[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-hotovo--aider-desk":3,"tool-hotovo--aider-desk":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":80,"owner_url":81,"languages":82,"stars":121,"forks":122,"last_commit_at":123,"license":124,"difficulty_score":23,"env_os":125,"env_gpu":125,"env_ram":125,"env_deps":126,"category_tags":128,"github_topics":129,"view_count":10,"oss_zip_url":79,"oss_zip_packed_at":79,"status":16,"created_at":137,"updated_at":138,"faqs":139,"releases":165},648,"hotovo\u002Faider-desk","aider-desk","AI coding agent for software engineers","AiderDesk 是一款专为软件工程师打造的桌面端 AI 编程平台，将自主智能代理、代码生成能力与综合工具集融为一体。它致力于解决开发者在面对复杂功能构建、大型代码库重构或重复性自动化任务时效率低下的痛点，让 AI 真正成为懂工作流的智能搭档。\n\n适合广大开发者使用，无论是初创项目的快速迭代还是企业级系统的维护。AiderDesk 的独特亮点在于其强大的自主代理模式，能够自动规划并执行复杂任务。技术上，它采用基于 LanceDB 的本地持久化记忆，能记住项目架构与个人偏好；支持与 VSCode、IntelliJ 等主流 IDE 无缝同步上下文；提供任务管理、Token 成本追踪以及 MCP 协议扩展能力。此外，用户还可配置专用子代理以优化特定场景成本。AiderDesk 旨在通过深度集成与智能化辅助，显著提升软件开发效率与体验。","\n# ✨ AiderDesk: The AI-Powered Development Platform\n\n[![Documentation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDocs-blue?logo=ReadTheDocs&logoColor=white)](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs)\n[![Discord](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDiscord-5865F2?logo=discord&logoColor=white)](https:\u002F\u002Fdiscord.com\u002Finvite\u002FdyM3G9nTe4)\n[![Ask DeepWiki](https:\u002F\u002Fdeepwiki.com\u002Fbadge.svg)](https:\u002F\u002Fdeepwiki.com\u002Fhotovo\u002Faider-desk)\n[![GitMCP](https:\u002F\u002Fimg.shields.io\u002Fendpoint?url=https:\u002F\u002Fgitmcp.io\u002Fbadge\u002Fhotovo\u002Faider-desk)](https:\u002F\u002Fgitmcp.io\u002Fhotovo\u002Faider-desk)\n\n**AiderDesk transforms how you write software** by combining an autonomous AI agent, powerful code generation, and a comprehensive toolset into a single desktop application. Whether you're building complex features, refactoring large codebases, or automating repetitive tasks, AiderDesk gives you an intelligent coding companion that works the way you do.\n\n## 🎬 Overview\n\n\u003Cdiv align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=KSWlhB-O2SE\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fhotovo_aider-desk_readme_d5a787e6da33.jpg\" alt=\"AiderDesk Overview Video\" width=400>\n  \u003C\u002Fa>\n\u003C\u002Fdiv>\n\n## ✨ Key Features\n\nAiderDesk is packed with features designed for modern software development:\n\n* **🤖 Autonomous Agent Mode**: An intelligent AI agent (powered by Vercel AI SDK) that autonomously plans and executes complex tasks—from implementing features to refactoring entire modules. Just describe what you need, and the agent breaks it down, explores your codebase, uses the right tools, and delivers results.\n* **🧩 Extensible Tool Ecosystem**:\n  * **Power Tools**: Direct file operations, semantic search, grep, shell commands, web fetching\n  * **Aider Tools**: Deep integration with Aider's intelligent code generation and modification\n  * **Memory Tools**: Vector-based persistent knowledge storage that learns project patterns and preferences\n  * **Skills**: Progressive disclosure system to add domain expertise and custom workflows\n  * **Task Tools**: Complete task management with cost tracking and todo lists\n  * **Todo Tools**: Built-in checklist management that agents can use to track progress\n* **👥 Specialized Subagents**: Create cost-optimized AI specialists for specific tasks (code review, testing, documentation). Automatic or on-demand invocation with configurable permissions and context memory.\n* **🎛️ Customizable Agent Profiles**: Configure different agent behaviors with system prompts, custom instructions, tool approvals, and rule files. Pre-built profiles included for common workflows.\n* **🧠 Persistent Memory**: The agent remembers project patterