task Execution Screenshot\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FOpenLAIR_dr-claw_readme_264d8c2b05ba.png\" alt=\"Task list and execution flow\" width=\"1000\">\n\u003C\u002Fp>\n\n\u003C\u002Fdetails>\n\n1. Choose a CLI backend from the **CLI selector** (Claude Code, Gemini CLI, or Codex).\n2. In **Research Lab**, click **Go to Chat** or **Use in Chat** on a pending task.\n3. The agent executes the task and writes results back to the project.\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>Optional — Run Auto Research from the Project Dashboard\u003C\u002Fstrong>\u003C\u002Fsummary>\n\nIf you want Dr. Claw to execute the generated task list end-to-end for you, use **Auto Research**:\n1. Open **Settings → Email** and configure `Notification Email`, `Sender Email`, and `Resend API Key`.\n2. Make sure your project already contains `.pipeline\u002Fdocs\u002Fresearch_brief.json` and `.pipeline\u002Ftasks\u002Ftasks.json`.\n3. Open the **Project Dashboard** and click **Auto Research** on the project card.\n4. Use **Open Session** to jump into the live Claude session created for the run.\n5. When all tasks finish, Dr. Claw sends a completion email. If the session is interrupted, stale runs are recovered automatically so they can be cancelled cleanly instead of staying stuck in `running`.\n\n\u003C\u002Fdetails>\n\n\u003Ca id=\"troubleshooting-web-search\">\u003C\u002Fa>\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>Step 4 — Troubleshooting Web Search\u003C\u002Fstrong>\u003C\u002Fsummary>\n\nIf the agent cannot search webpages, your current permission settings are likely too restrictive. Also check whether a runtime network lock is still active for the process.\n\n1. Check the runtime network lock:\n```bash\necho \"${CODEX_SANDBOX_NETWORK_DISABLED:-0}\"\n```\n\nIf the output is `1`, network requests can remain blocked even if Settings permissions are opened. Remove or override this variable in your deployment or startup layer (shell profile, systemd, Docker, PM2), then restart Dr. Claw.\n\n2. Open **Settings** (gear icon in sidebar).\n3. Go to **Permissions**, then choose your current agent:\n- **Claude Code**:\n  - Enable `WebSearch` and `WebFetch` in **Allowed Tools**.\n  - Ensure they are not present in **Blocked Tools**.\n  - Optionally enable **Skip permission prompts** if you want fewer confirmations.\n- **Gemini CLI**:\n  - Choose an appropriate **Permission Mode**.\n  - Allow `google_web_search` and `web_fetch` in **Allowed Tools** when web access is required.\n  - Ensure they are not present in **Blocked Tools**.\n- **Codex**:\n  - In **Permission Mode**, switch to **Bypass Permissions** when web access is required.\n4. Return to **Chat**, start a new message, and retry your web-search prompt.\n\nCodex permission mode notes:\n- **Default \u002F Accept Edits**: sandboxed execution; network may still be restricted by session policy.\n- **Bypass Permissions**: `sandboxMode=danger-full-access` with full disk and network access.\n\nSecurity note:\n- Use permissive settings only in trusted projects\u002Fenvironments.\n- After finishing web search tasks, switch back to safer settings.\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>Step 5 — Resolve \"Workspace Trust\" or First-Run Errors\u003C\u002Fstrong>\u003C\u002Fsummary>\n\nEach agent may require a one-time trust confirmation before it can execute code in your project directory. If Chat freezes or shows a trust prompt, switch to the **Shell** tab inside Dr. Claw and approve the prompt there.\n\nSteps:\n1. Switch to the **Shell** tab in Dr. Claw.\n2. Approve the trust\u002Fauth prompt shown in Shell.\n3. Return to **Chat** and resend your message.\n\nBy default, trust flow is already enabled in Dr. Claw, so you usually do **not** need to manually run extra trust commands.\n\nThe trust decision is persisted per directory — you only need to do this once per project.\n\n> **Shell tab not working?** If the Shell tab shows `Error: posix_spawnp failed`, see [docs\u002Ffaq.md](docs\u002Ffaq.md) for the fix, then retry.\n\nYou can switch tabs at any time:\n\n| Tab | What it does |\n|-----|-------------|\n| **Chat** | Start here. Use it to describe your research idea, generate a pipeline, and run tasks with the selected agent. |\n| **Survey** | Review papers, literature graphs, notes, and survey-stage tasks for the current project. |\n| **Research Lab** | Review the research brief, task list, progress, and generated artifacts in one place. |\n| **Skills** | Browse installed skills, inspect their contents, and import additional local skills. |\n| **Compute** | Manage compute resources and run experiment workloads from one place. |\n| **Shell** | Use the embedded terminal when you need direct CLI access, trust prompts, or manual commands. |\n| **Files** | Browse, open, create, rename, and edit project files with syntax highlighting. |\n| **Git** | Inspect diffs, stage changes, commit, and switch branches without leaving the app. |\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>Research Skills\u003C\u002Fstrong>\u003C\u002Fsummary>\n\nDr. Claw now uses the generated **Pipeline Task List** as the execution flow.\nThe project includes **100+ skills** under `skills\u002F` to support research tasks (idea exploration, code survey, experiment development\u002Fanalysis, writing, review, and delivery).\nThese skills are discovered by the agent and can be applied as task-level assistance throughout the workflow.\n\n\u003C\u002Fdetails>\n\n## Additional Details\n\u003Cdetails>\n\u003Csummary>\u003Cspan style=\"font-size: 1.17em; font-weight: 600;\">Mobile, architecture, and security notes\u003C\u002Fspan>\u003C\u002Fsummary>\n\n### Mobile & Tablet\n\nDr. Claw is fully responsive. On mobile devices:\n\n- **Bottom tab bar** for thumb-friendly navigation\n- **Swipe gestures** and touch-optimized controls\n- **Add to Home Screen** to use it as a PWA (Progressive Web App)\n\n### Architecture\n\n#### System Overview\n\n```\n┌─────────────────┐    ┌─────────────────┐    ┌─────────────────┐\n│   Frontend      │    │   Backend       │    │  Agent          │\n│   (React\u002FVite)  │◄──►│ (Express\u002FWS)    │◄──►│  Integration    │\n│                 │    │                 │    │                │\n└─────────────────┘    └─────────────────┘    └─────────────────┘\n```\n\n#### Backend (Node.js + Express)\n- **Express Server** - RESTful API with static file serving\n- **WebSocket Server** - Communication for chats and project refresh\n- **Agent Integration (Claude Code, Gemini CLI, Codex, OpenRouter)** - Process spawning, streaming, and session management\n- **File System API** - Exposing file browser for projects\n\n#### Frontend (React + Vite)\n- **React 18** - Modern component architecture with hooks\n- **CodeMirror** - Advanced code editor with syntax highlighting\n\n### Security & Tools Configuration\n\n**🔒 Important Notice**: Agent permissions are configurable per provider. Review **Settings → Permissions** before enabling broad file, shell, or web access.\n\n#### Enabling Tools\n\nTo use web and tool-heavy workflows safely:\n\n1. **Open Settings** - Click the gear icon in the sidebar\n2. **Choose an Agent** - Claude Code, Gemini CLI, or Codex\n3. **Enable Selectively** - Turn on only the tools or permission mode you need\n4. **Apply Settings** - Your preferences are saved locally\n\n**Recommended approach**: Start with the safest permission mode that still lets you complete the task, then relax settings only when needed.\n\n\u003C\u002Fdetails>\n\n## Contributing\n\u003Cdetails>\n\u003Csummary>\u003Cspan style=\"font-size: 1.17em; font-weight: 600;\">Show details\u003C\u002Fspan>\u003C\u002Fsummary>\n\nWe welcome contributions! Please follow these guidelines:\n\n#### Getting Started\n1. **Fork** the repository\n2. **Clone** your fork: `git clone \u003Cyour-fork-url>`\n3. **Install** dependencies: `npm install`\n4. **Create** a feature branch: `git checkout -b feature\u002Famazing-feature`\n\n#### Development Process\n1. **Make your changes** following the existing code style\n2. **Test thoroughly** - ensure all features work correctly\n3. **Run quality checks**: `npm run typecheck && npm run build`\n4. **Commit** with descriptive messages following [Conventional Commits](https:\u002F\u002Fconventionalcommits.org\u002F)\n5. **Push** to your branch: `git push origin feature\u002Famazing-feature`\n6. **Submit** a Pull Request with:\n   - Clear description of changes\n   - Screenshots for UI changes\n   - Test results if applicable\n\n#### What to Contribute\n- **Bug fixes** - Help us improve stability\n- **New features** - Enhance functionality (discuss in issues first)\n- **Documentation** - Improve guides and API docs\n- **UI\u002FUX improvements** - Better user experience\n- **Performance optimizations** - Make it faster\n\n\u003C\u002Fdetails>\n\nFor setup help and troubleshooting, see [FAQ](docs\u002Ffaq.md).\n\n## Legacy Compatibility & Deprecation\n\nDr. Claw was previously known as **VibeLab**. For users migrating from VibeLab, we provide a compatibility layer during the transition phase:\n\n- **CLI Alias**: The `vibelab` command is still supported as an alias for `drclaw` but will issue a deprecation warning.\n- **Python Package**: The `VibeLab` class in the `agent-harness` is deprecated; please use the `DrClaw` class instead.\n- **Session Files**: The CLI now defaults to `~\u002F.drclaw_session.json` but will automatically check for and migrate `~\u002F.vibelab_session.json` if found.\n- **Environment Variables**: `DRCLAW_URL` and `DRCLAW_TOKEN` are preferred, but `VIBELAB_URL` and `VIBELAB_TOKEN` are still supported as fallbacks.\n\n**Timeline**: We plan to remove legacy `vibelab` support in Version 2.0 (estimated Q3 2026). Please update your scripts and integrations as soon as possible.\n\n## License\n\nThis repository contains a combined work.\n\nUpstream portions derived from Claude Code UI remain under GNU General Public License v3.0 (GPL-3.0), while original modifications and additions by Dr. Claw Contributors are licensed under GNU Affero General Public License v3.0 (AGPL-3.0).\n\nSee [LICENSE](LICENSE) and [NOTICE](NOTICE) for the full license texts and scope details.\n\n## Citation\n\nIf you find Dr. Claw useful in your research, please cite:\n\n```bibtex\n@misc{song2026drclaw,\n  author       = {Dingjie Song and Hanrong Zhang and Dawei Liu and Yixin Liu and Zongxia Li and Zhengqing Yuan and Siqi Zhang and Lichao Sun},\n  title        = {Dr. Claw: An AI Research Workspace from Idea to Paper},\n  year         = {2026},\n  organization = {GitHub},\n  url          = {https:\u002F\u002Fgithub.com\u002FOpenLAIR\u002Fdr-claw},\n  homepage     = {https:\u002F\u002Fopenlair.github.io\u002Fdr-claw},\n}\n```\n\n## Acknowledgments\n\n### Built With\n- **[Claude Code](https:\u002F\u002Fdocs.anthropic.com\u002Fen\u002Fdocs\u002Fclaude-code)** - Anthropic's official CLI\n- **[Gemini CLI](https:\u002F\u002Fgeminicli.com\u002Fdocs\u002Fget-started\u002Finstallation\u002F)** - Google's Gemini command-line agent\n- **[Codex](https:\u002F\u002Fdevelopers.openai.com\u002Fcodex)** - OpenAI Codex\n- **[React](https:\u002F\u002Freact.dev\u002F)** - User interface library\n- **[Vite](https:\u002F\u002Fvitejs.dev\u002F)** - Fast build tool and dev server\n- **[Tailwind CSS](https:\u002F\u002Ftailwindcss.com\u002F)** - Utility-first CSS framework\n- **[CodeMirror](https:\u002F\u002Fcodemirror.net\u002F)** - Advanced code editor\n\n### Also Thanks To\n- **[Claude Code UI](https:\u002F\u002Fgithub.com\u002Fsiteboon\u002Fclaudecodeui)** — Dr. Claw is based on it. See [NOTICE](NOTICE) for details.\n- **[AI Researcher](https:\u002F\u002Fgithub.com\u002FHKUDS\u002FAI-Researcher\u002F)** (HKUDS) — Inspiration for research workflow and agentic research.\n- **[Vibe-Scholar](https:\u002F\u002Fgithub.com\u002FMr-Tieguigui\u002FVibe-Scholar)** — Inspiration for the AI-native research workspace direction.\n- **[autoresearch](https:\u002F\u002Fgithub.com\u002Fkarpathy\u002Fautoresearch)** — Inspiration for autonomous research orchestration and end-to-end execution.\n\n## Support & Community\n\n### Stay Updated\n- **Star** this repository to show support\n- **Watch** for updates and new releases\n- **Follow** the project for announcements\n\n---\n\n\u003Cdiv align=\"center\">\n  \u003Cstrong>Dr. Claw — From idea to paper.\u003C\u002Fstrong>\n\u003C\u002Fdiv>\n","\u003Cdiv align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FOpenLAIR_dr-claw_readme_5147600a5861.png\" alt=\"Dr. Claw\" width=\"128\" height=\"128\">\n  \u003Ch1>Dr. Claw：您的AI研究助手\u003C\u002Fh1>\n  \u003Cp>\u003Cstrong>全栈式研究工作空间。\u003C\u002Fstrong>\u003C\u002Fp>\n\u003C\u002Fdiv>\n\n\u003Cp align=\"center\">\n\u003Ca href=\"https:\u002F\u002Fopenlair.github.io\u002Fdr-claw\">\n\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F%F0%9F%8C%90-Homepage-CB2B3E?style=for-the-badge\" alt=\"首页\" \u002F>\n\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenLAIR\u002Fdr-claw\">\n\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F%F0%9F%A6%9E-Dr.%20Claw-CB2B3E?style=for-the-badge\" alt=\"Dr. Claw\" \u002F>\n\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenLAIR\u002Fdr-claw\u002Fblob\u002Fmain\u002FLICENSE\">\n\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-GPL--3.0%20%2B%20AGPL--3.0-blue?style=for-the-badge\" alt=\"许可证：GPL-3.0 + AGPL-3.0\" \u002F>\n\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fjoin.slack.com\u002Ft\u002Fvibe-lab-group\u002Fshared_invite\u002Fzt-3r4bkcx5t-iGyRMI~r09zt7p_ND2eP9A\">\n\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FJoin-Slack-4A154B?style=for-the-badge&logo=slack\" alt=\"加入Slack\" \u002F>\n\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fx.com\u002FVibe2038004\">\n\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FFollow-on%20X-black?style=for-the-badge&logo=x\" alt=\"在X上关注\" \u002F>\n\u003C\u002Fa>\n\u003Ca href=\".\u002Fpublic\u002Fwechat-group-qr.jpg\">\n\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FJoin-WeChat-07C160?style=for-the-badge&logo=wechat&logoColor=white\" alt=\"加入微信\" \u002F>\n\u003C\u002Fa>\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  \u003Ca href=\".\u002FREADME.md\">英文\u003C\u002Fa> | \u003Ca href=\".\u002FREADME.zh-CN.md\">中文\u003C\u002Fa>\n\u003C\u002Fp>\n\n## 目录\n\n- [概述](#overview)\n- [亮点](#highlights)\n- [快速入门](#quick-start)\n- [配置](#configuration)\n- [OpenRouter](#openrouter)\n- [OpenClaw集成](#openclaw-integration)\n- [研究实验室——快速示例](#research-lab-quick-example)\n- [使用指南](#usage-guide)\n- [更多细节](#additional-details)\n- [贡献](#contributing)\n- [常见问题](.\u002Fdocs\u002Ffaq.md)\n- [许可证](#license)\n- [引用](#citation)\n- [致谢](#acknowledgments)\n- [支持与社区](#support--community)\n\n## 概述\n\nDr. Claw是一款通用型AI研究助手，旨在帮助研究人员和开发者在不同领域开展端到端项目。从构思初始想法到执行实验并准备可发表的成果，Dr. Claw将整个工作流程整合在一个平台上，让团队能够专注于研究质量和迭代速度。\n\n## 产品截图\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FOpenLAIR_dr-claw_readme_b620dae813d2.png\" alt=\"Dr. Claw聊天界面\" width=\"1000\">\n\u003C\u002Fp>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>理念：杠杆认知\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FOpenLAIR_dr-claw_readme_bb46b412f1f1.png\" alt=\"杠杆认知\" width=\"900\">\n\u003C\u002Fp>\n\n**人工操作太慢，完全自动化的AI又过于泛化。Vibe Researching才是新的前沿。** Dr. Claw通过“代理执行”将您的“研究品味”转化为超大规模的成果——让您跑得更快、想得更大，同时依然坚守科学严谨性。\n\n\u003C\u002Fdetails>\n\n## 亮点\n\n- **🔬 研究实验室** — 结构化的仪表盘，支持端到端研究：定义研究简报、生成任务流水线、跟踪从调研→构思→实验→发表→推广的全流程进度，并可一键查看源论文、以LaTeX数学格式呈现的研究思路以及缓存产物——所有信息一目了然\n- **⚡ 自动研究** — 可直接从项目仪表盘一键启动连续任务执行，实时打开生成的会话，并在运行结束后收到邮件通知\n- **📚 100+项研究技能** — 精选的技能库涵盖创意生成、代码调研、实验开发与分析、论文撰写、审稿回复及成果交付——由智能代理自动发现并作为任务级辅助应用\n- **🗂️ 聊天驱动的流水线** — 在聊天中描述您的研究想法；代理会利用“inno-pipeline-planner”技能交互式生成结构化的研究简报和任务清单——无需手动模板\n- **🤖 多代理后端** — 可无缝切换Claude Code、Gemini CLI、Codex和OpenRouter作为执行引擎\n\n### 流水线产出\n\n| | 成果 | 位置 | 描述 |\n|---|---|---|---|\n| 📚 | 调研报告 | `Survey\u002Freports\u002F` | 带有arXiv、Semantic Scholar及网络来源引用的文献综述 |\n| 💡 | 研究想法 | `Ideation\u002Fideas\u002F` | 带有多角色评估分数的头脑风暴成果 |\n| 🔬 | 实验代码 | `Experiment\u002Fcore_code\u002F` | 从计划到实现再到评估的循环 |\n| 📊 | 分析结果 | `Experiment\u002Fanalysis\u002F` | 统计分析、表格及可发表的图表 |\n| 📝 | 论文草稿 | `Publication\u002Fpaper\u002F` | 学术手稿（IEEE\u002FACM格式），包含引用和LaTeX数学 |\n| 🎞️ | 演示文稿 | `Promotion\u002Fslides\u002F` | 幻灯片演示、TTS旁白音频及演示视频 |\n\n> 请参阅[docs\u002Fpipeline-outputs.md](docs\u002Fpipeline-outputs.md)了解完整的成果列表及项目目录结构。\n\n\u003Cdetails>\n\u003Csummary>\u003Cspan style=\"font-size: 1.17em; font-weight: 600;\">更多功能\u003C\u002Fspan>\u003C\u002Fsummary>\n\n- **💬 交互式聊天+终端** — 可与代理聊天或直接进入完整终端——与研究上下文并排显示\n- **📁 文件与Git浏览器** — 支持语法高亮的文件浏览、实时编辑、暂存更改、提交及切换分支，全程无需离开UI\n- **📱 响应式设计与PWA支持** — 适配桌面、平板及移动端布局，配备底部标签栏、滑动手势及添加到主屏幕支持\n- **🔄 会话管理** — 支持恢复对话、管理多个会话，并跨项目追踪完整历史记录\n\n\u003C\u002Fdetails>\n\n### 功能图集\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>展开截图\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n\u003Cp>\u003Cstrong>项目仪表盘\u003C\u002Fstrong> — 从项目概览开始，查看状态并启动端到端自动化。\u003C\u002Fp>\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FOpenLAIR_dr-claw_readme_a6f713435c62.png\" alt=\"项目仪表盘界面\" width=\"1000\">\n\u003C\u002Fp>\n\n\u003Cp>\u003Cstrong>技能库\u003C\u002Fstrong> — 浏览可用于构思、实验及撰写的可复用研究技能。\u003C\u002Fp>\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FOpenLAIR_dr-claw_readme_66752d79fcbb.png\" alt=\"技能库界面\" width=\"1000\">\n\u003C\u002Fp>\n\n\u003Cp>\u003Cstrong>新闻仪表盘\u003C\u002Fstrong> — 在不离开工作区的情况下关注研究相关动态。\u003C\u002Fp>\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FOpenLAIR_dr-claw_readme_3c254c2a7a78.png\" alt=\"新闻仪表盘界面\" width=\"1000\">\n\u003C\u002Fp>\n\n\u003C\u002Fdetails>\n\n\n## 快速入门\n\n### 先决条件\n\n- [Node.js](https:\u002F\u002Fnodejs.org\u002F) v20及以上版本（**推荐使用v22 LTS**，详见`.nvmrc`）\n- 至少安装并配置以下CLI工具之一：\n  - [Claude Code CLI](https:\u002F\u002Fdocs.anthropic.com\u002Fen\u002Fdocs\u002Fclaude-code)\n  - [Gemini CLI](https:\u002F\u002Fgeminicli.com\u002Fdocs\u002Fget-started\u002Finstallation\u002F)\n  - [Codex CLI](https:\u002F\u002Fdevelopers.openai.com\u002Fcodex\u002Fcli\u002F)\n- 部分系统需要原生构建工具来处理`node-pty`和`better-sqlite3`等依赖。如果`npm install`失败，请参阅[常见问题](docs\u002Ffaq.md)。\n\nCursor代理支持正在开发中，即将上线。\n\n### 安装\n\n1. **克隆仓库：**\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FOpenLAIR\u002Fdr-claw.git\ncd dr-claw\n```\n\n2. **安装依赖：**\n```bash\nnpm install\n```\n\n3. **配置环境：**\n```bash\ncp .env.example .env\n# 编辑 .env 文件，根据需要设置端口等参数\n```\n\n如果需要自定义端口、认证或工作空间设置，请参阅 [docs\u002Fconfiguration.md](docs\u002Fconfiguration.md)。\n\n4. **启动应用：**\n\n```bash\n# 开发模式（带热重载）\nnpm run dev\n```\n\n然后通过浏览器访问 `http:\u002F\u002Flocalhost:5173` 创建您的账户。\n\n5. **使用应用**\n\n与 Dr. Claw 的交互方式有两种：**前端 UI 流程**和**纯终端模式**。UI 提供更丰富的可视化体验，但可能会偶尔出现 bug；而终端模式则更加稳定、轻量。\n\n#### 方案 A：前端 UI\n\n在浏览器中打开 `http:\u002F\u002Flocalhost:5173`（或您在 `.env` 中配置的端口）。\n\n\n\n#### 方案 B：纯终端模式\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FOpenLAIR_dr-claw_readme_937babfcfd12.png\" alt=\"终端流程示例\" width=\"800\">\n\u003C\u002Fp>\n\n打开一个**第二个终端**（第一个终端保持运行 `npm run dev`），并安装 `drclaw` CLI 工具包：\n\n```bash\npip install -e .\u002Fagent-harness\n```\n\n然后使用您在设置过程中创建的凭据登录：\n\n```bash\ndrclaw auth login --username YOUR_USERNAME --password YOUR_PASSWORD\n```\n\n至少安装一个代理 CLI（如果您尚未安装）：\n\n| 代理 | 安装 | 认证 |\n|-------|---------|------|\n| Claude Code | `npm install -g @anthropic-ai\u002Fclaude-code` | `claude` → 按照 OAuth 提示操作 |\n| Gemini CLI | `npm install -g @google\u002Fgemini-cli` | `gemini` → Google 登录，或 `export GOOGLE_API_KEY=...` |\n| Codex CLI | `npm install -g @openai\u002Fcodex` | `codex login`，或 `export OPENAI_API_KEY=...` |\n| **OpenRouter** | 无需 CLI | `export OPENROUTER_API_KEY=sk-or-...`（请在 [openrouter.ai\u002Fkeys](https:\u002F\u002Fopenrouter.ai\u002Fkeys) 获取密钥） |\n\n> **OpenRouter** 允许您通过一个 API 密钥使用*任何*模型（GPT-5、Claude、Gemini、DeepSeek、Llama、Mistral、Qwen 等）。您可以在 UI 中选择模型，或在 `.env` 中设置 `OPENROUTER_MODEL`。\n\n进入您想要工作的项目目录，并启动任意一个代理：\n\n```bash\ncd \u002Fpath\u002Fto\u002Fyour\u002Fproject\nclaude    # 或：gemini | codex\n```\n\n`dr-claw\u002Fskills\u002F` 中的技能会在项目创建时自动符号链接到每个项目的 `.claude\u002Fskills\u002F` 目录，因此代理无需额外配置即可发现这些技能。您也可以在会话中手动引用任意技能：\n\n```\n> 阅读 .claude\u002Fskills\u002Finno-experiment-analysis\u002FSKILL.md 并按照其指示分析我的结果。\n```\n\n\n\n#### 方案 C：OpenRouter 终端聊天\n\n若希望以轻量级的纯终端模式使用任意 [OpenRouter](https:\u002F\u002Fopenrouter.ai\u002F) 模型，可使用内置的 `dr-claw chat` 命令。无需浏览器或 UI——只需进行一次具备完整工具调用能力的交互式代理会话（文件 I\u002FO、Shell、grep、glob、网络搜索\u002F获取）。\n\n```bash\n# 确保已设置 OPENROUTER_API_KEY（或直接传入 --key 参数）\nexport OPENROUTER_API_KEY=sk-or-...\n\n# 启动与任意模型的聊天会话\nnode server\u002Fcli.js chat --model moonshotai\u002Fkimi-k2.5\n```\n\n您也可以直接在命令行中传入 API 密钥：\n\n```bash\nnode server\u002Fcli.js chat --model anthropic\u002Fclaude-sonnet-4 --key sk-or-your-key\n```\n\n| 标志 | 描述 |\n|------|-------------|\n| `--model \u003Cslug>` | OpenRouter 模型标识符（例如 `moonshotai\u002Fkimi-k2.5`、`anthropic\u002Fclaude-sonnet-4`、`deepseek\u002Fdeepseek-r1`） |\n| `--key \u003Ckey>` | OpenRouter API 密钥（默认为 `OPENROUTER_API_KEY` 环境变量） |\n\n您可以在 [openrouter.ai\u002Fmodels](https:\u002F\u002Fopenrouter.ai\u002Fmodels) 查看所有可用模型。\n\n\n\n如果后续代理的网络搜索无法正常工作，请参阅下方的【网络搜索故障排除】。\n\n## OpenClaw 集成\n\n本节专为**首次将 OpenClaw 与 Dr. Claw 集成的新用户**编写。我们的目标并非面面俱到，而是帮助您快速完成一次可靠的首次集成：\n\n- OpenClaw 能够查看现有的 Dr. Claw 项目；\n- OpenClaw 能够找到等待用户输入的会话；\n- OpenClaw 能够回复到选定的会话，使 Dr. Claw 继续运行；\n- OpenClaw 能够总结项目\u002F组合进展，并推荐下一步的重点方向。\n\n清晰的思维模型是：\n- **Dr. Claw**：拥有真实的项目、会话、流水线、产出物以及执行过程；\n- **`drclaw` CLI**：通过稳定的本地控制界面暴露这些状态，主要以 `drclaw` 命令呈现；\n- **OpenClaw**：作为面向用户的秘书，可通过移动端、聊天或语音提供服务。\n\n### 新用户的最快路径\n\n如果您只想走最短的成功路径，只需完成以下五步：\n\n1. 启动 Dr. Claw；\n2. 安装 `drclaw` CLI；\n3. 授予 OpenClaw 本地 Shell \u002F `exec` 访问权限；\n4. 安装提供的 OpenClaw 技能；\n5. 确保 `chat waiting` 和 `digest portfolio` 能够端到端正常运行。\n\n一旦这两条命令能够正常工作，OpenClaw 就已经可以像一个可用的研究秘书一样发挥作用。\n\n### 步骤 0：确认先决条件\n\n在集成之前，请确保：\n- 您已经能够在本地运行 Dr. Claw；\n- 您已经拥有至少一个项目，或者可以在 `~\u002Fvibelab\u002F...` 下创建一个；\n- 您已经配置了至少一个执行后端，如 Claude Code、Gemini CLI 或 Codex；\n- 您的 OpenClaw 实例被允许运行本地工具。\n\n如果以上条件尚未满足，请先让 Dr. Claw 本身正常运行。\n\n### 步骤 1：启动并验证 Dr. Claw 服务器\n\n从仓库根目录开始：\n\n```bash\nnpm install\nnpm run dev\n```\n\n在另一个终端中：\n\n```bash\ndrclaw --json auth status\ndrclaw server status\n```\n\n优先使用 `auth status` 作为可达性检查。如果返回 JSON，则说明服务器已可访问。\n\n`drclaw server status` 只会报告由 `drclaw server on` 启动的守护进程。如果您是手动通过 `npm run dev` 启动的应用程序，即使 `http:\u002F\u002Flocalhost:3001` 正常运行，它也可能显示 `STOPPED`。\n\n如果您希望 Dr. Claw 自动管理后台进程：\n\n```bash\ndrclaw server on\n```\n\n### 步骤 2：安装并验证 `drclaw` CLI\n\n从仓库根目录开始：\n\n```bash\npip install -e .\u002Fagent-harness\n```\n\n然后进行验证：\n\n```bash\ndrclaw --help\ndrclaw --json auth status\ndrclaw --json projects list\n```\n\n理想的顺序是：\n- `drclaw --json auth status` 应返回 JSON，表明服务器已可访问；\n- `drclaw --json projects list` 只有在登录后或已有保存的令牌时才会生效。\n\n如果 `projects list` 返回 `Not logged in`，请先进行认证：\n\n```bash\ndrclaw auth login --username \u003Cusername> --password \u003Cpassword>\n```\n\n### 第3步：赋予OpenClaw本地CLI执行能力\n\n关键的集成并不在于构建一个深度的API桥接。关键在于**OpenClaw能够直接在本地执行`drclaw ...`命令**。\n\n至少，OpenClaw应当能够运行如下命令：\n\n```bash\ndrclaw --json chat waiting\ndrclaw --json digest portfolio\ndrclaw --json chat reply --project \u003Cproject> --session \u003Csession-id> -m \"\u003Cmessage>\"\ndrclaw --json workflow continue --project \u003Cproject> --session \u003Csession-id> -m \"\u003Cinstruction>\"\n```\n\n推荐的做法：\n- 为OpenClaw启用本地`exec`\u002Fshell功能\n- 尽量采用直接的本地CLI调用\n- 避免一开始就搭建额外的代理层\n\n这一层越薄，调试就越容易，可靠性也越高。\n\n### 第4步：通过一条命令完成OpenClaw的链接\n\n运行：\n\n```bash\ndrclaw install --server-url http:\u002F\u002Flocalhost:3001\n```\n\n这将自动完成以下操作：\n- 将Dr. Claw技能复制到`~\u002F.openclaw\u002Fworkspace\u002Fskills\u002Fdrclaw`\n- 安装OpenClaw用于序列化本地轮次交互的辅助脚本\n- 保存Dr. Claw服务器URL，以便后续CLI或OpenClaw使用\n- 记住本地`drclaw`可执行文件的路径\n\n如果在设置过程中还想保存默认的推送通道：\n\n```bash\ndrclaw install --server-url http:\u002F\u002Flocalhost:3001 --push-channel feishu:\u003Cchat_id>\n```\n\n兼容性形式仍然可用：\n\n```bash\ndrclaw openclaw install --server-url http:\u002F\u002Flocalhost:3001\n```\n\n### 第5步：先让两个核心命令正常工作\n\n对于新用户，不要一开始就尝试所有功能。先从这两个命令入手：\n\n1. **查找哪些会话正在等待用户输入**\n```bash\ndrclaw --json chat waiting\n```\n\n2. **获取投资组合层面的进展与建议**\n```bash\ndrclaw --json digest portfolio\n```\n\n如果OpenClaw能够调用这两个命令，并将结果汇总后反馈给用户，那么你的最小可行集成就已经成功了。\n\n### 第6步：添加回复循环\n\n接下来常见的用户操作是：查看有等待的会话，然后让OpenClaw为其作答。\n\n固定的流程是：\n\n1. 查找等待中的会话：\n```bash\ndrclaw --json chat waiting\n```\n\n2. 让用户选择项目和会话\n\n3. 发送回复：\n```bash\ndrclaw --json chat reply --project \u003Cproject> --session \u003Csession-id> -m \"\u003Cmessage>\"\n```\n\n4. 立即再次检查是否仍处于等待状态：\n```bash\ndrclaw --json chat waiting --project \u003Cproject>\n```\n\n如果用户希望继续讨论同一个项目\u002F会话，则切换为：\n```bash\ndrclaw --json chat project --project \u003Cproject> --session \u003Csession-id> -m \"\u003Cinstruction>\"\n```\n\n这是进行多轮、基于项目的讨论时更优的模式。\n\n### 第7步：推荐的固定操作模式\n\n推荐的OpenClaw使用流程：\n\n1. **用户提问：目前有哪些事项需要我关注？**\n```bash\ndrclaw --json digest portfolio\n```\n\n2. **用户提问：哪些会话正在等我？**\n```bash\ndrclaw --json chat waiting\n```\n\n3. **用户提问：这个项目的最新状态如何？**\n```bash\ndrclaw --json projects latest \u003Cproject>\ndrclaw --json projects progress \u003Cproject>\n```\n\n4. **用户说：回复这个会话并推动其继续进展**\n```bash\ndrclaw --json chat reply --project \u003Cproject> --session \u003Csession-id> -m \"\u003Cmessage>\"\n```\n\n5. **用户说：我刚有个新想法，创建一个项目并帮我把它落地**\n```bash\ndrclaw --json projects idea \u002Fabsolute\u002Fpath\u002Fto\u002Fproject --name \"\u003Cdisplay-name>\" --idea \"\u003Cidea text>\"\n```\n\n### 第8步：优先采用序列化的本地轮次\n\n当OpenClaw反复运行`openclaw agent --local`时，应使用序列化封装器来避免会话锁定冲突：\n\n```bash\nagent-harness\u002Fskills\u002Fdr-claw\u002Fscripts\u002Fopenclaw_drclaw_turn.sh\n```\n\n示例：\n\n```bash\nopenclaw_drclaw_turn.sh --json -m \"使用你的exec工具运行`drclaw --json digest portfolio`。仅返回原始stdout。\"\n```\n\n实际上：**当OpenClaw在本地调用Dr. Claw时，稳定的串行轮次比冒险的并行轮次更可靠。**\n\n### 第9步：如何判断集成是否成功\n\n新用户可以认为集成已完成，只要满足以下四个条件：\n- OpenClaw能够列出Dr. Claw的项目\n- OpenClaw能够识别出等待中的会话\n- OpenClaw能够成功向选定的会话发送一条回复\n- OpenClaw能够生成一份包含建议的“digest portfolio”式总结\n\n到那时，OpenClaw就不再只是一个聊天界面，而真正成为了Dr. Claw的移动秘书。\n\n### 第10步：最终用户可以怎么说\n\n设置完成后，用户应该能够自然地与OpenClaw对话：\n- “帮我看看有哪些Dr. Claw项目正在等我回复。”\n- “把这个项目的最新消息和当前进展给我总结一下。”\n- “回复这个会话：继续选项B，完成后告诉我结果。”\n- “总结一下最近各项目的实验进展，并推荐我今天该重点关注什么。”\n- “我刚有个新想法，帮我创建一个Dr. Claw项目，和我一起讨论、打磨，然后开始制定执行计划。”\n\n我们的目标不是取代Dr. Claw，而是让Dr. Claw通过OpenClaw变得**可调用、可汇报、可引导、可远程管理**。\n\n## 配置\n\nDr. Claw从`.env`文件中读取本地配置。对大多数用户来说，唯一需要做的就是将`.env.example`复制到`.env`，但以下这些配置是你最早可能需要调整的：\n\n- `PORT`：后端服务器端口\n- `VITE_PORT`：前端开发服务器端口\n- `HOST`：前端和后端的绑定地址\n- `JWT_SECRET`：在将Dr. Claw暴露到本地以外之前必需\n- `WORKSPACES_ROOT`：新建项目工作空间的默认根目录\n\n完整的环境参考及部署说明，请参阅[docs\u002Fconfiguration.md](docs\u002Fconfiguration.md)。\n\n自动研究的邮件通知在应用内通过**设置→邮件**进行配置。v1版本支持Claude Code、Codex、Gemini和OpenRouter引擎进行无人值守的任务执行，且中断的运行会自动恢复，不会一直卡在“运行中”状态。\n\n## OpenRouter\n\n[OpenRouter](https:\u002F\u002Fopenrouter.ai\u002F)被集成为一级提供商，让你只需一个API密钥就能访问**数百种模型**（GPT-5、Claude、Gemini、DeepSeek、Llama、Mistral、Qwen、Kimi等）。\n\n### 设置\n\n1. 在[openrouter.ai\u002Fkeys](https:\u002F\u002Fopenrouter.ai\u002Fkeys)获取一个API密钥。\n2. 通过三种方式之一设置密钥：\n   - **环境变量**：`export OPENROUTER_API_KEY=sk-or-...`\n   - **`.env`文件**：在项目`.env`中添加`OPENROUTER_API_KEY=sk-or-...`\n   - **UI**：进入**设置→OpenRouter**，粘贴你的密钥\n\n### 在UI中使用OpenRouter\n\n1. 打开一个项目，进入**聊天**。\n2. 在**选择你的AI助手**下，点击**OpenRouter**。\n3. 在下拉菜单中搜索模型（会从OpenRouter获取完整列表），或输入自定义模型的slug。\n4. 开始聊天——该代理具备与Claude、Gemini和Codex相同的工具调用能力（文件读写、shell、grep、glob、网络搜索\u002F抓取、待办事项）。\n\nOpenRouter也在项目仪表板的**自动研究**中可用——选择它作为提供商，并挑选任意模型。\n\n### 在终端中使用 OpenRouter\n\n无需浏览器。`dr-claw chat` CLI 为您提供一个完全自主的终端会话：\n\n```bash\n# 基本用法\nnode server\u002Fcli.js chat --model moonshotai\u002Fkimi-k2.5\n\n# 使用显式 API 密钥\nnode server\u002Fcli.js chat --model deepseek\u002Fdeepseek-r1 --key sk-or-your-key\n```\n\n该 CLI 支持与 UI 相同的工具（文件 I\u002FO、Shell、grep、glob、网页搜索、网页抓取、待办事项）。输入您的消息，代理将自动执行多步骤的研究任务。\n\n### 默认模型\n\n在 `.env` 文件中设置 `OPENROUTER_MODEL`，以更改未指定模型时使用的默认模型：\n\n```env\nOPENROUTER_MODEL=moonshotai\u002Fkimi-k2.5\n```\n\n如果未设置，则默认为 `anthropic\u002Fclaude-sonnet-4`。\n\n\u003Ca id=\"research-lab-quick-example\">\u003C\u002Fa>\n\n## 研究实验室 — 快速示例\n\nDr. Claw 的核心功能是 **研究实验室**。\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>研究实验室截图\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FOpenLAIR_dr-claw_readme_2f54d80ef79c.png\" alt=\"研究实验室工作流\" width=\"1000\">\n\u003C\u002Fp>\n\n\u003C\u002Fdetails>\n\n典型流程如下：\n\n1. 在 **设置** 中配置一个支持的代理。\n2. 如果需要完成邮件通知，可在 **设置 → 邮件** 中配置通知设置。\n3. 在 **聊天** 中描述您的研究想法。\n4. 让代理生成 `.pipeline\u002Fdocs\u002Fresearch_brief.json` 和 `.pipeline\u002Ftasks\u002Ftasks.json`。\n5. 在 **研究实验室** 中查看流水线，并手动将任务发回 **聊天**，或在项目仪表板上点击 **自动研究** 以按顺序运行这些任务。\n\n有关完整的分步操作，请参阅下方的 **使用指南**。\n\n## 使用指南\n\n启动 Dr. Claw 后，打开浏览器并按照以下步骤操作。\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>步骤 1 — 创建或打开项目\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n首次打开 Dr. Claw 时，您将看到 **项目** 侧边栏。您有两种选择：\n\n- **打开现有项目** — Dr. Claw 会自动发现从 Claude Code、Codex 和 Gemini 注册的项目及关联的会话。\n- **创建新项目** — 点击 **“+”** 按钮，选择您电脑上的一个目录，Dr. Claw 将设置工作空间：包括 `.claude\u002F`、`.agents\u002F`、`.gemini\u002F` 等代理文件夹，标准工作空间元数据，关联的 `skills\u002F` 目录，预设的研究目录（调查\u002F参考文献、调查\u002F报告、构思\u002F创意、构思\u002F参考文献、实验\u002F代码参考、实验\u002F数据集、实验\u002F核心代码、实验\u002F分析、出版\u002F论文、推广\u002F主页、推广\u002F幻灯片、推广\u002F音频、推广\u002F视频），以及项目根目录下的 **instance.json** 文件，其中包含这些目录的绝对路径。光标代理支持即将推出。\n\n> **默认项目存储路径：** 新项目默认存储在 `~\u002Fdr-claw` 下。您可以在 **设置 → 外观 → 默认项目路径** 中更改此设置，或设置 `WORKSPACES_ROOT` 环境变量。该设置会持久化在 `~\u002F.claude\u002Fproject-config.json` 中。\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>步骤 2 — 通过聊天生成您的研究流水线\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n创建或打开项目后，Dr. Claw 默认会打开 **聊天**。如果尚未存在研究流水线，会出现一个引导横幅，其中包含 **在聊天中使用** 按钮，用于注入一个入门提示。\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>聊天截图\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FOpenLAIR_dr-claw_readme_7dae94904768.png\" alt=\"聊天界面\" width=\"1000\">\n\u003C\u002Fp>\n\n\u003C\u002Fdetails>\n\n描述您的研究想法——即使是粗略的想法也没关系。代理会使用 `inno-pipeline-planner` 技能提出澄清问题，然后生成：\n- `.pipeline\u002Fdocs\u002Fresearch_brief.json`（您的结构化研究简报）\n- `.pipeline\u002Ftasks\u002Ftasks.json`（任务流水线）\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>步骤 3 — 在研究实验室中查看并执行任务\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n切换到 **研究实验室**，查看生成的任务、进度指标和成果。然后执行任务：\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>任务执行截图\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FOpenLAIR_dr-claw_readme_264d8c2b05ba.png\" alt=\"任务列表与执行流程\" width=\"1000\">\n\u003C\u002Fp>\n\n\u003C\u002Fdetails>\n\n1. 从 **CLI 选择器** 中选择一个 CLI 后端（Claude Code、Gemini CLI 或 Codex）。\n2. 在 **研究实验室** 中，点击待处理任务上的 **前往聊天** 或 **在聊天中使用**。\n3. 代理执行任务并将结果写回项目。\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>可选 — 从项目仪表板运行自动研究\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n如果您希望 Dr. Claw 为您端到端地执行生成的任务列表，可以使用 **自动研究**：\n1. 打开 **设置 → 邮件**，配置 `通知邮箱`、`发件人邮箱` 和 `重发 API 密钥`。\n2. 确保您的项目已包含 `.pipeline\u002Fdocs\u002Fresearch_brief.json` 和 `.pipeline\u002Ftasks\u002Ftasks.json`。\n3. 打开 **项目仪表板**，并在项目卡片上点击 **自动研究**。\n4. 使用 **打开会话** 跳转到为此次运行创建的实时 Claude 会话。\n5. 当所有任务完成后，Dr. Claw 会发送一封完成邮件。如果会话中断，过时的运行会自动恢复，以便能够干净地取消，而不会一直卡在“运行中”状态。\n\n\u003C\u002Fdetails>\n\n\u003Ca id=\"troubleshooting-web-search\">\u003C\u002Fa>\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>步骤 4 — 解决网页搜索问题\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n如果代理无法搜索网页，很可能是当前的权限设置过于严格。同时请检查进程是否仍处于运行时网络锁定状态。\n\n1. 检查运行时网络锁定：\n```bash\necho \"${CODEX_SANDBOX_NETWORK_DISABLED:-0}\"\n```\n\n如果输出为 `1`，即使已开放设置权限，网络请求仍可能被阻止。请在部署或启动层（shell 配置文件、systemd、Docker、PM2）中移除或覆盖此变量，然后重启 Dr. Claw。\n\n2. 打开 **设置**（侧边栏中的齿轮图标）。\n3. 进入 **权限**，然后选择您当前的代理：\n- **Claude Code**：\n  - 在 **允许的工具** 中启用 `WebSearch` 和 `WebFetch`。\n  - 确保它们未出现在 **阻止的工具** 中。\n  - 如需减少确认次数，可选择 **跳过权限提示**。\n- **Gemini CLI**：\n  - 选择合适的 **权限模式**。\n  - 在需要访问网页时，在 **允许的工具** 中允许 `google_web_search` 和 `web_fetch`。\n  - 确保它们未出现在 **阻止的工具** 中。\n- **Codex**：\n  - 在 **权限模式** 中，当需要访问网页时切换到 **绕过权限**。\n4. 返回 **聊天**，开始新消息，并重新尝试您的网页搜索提示。\n\nCodex 权限模式说明：\n- **默认 \u002F 接受编辑**：沙箱执行；网络仍可能受会话策略限制。\n- **绕过权限**：`sandboxMode=danger-full-access`，具有完全的磁盘和网络访问权限。\n\n安全提示：\n- 仅在可信的项目\u002F环境中使用宽松设置。\n- 完成网页搜索任务后，应切换回更安全的设置。\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>步骤 5 — 解决“工作空间信任”或首次运行错误\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n每个代理在您的项目目录中执行代码之前，可能都需要进行一次信任确认。如果 Chat 卡顿或弹出信任提示，请切换到 Dr. Claw 内的 **Shell** 选项卡，并在那里批准该提示。\n\n步骤：\n1. 切换到 Dr. Claw 的 **Shell** 选项卡。\n2. 在 Shell 中批准显示的信任\u002F身份验证提示。\n3. 返回 **Chat** 并重新发送您的消息。\n\n默认情况下，Dr. Claw 已启用信任流程，因此您通常无需手动运行额外的信任命令。\n\n信任决策会按目录持久化——每个项目只需执行一次即可。\n\n> **Shell 选项卡无法使用？** 如果 Shell 选项卡显示“错误：posix_spawnp 失败”，请参阅 [docs\u002Ffaq.md](docs\u002Ffaq.md) 获取修复方法，然后重试。\n\n您可以随时切换选项卡：\n\n| 选项卡 | 功能 |\n|-----|-------------|\n| **Chat** | 从这里开始。用于描述您的研究想法、生成流水线并使用选定的代理运行任务。 |\n| **Survey** | 查看当前项目的论文、文献图谱、笔记以及调研阶段的任务。 |\n| **Research Lab** | 在一处查看研究简报、任务列表、进展及生成的成果。 |\n| **Skills** | 浏览已安装的技能、检查其内容并导入额外的本地技能。 |\n| **Compute** | 在一处管理计算资源并运行实验工作负载。 |\n| **Shell** | 当需要直接访问 CLI、处理信任提示或执行手动命令时，使用内置终端。 |\n| **Files** | 浏览、打开、创建、重命名并编辑项目文件，支持语法高亮。 |\n| **Git** | 在不离开应用的情况下查看差异、暂存更改、提交并切换分支。 |\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>研究技能\u003C\u002Fstrong>\u003C\u002Fsummary>\n\nDr. Claw 现在将生成的 **流水线任务列表** 作为执行流程。\n该项目在 `skills\u002F` 目录下包含 **100 多项技能**，以支持研究任务（想法探索、代码调研、实验开发\u002F分析、写作、评审与交付）。\n这些技能由代理发现，并可在整个工作流中作为任务级辅助加以应用。\n\n\u003C\u002Fdetails>\n\n\n\n## 其他详细信息\n\u003Cdetails>\n\u003Csummary>\u003Cspan style=\"font-size: 1.17em; font-weight: 600;\">移动端、架构与安全注意事项\u003C\u002Fspan>\u003C\u002Fsummary>\n\n### 移动端与平板\n\nDr. Claw 完全响应式设计。在移动设备上：\n\n- **底部标签栏**，便于用拇指操作导航\n- **滑动手势**和触控优化的控件\n- **添加到主屏幕**，可将其用作 PWA（渐进式 Web 应用）\n\n### 架构\n\n#### 系统概览\n\n```\n┌─────────────────┐    ┌─────────────────┐    ┌─────────────────┐\n│   前端      │    │   后端       │    │  代理          │\n│   (React\u002FVite)  │◄──►│ (Express\u002FWS)    │◄──►│  集成    │\n│                 │    │                 │    │                │\n└─────────────────┘    └─────────────────┘    └─────────────────┘\n```\n\n#### 后端（Node.js + Express）\n- **Express 服务器** - 提供 RESTful API 并支持静态文件服务\n- **WebSocket 服务器** - 用于聊天和项目刷新的通信\n- **代理集成（Claude Code、Gemini CLI、Codex、OpenRouter）** - 进程启动、流式传输及会话管理\n- **文件系统 API** - 暴露项目文件浏览器\n\n#### 前端（React + Vite）\n- **React 18** - 使用 Hooks 的现代组件架构\n- **CodeMirror** - 支持语法高亮的高级代码编辑器\n\n### 安全与工具配置\n\n**🔒 重要提示**：代理权限可根据不同提供商进行配置。在启用广泛的文件、Shell 或 Web 访问权限之前，请先查看 **设置 → 权限**。\n\n#### 启用工具\n\n为安全地使用 Web 和工具密集型工作流：\n\n1. 打开设置 - 点击侧边栏中的齿轮图标\n2. 选择代理 - Claude Code、Gemini CLI 或 Codex\n3. 有选择地启用 - 仅开启您所需的工具或权限模式\n4. 应用设置 - 您的偏好将被本地保存\n\n**推荐做法**：先从最安全的权限模式入手，确保能够完成任务，随后再根据需要逐步放宽设置。\n\n\u003C\u002Fdetails>\n\n## 贡献\n\u003Cdetails>\n\u003Csummary>\u003Cspan style=\"font-size: 1.17em; font-weight: 600;\">显示详情\u003C\u002Fspan>\u003C\u002Fsummary>\n\n我们欢迎贡献！请遵循以下指南：\n\n#### 入门\n1. **Fork** 仓库\n2. **Clone** 您的 Fork：`git clone \u003Cyour-fork-url>`\n3. **安装** 依赖：`npm install`\n4. **创建** 功能分支：`git checkout -b feature\u002Famazing-feature`\n\n#### 开发流程\n1. **进行修改**，遵循现有代码风格\n2. **全面测试** - 确保所有功能正常运行\n3. **运行质量检查**：`npm run typecheck && npm run build`\n4. **提交**，使用符合 [Conventional Commits](https:\u002F\u002Fconventionalcommits.org\u002F) 规范的描述性提交信息\n5. **推送** 至您的分支：`git push origin feature\u002Famazing-feature`\n6. **提交 Pull Request**，需包含：\n   - 清晰的变更说明\n   - UI 变更的截图\n   - 如适用，测试结果\n\n#### 可贡献的内容\n- **Bug 修复** - 帮助我们提升稳定性\n- **新功能** - 增强功能（请先在 Issue 中讨论）\n- **文档** - 改善指南和 API 文档\n- **UI\u002FUX 改进** - 提升用户体验\n- **性能优化** - 让它运行得更快\n\n\u003C\u002Fdetails>\n\n如需设置帮助与故障排除，请参阅 [FAQ](docs\u002Ffaq.md)。\n\n## 兼容性与弃用\n\nDr. Claw 曾名为 **VibeLab**。对于从 VibeLab 迁移的用户，我们在过渡阶段提供兼容层：\n\n- **CLI 别名**：`vibelab` 命令仍作为 `drclaw` 的别名被支持，但会发出弃用警告。\n- **Python 包**：`agent-harness` 中的 `VibeLab` 类已被弃用，请改用 `DrClaw` 类。\n- **会话文件**：CLI 现已默认使用 `~\u002F.drclaw_session.json`，但如果检测到 `~\u002F.vibelab_session.json`，则会自动检查并迁移。\n- **环境变量**：优先使用 `DRCLAW_URL` 和 `DRCLAW_TOKEN`，但 `VIBELAB_URL` 和 `VIBELAB_TOKEN` 仍作为备用方案被支持。\n\n**时间表**：我们计划在 Version 2.0（预计 2026 年第三季度）中移除对旧版 `vibelab` 的支持。请尽快更新您的脚本和集成。\n\n## 许可证\n\n本仓库包含合并作品。\n\n源自 Claude Code UI 的上游部分仍受 GNU 通用公共许可证 v3.0（GPL-3.0）约束，而 Dr. Claw 贡献者所做的原始修改与新增内容则依据 GNU Affero 通用公共许可证 v3.0（AGPL-3.0）授权。\n\n完整许可文本及适用范围详情，请参阅 [LICENSE](LICENSE) 和 [NOTICE](NOTICE)。\n\n## 引用\n\n如果您在研究中发现 Dr. Claw 有用，请引用：\n\n```bibtex\n@misc{song2026drclaw,\n  author       = {Dingjie Song and Hanrong Zhang and Dawei Liu and Yixin Liu and Zongxia Li and Zhengqing Yuan and Siqi Zhang and Lichao Sun},\n  title        = {Dr. Claw：从创意到论文的人工智能研究工作空间},\n  year         = {2026},\n  organization = {GitHub},\n  url          = {https:\u002F\u002Fgithub.com\u002FOpenLAIR\u002Fdr-claw},\n  homepage     = {https:\u002F\u002Fopenlair.github.io\u002Fdr-claw},\n}\n```\n\n## 致谢\n\n### 感谢构建\n- **[Claude Code](https:\u002F\u002Fdocs.anthropic.com\u002Fen\u002Fdocs\u002Fclaude-code)** — Anthropic 官方 CLI\n- **[Gemini CLI](https:\u002F\u002Fgeminicli.com\u002Fdocs\u002Fget-started\u002Finstallation\u002F)** — Google 的 Gemini 命令行代理\n- **[Codex](https:\u002F\u002Fdevelopers.openai.com\u002Fcodex)** — OpenAI Codex\n- **[React](https:\u002F\u002Freact.dev\u002F)** — 用户界面库\n- **[Vite](https:\u002F\u002Fvitejs.dev\u002F)** — 高速构建工具与开发服务器\n- **[Tailwind CSS](https:\u002F\u002Ftailwindcss.com\u002F)** — 以实用优先的 CSS 框架\n- **[CodeMirror](https:\u002F\u002Fcodemirror.net\u002F)** — 高级代码编辑器\n\n### 同时感谢\n- **[Claude Code UI](https:\u002F\u002Fgithub.com\u002Fsiteboon\u002Fclaudecodeui)** — Dr. Claw 以此为基础。详情请参阅 [NOTICE](NOTICE)。\n- **[AI Researcher](https:\u002F\u002Fgithub.com\u002FHKUDS\u002FAI-Researcher\u002F)** (HKUDS) — 研究工作流与代理式研究的灵感来源。\n- **[Vibe-Scholar](https:\u002F\u002Fgithub.com\u002FMr-Tieguigui\u002FVibe-Scholar)** — AI 原生研究工作空间方向的灵感来源。\n- **[autoresearch](https:\u002F\u002Fgithub.com\u002Fkarpathy\u002Fautoresearch)** — 自主研究编排与端到端执行的灵感来源。\n\n## 支持与社区\n\n### 关注最新动态\n- 为本仓库**加星标**以示支持\n- **关注**更新与新版本发布\n- **关注**项目公告以获取最新消息\n\n---\n\n\u003Cdiv align=\"center\">\n  \u003Cstrong>Dr. Claw — 从创意到论文。\u003C\u002Fstrong>\n\u003C\u002Fdiv>","# Dr. Claw 快速上手指南\n\n## 环境准备\n\n### 系统要求\n\n- 操作系统：支持 Linux、macOS 或 Windows（推荐使用 Linux\u002FmacOS）\n- Node.js v20 或更高版本（推荐使用 v22 LTS）\n\n### 前置依赖\n\n- 安装并配置以下至少一个 CLI 工具：\n  - [Claude Code CLI](https:\u002F\u002Fdocs.anthropic.com\u002Fen\u002Fdocs\u002Fclaude-code)\n  - [Gemini CLI](https:\u002F\u002Fgeminicli.com\u002Fdocs\u002Fget-started\u002Finstallation\u002F)\n  - [Codex CLI](https:\u002F\u002Fdevelopers.openai.com\u002Fcodex\u002Fcli\u002F)\n- 部分系统需要安装本地构建工具，如 `node-pty` 和 `better-sqlite3`。如果 `npm install` 失败，请参考 [FAQ](docs\u002Ffaq.md)。\n\n> 国内用户可考虑使用 [nvm](https:\u002F\u002Fgithub.com\u002Fnvm-sh\u002Fnvm) 管理 Node.js 版本，并通过国内镜像源加速安装过程。\n\n---\n\n## 安装步骤\n\n1. **克隆仓库**\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FOpenLAIR\u002Fdr-claw.git\ncd dr-claw\n```\n\n2. **安装依赖**\n```bash\nnpm install\n```\n\n3. **配置环境**\n```bash\ncp .env.example .env\n# 编辑 .env 文件，设置端口等参数\n```\n\n4. **启动应用**\n```bash\n# 开发模式（支持热重载）\nnpm run dev\n```\n\n然后通过浏览器访问 `http:\u002F\u002Flocalhost:5173` 创建账户。\n\n---\n\n## 基本使用\n\n### 方式 A：前端 UI\n\n打开浏览器访问 `http:\u002F\u002Flocalhost:5173`（或根据 `.env` 中配置的端口）即可开始使用。\n\n### 方式 B：终端操作\n\n1. 打开一个新的终端窗口（保持 `npm run dev` 在第一个终端中运行），并安装 `drclaw` CLI 工具：\n\n```bash\npip install -e .\u002Fagent-harness\n```\n\n2. 使用创建账户时的用户名和密码登录：\n\n```bash\ndrclaw auth login --username YOUR_USERNAME --password YOUR_PASSWORD\n```\n\n3. 安装至少一个代理 CLI（如尚未安装）：\n\n| Agent | 安装命令 | 认证方式 |\n|-------|----------|----------|\n| Claude Code | `npm install -g @anthropic-ai\u002Fclaude-code` | `claude` → 跟随 OAuth 提示 |\n| Gemini CLI | `npm install -g @google\u002Fgemini-cli` | `gemini` → Google 登录，或 `export GOOGLE_API_KEY=...` |\n| Codex CLI | `npm install -g @openai\u002Fcodex` | `codex login`，或 `export OPENAI_API_KEY=...` |\n\n4. 进入项目目录并启动代理：\n\n```bash\ncd \u002Fpath\u002Fto\u002Fyour\u002Fproject\nclaude    # 或 gemini | codex\n```\n\n> 技能文件会自动链接到每个项目的 `.claude\u002Fskills\u002F` 目录，无需额外配置。\n\n### 方式 C：OpenRouter 终端聊天（轻量级）\n\n使用 OpenRouter API 密钥进行终端交互，无需浏览器或 UI：\n\n```bash\n# 设置 OpenRouter API 密钥（或在命令中直接传入）\nexport OPENROUTER_API_KEY=sk-or-...\n\n# 启动聊天会话\nnode server\u002Fcli.js chat --model moonshotai\u002Fkimi-k2.5\n```\n\n你也可以直接传递密钥：\n\n```bash\nnode server\u002Fcli.js chat --model anthropic\u002Fclaude-sonnet-4 --key sk-or-your-key\n```\n\n> 可用模型列表请查看 [openrouter.ai\u002Fmodels](https:\u002F\u002Fopenrouter.ai\u002Fmodels)。","某高校人工智能实验室的研究团队正在开展一项关于自然语言处理（NLP）模型优化的课题，需要从文献调研、实验设计、代码实现到论文撰写完成全流程研究工作。\n\n### 没有 dr-claw 时\n- 团队成员需要手动搜索和整理大量学术论文，耗时且容易遗漏关键信息。\n- 实验流程需反复调整参数并记录结果，缺乏系统化的任务管理工具，导致进度混乱。\n- 编写代码和调试过程中遇到问题，需要频繁切换多个平台进行查询和测试，效率低下。\n- 论文撰写阶段依赖人工排版与格式校对，耗费大量时间且易出错。\n- 多人协作时，任务分配不明确，沟通成本高，难以统一研究方向。\n\n### 使用 dr-claw 后\n- 通过内置的“Research Lab”功能，自动抓取并整理相关领域的最新论文，生成结构化摘要和参考文献列表，节省了大量文献调研时间。\n- 利用“Auto Research”一键启动任务流，从实验设计到结果分析全程自动化执行，并实时跟踪进度，确保研究过程高效有序。\n- “Chat-Driven Pipeline”支持在聊天界面中描述研究目标，AI助手自动生成任务清单和代码框架，显著提升了开发效率。\n- 内置的“100+ Research Skills”库可自动完成论文初稿撰写、公式排版及格式检查，大幅减少人工校对工作量。\n- 多人协作时，dr-claw 提供统一的任务分配和进度追踪界面，便于团队成员协同推进项目。\n\n核心价值：dr-claw 将复杂的研究流程智能化、一体化，让研究人员专注于创新本身，而非重复性劳动。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FOpenLAIR_dr-claw_5147600a.png","OpenLAIR","Lehigh AI Research Lab (LAIR)","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002FOpenLAIR_17ea0e2b.png","",null,"https:\u002F\u002Fgithub.com\u002FOpenLAIR",[82,86,90,94,98,101,105,109,112],{"name":83,"color":84,"percentage":85},"JavaScript","#f1e05a",45.1,{"name":87,"color":88,"percentage":89},"TypeScript","#3178c6",26.7,{"name":91,"color":92,"percentage":93},"Python","#3572A5",13.4,{"name":95,"color":96,"percentage":97},"TeX","#3D6117",9.5,{"name":99,"color":79,"percentage":100},"BibTeX Style",3.4,{"name":102,"color":103,"percentage":104},"HTML","#e34c26",0.8,{"name":106,"color":107,"percentage":108},"CSS","#663399",0.5,{"name":110,"color":111,"percentage":108},"Shell","#89e051",{"name":113,"color":114,"percentage":115},"Makefile","#427819",0,793,82,"2026-04-05T11:09:49","NOASSERTION","Linux, macOS, Windows","未说明",{"notes":123,"python":121,"dependencies":124},"需要安装 Node.js v20 或更高版本，推荐使用 v22 LTS。需配置至少一个 CLI 工具（如 Claude Code、Gemini、Codex 或 OpenRouter）。部分系统可能需要安装原生构建工具（如 node-pty 和 better-sqlite3）。首次运行时需通过浏览器创建账户，并根据项目需求选择合适的模型和 API 密钥。",[125,126,127],"node.js v20+","npm","Claude Code CLI \u002F Gemini CLI \u002F Codex CLI \u002F OpenRouter API key",[15,13,26,54],"2026-03-27T02:49:30.150509","2026-04-06T06:45:33.228621",[132,137,142,147,152,156],{"id":133,"question_zh":134,"answer_zh":135,"source_url":136},5640,"如何解决 VibeLab 无法正确读取本地 Skill 文件的问题？","请下载最新版本的 VibeLab，并确保使用支持 Claude Code 的实验性自定义 API。如果在中国大陆使用，可以尝试通过 ccswitch 重定向 API 请求，但需注意在 VibeLab 的沙箱环境中可能仍需要登录。","https:\u002F\u002Fgithub.com\u002FOpenLAIR\u002Fdr-claw\u002Fissues\u002F2",{"id":138,"question_zh":139,"answer_zh":140,"source_url":141},5641,"如何解决 VibeLab 中技能名称中连字符与下划线不匹配的问题？","该问题已被标记为 P0 优先级并已修复。维护者已提交 PR #45 来改进 CLI 工具的跨平台检测能力，以确保技能名称中的连字符能被正确识别和调用。","https:\u002F\u002Fgithub.com\u002FOpenLAIR\u002Fdr-claw\u002Fissues\u002F32",{"id":143,"question_zh":144,"answer_zh":145,"source_url":146},5642,"如何解决 VibeLab 中 Thinking Mode 下拉菜单被遮挡且无滚动条的问题？","此问题已在最新版本中修复。现在 Thinking Mode 下拉菜单支持滚动功能，选项将不再被遮挡。请从 `main` 分支拉取最新版本进行测试。","https:\u002F\u002Fgithub.com\u002FOpenLAIR\u002Fdr-claw\u002Fissues\u002F8",{"id":148,"question_zh":149,"answer_zh":150,"source_url":151},5643,"如何解决 Claude 登录成功但设置界面仍显示未连接的问题？","此问题已修复。维护者建议设置 `allowDangerouslySkipPermissions: true` 在 SDK 配置中，以跳过首次启动时的交互式信任对话框，避免因无 TTY 导致的挂起问题。","https:\u002F\u002Fgithub.com\u002FOpenLAIR\u002Fdr-claw\u002Fissues\u002F1",{"id":153,"question_zh":154,"answer_zh":155,"source_url":141},5644,"如何解决 Python 技能缺少依赖项的问题？","当前设计决定不自动安装 Python 依赖项，因为这会增加项目环境管理的复杂性。建议手动安装所需依赖或在 `.venv` 环境中处理。",{"id":157,"question_zh":158,"answer_zh":159,"source_url":141},5645,"如何解决 Gemini CLI 未安装导致的静默失败问题？","此问题已在提交 `58cc2e0` 中修复。新增了 `cliResolution.js` 文件，用于实现跨平台的 CLI 可用性检测，并支持通过环境变量（如 `GEMINI_CLI_PATH`）覆盖默认路径。",[161,166,171,176,181,186],{"id":162,"version":163,"summary_zh":164,"released_at":165},105279,"v1.1.2","## Highlights\n\nDr. Claw v1.1.2 covers all changes merged after `v1.1.1` on March 30, 2026 through the `v1.1.2` release on April 5, 2026.\n\n- **Local GPU Detection**: Added full-stack local GPU detection, enabling the platform to discover and report available GPU resources on the host machine. This lays the groundwork for future local compute features.\n- **Multi-Provider API Key Support**: Fixed and enabled API key configuration for Claude Code, OpenAI, and Gemini, ensuring all three major providers work correctly out of the box.\n- **Windows Compatibility for News Search**: Resolved runtime compatibility issues that prevented the research news search scripts from running on Windows, broadening platform support.\n- **Codex Session Stability**: Fixed Codex session ID tracking and added metadata fallback logic to prevent session loss during edge-case failures.\n- **UI Improvements**: Added a sidebar version indicator for at-a-glance release awareness, fixed the stop button behavior, and improved editable numeric inputs in the news settings dialog.\n\n## Commit Summary\n\nSince `v1.1.1`, this release includes 17 commits total:\n- 14 non-merge commits\n- 3 merge commits\n\n## Full Commit Report\n\n### New Features\n- `9d8e3fc` feat: add local GPU detection (Zongxia Li)\n- `8637d18` feat: add local GPU detection backend implementation (Zongxia Li)\n- `c2ec149` Add sidebar version indicator (bbsngg)\n\n### Bug Fixes\n- `ff7781c` fix codex session id tracking (cnpcshangbo)\n- `127b377` test codex session metadata fallback (cnpcshangbo)\n- `9166a3b` patch Claude Code API key not working (Zongxia Li)\n- `ae81588` Open AI API key enabled (Zongxia Li)\n- `4ef90c7` Gemini AI API key enabled (Zongxia Li)\n- `b9e2bcb` stop button fix (Zongxia Li)\n- `5883ef0` Fix Windows news search runtime compatibility (r1ng13)\n- `f138a05` Fix editable numeric news settings inputs (r1ng13)\n\n### Documentation & Miscellaneous\n- `9827f2d` Update Cite (LiKid)\n- `6d498d1` Fix indentation for family-names entry in CITATION.cff (LiKid)\n- `71e76ef` mod: update WeChat group QR code (David Liu)\n\n### Merge Commits\n- `5ef1370` Merge pull request #121 from r1ng13\u002Ffix\u002Fwindows-news-runtime\n- `beba1ec` Merge pull request #118 from cnpcshangbo\u002Fpr\u002Fcodex-session-fallback\n- `57d2e31` Merge remote-tracking branch 'origin\u002Fmain' into merge-pr119\n\n## Contributors\n\n- **Zongxia Li** — GPU detection, API key fixes, stop button fix\n- **cnpcshangbo** — Codex session tracking and metadata fallback\n- **r1ng13** — Windows compatibility and news settings input fix\n- **LiKid** — Citation updates\n- **bbsngg** — Sidebar version indicator\n- **David Liu** — WeChat QR code update\n\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002FOpenLAIR\u002Fdr-claw\u002Fcompare\u002Fv1.1.1...v1.1.2","2026-04-05T05:09:32",{"id":167,"version":168,"summary_zh":169,"released_at":170},105280,"v1.1.1","## Highlights\n\nDr. Claw v1.1.1 covers all changes merged after `v1.1.0` on March 27, 2026 through the `v1.1.1` release commit on March 30, 2026.\n\n- Added OpenRouter as a full agentic provider with restored tool-use history, persisted permissions, and corrected Codex token accounting in mixed-provider workflows.\n- Added a session context sidebar and review queue to make active session state and follow-up review work more visible in Chat.\n- Added explicit chat thinking controls for Codex and Gemini, including Codex reasoning effort selection plus Gemini 3 `thinkingLevel` and Gemini 2.5 `thinkingBudget` presets.\n- Improved chat runtime feedback with a dedicated `Running code` status, release reminder follow-up handling, and package\u002Fbuild fixes for the `v1.1.1` line.\n\n## Commit Summary\n\nSince `v1.1.0`, this release includes 21 commits total:\n- 15 non-merge commits\n- 6 merge commits\n\n## Full Commit Report\n\n### Non-merge commits\n- `fd9dc34` Release 1.1.1\n- `16e8549` Add Gemini thinking strength controls\n- `974808e` Add Codex reasoning effort selector to chat\n- `4ba3f2e` Add session context sidebar and review queue\n- `b718856` 国内用户\n- `3c5d647` openrouter instruction\n- `e0ae59f` fix package build patch\n- `0210672` fix: restore OpenRouter history tool results\n- `27774cb` fix: apply persisted tool permissions and restore tool-use history for OpenRouter\n- `e5aefd9` fix: stop treating Codex lifetime tokens as context usage\n- `9abcf1f` feat: add OpenRouter as a full agentic provider\n- `0e2d094` fix: snooze release reminder on modal dismiss\n- `69c11d1` fix(chat): use statusTextOverride to decouple \"Running code\" from backend status\n- `0db5174` feat(chat): show \"Running code\" status when executing Bash commands\n- `ba4d47d` Add release update reminders and tests\n\n### Merge commits\n- `c1b1ad3` Merge pull request #117 from OpenLAIR\u002Ffeat\u002Fcodex-reasoning-effort-chat\n- `d03a60c` Merge pull request #116 from OpenLAIR\u002Ffeat\u002Fsession-context-sidebar\n- `bd5526f` Merge pull request #115 from OpenLAIR\u002Ffix\u002Fcodex-context-indicator-semantics\n- `d8d9da3` Merge pull request #114 from OpenLAIR\u002Fopenrouter\n- `bf118c4` Merge pull request #62 from OpenLAIR\u002Ffeat\u002Frelease-update-reminder\n- `e70918b` Merge pull request #82 from OpenLAIR\u002Ffeat\u002Fstatus-running-code\n\n## Validation\n\n- `npm run typecheck`\n- `npm run build`\n\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002FOpenLAIR\u002Fdr-claw\u002Fcompare\u002Fv1.1.0...v1.1.1","2026-03-30T04:31:47",{"id":172,"version":173,"summary_zh":174,"released_at":175},105281,"v1.1.0","## Highlights\n\nDr. Claw v1.1.0 covers all changes merged after v1.0.0 on March 17, 2026 through March 26, 2026, plus the March 27, 2026 release metadata update.\n\n- Added multi-session support, improved session naming, and hardened session recovery, timer consistency, index reconciliation, and stuck-loading handling.\n- Added the Dr. Claw CLI and OpenClaw integration workflow, improved cross-platform CLI detection, Claude auth browser opening, Windows support, and automatic port fallback for the server and Vite.\n- Expanded chat capabilities with better file and image attachments, Codex PDF handling, markdown file-link opening, implicit prompt loading, thinking block fixes, and stage tags with auto-linking.\n- Added project trash management, faster project loading, improved git onboarding and hidden-file visibility, terminal workspace persistence, and better shell history handoff.\n- Added citations and figure-generation improvements, a simplified local-only reference library without cloud Zotero, and new academic skills including inno-grant-proposal and inno-rebuttal.\n- Continued the Dr. Claw branding migration across the CLI, server paths, auth screen, docs, screenshots, and README content.\n\n## Commit Summary\n\nSince v1.0.0, this release includes 165 commits from contributors across product, CLI, chat, project management, skills, docs, and reliability work.\n\n## Notable PRs and topics\n\n- #65 multi-session support and improved session naming\n- #68 faster project loading and trash bin project management\n- #36 Dr. Claw CLI and OpenClaw integration workflow\n- #44 source control onboarding improvements\n- #49 citations support and figure generation improvements\n- #51 chat file preview and file-link opening\n- #60 thinking block integration and display fixes\n- #61 reference library simplification and cloud Zotero removal\n- #71 Gemini legacy thought formatting fixes\n- #72 session index reconciliation and stuck loading fixes\n- #75 session race handling and Python launcher fix\n- #76 implicit prompt loading with badge dropdown\n- #78 session stage tags and auto-linking\n- #80 shell history handoff and cleanup improvements\n- #81 session recovery with retry on errors\n- #84 Claude CLI detection fix on Windows\n- #92 inno-rebuttal skill and guided prompt additions\n- #95 Gemini and Codex chat image attachment fixes\n- #96 automatic port fallback for server and Vite\n- #97 Codex PDF attachment handling\n- #101 skill shortcut append behavior fix\n- #102 project path encoding deduplication and legacy migration\n- #110 automatic browser opening for Claude CLI auth URLs\n\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002FOpenLAIR\u002Fdr-claw\u002Fcompare\u002Fv1.0.0...v1.1.0\n","2026-03-27T05:38:25",{"id":177,"version":178,"summary_zh":179,"released_at":180},105282,"v1.0.0","# Changelog\n\n## Dr. Claw v1.0.0 - 2026-03-17\n\n### Highlights\n- Marked the first stable `1.0.0` release for Dr. Claw across desktop and mobile experiences.\n- Consolidated the recent major product additions including Codex support, Gemini workflows, Research Lab expansion, and skill discovery improvements.\n- Established `Dr. Claw v1.0.0` as the named release milestone for the current product line.\n\n### Commit Report\n#### Product, Dashboard, and UX\n- `76185b1` Add Paper News dashboard and integrate research paper skills\n- `089f979` Polish dashboard and news page layouts\n- `579ee00` fix(pipeline): add promotion workspace structure\n- `592d0bb` Add dashboard token usage summaries\n- `7a85fed` feat: add file preview support for images and PDFs. Extend CodeEditor beyond text-only viewing to support binary file preview using browser-native elements (img for images, iframe for PDF).\n- `8cbf253` fix: remove download button from unsupported file preview\n- `0322763` Add files via upload\n- `2b260ab` Replace Whisper-based voice input with browser-native Web Speech API\n- `ec6a10b` fix: resolve session processing race condition on session switch (#29)\n- `485d621` fix: strip thinking blocks from displayed assistant messages\n- `72e4295` fix: improve MicButton error messages and use portal for tooltip\n- `36f52cf` feat(compute): add SSH port field for Direct GPU nodes\n- `95de26f` add immediate inline port validation with visual feedback\n- `6640782` Fix default new workspace root path\n- `f9011ed` Clarify Opus Plan model label to avoid confusion\n- `777735a` Change default Claude model to Opus 4.6\n- `6c3a493` Update chat empty state and remove inno-research-orchestrator\n- `946257b` feat: more alert when agent not installed\n- `ff58b65` Fix chat placeholder showing \"Claude\" when using Gemini CLI\n- `c9e23d3` Fix chat provider ready prompt copy\n- `56c7b4c` Add file chat action to preview header\n- `f06e144` Fix remembered tool permissions for Gemini\n\n#### Projects, Workspace, and Session Management\n- `3a04fc8` feat: update server script to use watch path and enhance session handling logic\n- `47cad7f` Add: multi-users support\n- `6384d5e` feat: support project and session management\n- `b3a9a85` feat: project creation modal\n- `bfda97a` fix: resolve typescript errors in sidebar components\n- `b6cae22` fix: correctly await normalizeComparablePath in codex session retrieval\n- `e2c9fce` Fix workspace root filtering for claimed projects in sidebar\n- `a7f49df` fix: refresh projects after creation and handle more tab switches\n- `e7fdb74` Fix stale project records on delete\n- `123f87a` Add auto research workflow and email settings\n- `f31b7e0` feat: analysis workspace when import from outside\n- `eaaed7a` Fix stale auto research session recovery\n- `a4b58c2` feat: add QA session type\n- `6113465` feat: direct to QA session from file systems\n- `0f7d5e0` fix: gemini session message count\n- `f9683aa` fix: file tag for qa sessions\n\n#### Editor, File Navigation, and Research Workflows\n- `15324ac` chore(templates): align agent prompts with claude template\n- `6ef6546` feat(skills): add gemini-deep-research skill to skill library\n- `c1defb1` Rebuild news dashboard as unified single-page feed with all sources\n- `59326f7` Add server\u002Fdata\u002F to gitignore\n- `3e68622` Add i18n support (en\u002Fzh-CN\u002Fko) to news dashboard, replace source icons with brand logos, and improve UX\n- `d57e57e` Move research-news Python scripts from skills\u002F to server\u002Fscripts\u002F\n- `eb47eb5` Add python-tools installer, news data hook updates, and ignore result files\n- `147cd5e` Fix script paths in news source registry and rename action labels\n- `045af52` Refactor news dashboard to single-tab browsing, improve UX and filter invalid data\n- `f1ffbae` feat(editor): resolve bare filenames and show picker for ambiguous matches\n- `ade3e3a` fix(editor): follow symlinks, resolve partial paths, and improve picker UX\n- `1de8fdc` fix(chat): recognize extensionless files like Dockerfile as clickable links\n- `c00c865` Document auto research in README\n\n#### Branding, Packaging, Docs, and Compliance\n- `ead7e5c` Update logo image in README.md\n- `49fa5be` Update image source in README.md\n- `e963538` Adjust image height for VibeLab logo\n- `f8386c8` Adjust image height in README\n- `a540603` Rename project from VibeLab to Dr.Claw\n- `2938f67` Update README.md\n- `662cf74` Update project description in README.md\n- `c1d30e3` strengthen beta user agreement restrictions\n- `530c5a0` Stop storing telemetry dialogue content\n- `149442e` rebrand user-facing product naming to Dr. Claw\n- `98ad0d8` rename package cli and workspace defaults to dr-claw\n- `cec4bed` Update WeChat group QR code\n- `830107b` fix: update .gitignore to exclude __pycache__ and ensure *_results.json is ignored\n- `81fbbbb` add dr-claw compatibility migrations\n- `346b40a` migrate legacy dr-claw projects and internal identifiers\n- `1539ced` clean up remaining Dr. Claw branding references\n- `2e8764b` fix remaining skills taxonomy branding references\n- `ec08d7f` upda","2026-03-17T06:31:09",{"id":182,"version":183,"summary_zh":184,"released_at":185},105283,"v0.1.5","### Highlights\n- Added broader agent and workspace support with Codex integration, Gemini CLI support, survey workspaces, graph previews, and multi-shell execution.\n- Expanded core product workflows with project dashboards, guided chat starter refinements, account recovery, file upload\u002Fdelete flows, and auth\u002Fwebsocket fixes.\n- Reworked skill discovery and research UX with a redesigned skills explorer, taxonomy browsing, global skills library surfacing, and multiple dashboard\u002Fpreview polish passes.\n- Refreshed onboarding and documentation with README improvements, badge updates, and cleanup of obsolete guide content.\n\n### Full Commit List\n- fix(projects): fallback to home dir instead of root when project path decodes to \u002F (b85b2e8)\n- docs: remove TODO.md (5601e87)\n- merge: fix\u002Fproject-path-root-fallback into main (1ce1ac7)\n- feat(researchlab): add task insertion and deletion in pipeline board (806998f)\n- Add survey stage to research pipeline (f6b6b9e)\n- Fix auth flow and Claude login commands (8f47c35)\n- Remove unused guide files (c8613f6)\n- fix(websocket): include auth token for chat websocket (1dfbbd3)\n- add codex support (61724c0)\n- Merge pull request #16 from OpenLAIR\u002Ffeat\u002Fsurvey-stage-pipeline (73f26fa)\n- Merge pull request #17 from OpenLAIR\u002Fcodex\u002Fsupport (92832b0)\n- feat(survey): add survey workspace and graph previews (9f1aa42)\n- feat(workspace): add multi-shell workspace support (039d417)\n- Merge branch 'feat\u002Fsurvey-stage-pipeline' (8008311)\n- feat(models): update defaults and fix selector UI (3652fd5)\n- feat: add file management features including upload and delete functionality (76a215d)\n- Merge main (ae4ec26)\n- feat(dashboard): add project overview dashboard (e3f3285)\n- Streamline guided chat starter selection (b0bfda9)\n- Merge pull request #18 from OpenLAIR\u002Ffeat\u002Fproject-dashboard (1adb314)\n- Add account recovery registration flow (07700fa)\n- Merge pull request #19 from OpenLAIR\u002Ffeat\u002Fresearchlab-taskList-updates (ab2be50)\n- feat: add gemini cli support (607fb9a)\n- feat: gemini session support (d37cebd)\n- feat: gemini cli agent follow (e99fe13)\n- fix: refine gemini cli tool result presents (d70cf9f)\n- Redesign the skills dashboard as an  explorer (d1528f2)\n- mod: refine gemini cli interaction (0e200df)\n- fix: explicit type for item in toolConfigs.ts to satisfy CI typecheck (ca3da4e)\n- Merge pull request #20 from OpenLAIR\u002Ffeat\u002Fskills-explorer (39441ee)\n- Merge pull request #21 from OpenLAIR\u002Ffeat\u002Fgemini-cli (57279b8)\n- Update README.md (e55be2c)\n- Add WeChat QR badge to README (495a591)\n- docs: tighten README onboarding and FAQ (9baf2b6)\n- Improve research lab artifact explorer preview (4358f95)\n- Polish Research Lab layout and previews (4460838)\n- Add global skills library entry and refresh project dashboard (60913df)\n- Improve research lab and survey preview UX (fab7cba)\n- Adopt taxonomy-based skill explorer (8f41650)\n- chore(release): v0.1.5 (131c985)\n\n### Validation\n- `npm run typecheck` passed.\n- `npm run build` passed.\n","2026-03-09T07:08:34",{"id":187,"version":188,"summary_zh":189,"released_at":190},105284,"v2026.3.4","## What's Changed\n\n### Changes\n- Initial commit (cf576ae)\n- chore: ignore .pipeline, move e2e to test, trim docs (70c7d36)\n- docs(readme): unify Vibe Lab to VibeLab, merge acknowledgments, port 5173, dedupe trust\u002Fnetwork, style summary (286226f)\n- chore: replace bbsngg links with OpenLAIR\u002FVibeLab (71151ef)\n- feat: add core skill loading and skills index generation for agents (b8c5cb9)\n- feat: update workspace path handling and improve workspace name generation (389da8d)\n- 1 (4c4c060)\n- update chat components and git panel (d8cf418)\n- 1 (e673050)\n- feat: remove login\u002Fpassword wall, revert allowDangerouslySkipPermissions to false (63d5a3b)\n- feat: make soul templates stage-aware with pipeline.startStage support (9577738)\n- feat: refactor skills to VibeLab root ground reference + add edit\u002Fdelete to skill cards (3c3a182)\n- feat: add agent turn grouping and task progress improvements (f636b2a)\n- feat: add new skills and update templates (ac75dc9)\n- feat: add skill shortcuts panel and UI improvements (abc3fcd)\n- feat: update onboarding, settings, and message handling (21bf4f1)\n- feat: update skills dashboard, chat interface, and sidebar (324b230)\n- docs: comment out Cursor CLI and Codex references in READMEs (8a9e5f6)\n- feat(chat): make file paths in agent replies clickable to open in editor (1f4e8a4)\n- feat: align pipeline banner, skill shortcuts, and task progress in one container (9d1f494)\n- docs: add TODO.md with known bug for file preview panel resize (fdb6a58)\n- docs: add bypass permission bug to TODO.md (1327004)\n- docs: add UI tags fade-out bug to TODO.md (ca3e33c)\n- docs: organize TODO bugs by priority (all low for now) (bf536c2)\n- docs: add TODO for refactoring stage-skill-map.json to reduce skill suggestion noise (c389c43)\n- Merge pull request #7 from OpenLAIR\u002Ffeat\u002Ffile-panel-resize (8bcfa42)\n- Merge pull request #4 from OpenLAIR\u002Ffeat\u002Fstage-aware-soul-templates (70015d1)\n- feat: support generic file attachments and improve token tracking (f29f1d6)\n- Merge branch 'feat\u002Fskills-refactor-edit-delete' into main (f1c6e1c)\n- change template (d1f3834)\n- Merge pull request #9 from OpenLAIR\u002Ffet\u002Fguide (8175472)\n- feat: add presentation pipeline as 4th research stage (019e41c)\n- Removed Open research_brief.json, Open tasks.json. Filtered these two files out of the artifact list and blocked them   from being selected for preview; simplified placeholder text. (f0f5c3a)\n- feat(researchlab): add inline task editing in pipeline board (2259138)\n- feat(chat): add guided starter with skill-aware prompt templates (06bb34b)\n- feat(researchlab): improve task card edit discoverability and i18n (f137e9b)\n- merge: feat\u002Fresearchlab-taskList-updates into main (7eafef0)\n- refactor(pipeline): rename presentation stage to promotion (148ddb5)\n- Merge branch 'feat\u002Fpresentation-pipeline' into main (cae13b0)\n- feat(chat): add presentation guided starter scenario (7a2efa0)\n- chore(release): v2026.3.0 (3f1848b)\n- chore(release): v2026.3.4 (6085f17)\n\n### Breaking\n\n### Fixes\n- fix(templates): remove deprecated orchestrator and idea_maturity refs (8195ad4)\n- fix: update shell command handling and improve localization for terminal commands (38aa7f0)\n- fix: refine message handling for assistant role in chat and update shell component props (9dcc079)\n- fix: pass --trust flag to cursor-agent to resolve workspace trust error (9f8bb07)\n- fix: Claude auth status now queries CLI instead of missing credentials file (#1) (dd371dd)\n- fix: skip interactive trust dialog for headless SDK sessions (100ec4b)\n- fix: revert WORKSPACES_ROOT default to os.homedir() so users can create workspaces anywhere (c4315e1)\n- feat: add Import Local Skills button + port Zhengqing Yuan UI fixes (fccaa63)\n- fix: actually delete project folder from disk when deleting a project (744391f)\n- Merge pull request #3 from OpenLAIR\u002Ffix\u002Fcursor-trust-flag (8545cae)\n- feat: compute_node_guid_and_fix_ui (9816a19)\n- fix: guard process.env access in shared modelConstants for browser compatibility (554715e)\n- fix: presentation pipeline sanity fixes — artifact collection, task generation, templates (2a86275)\n- fix(researchlab): reset inline edit state on project switch (74e1f4d)\n\n### Full Changelog\n- 56 commits included in this release.\n","2026-03-04T18:02:37"]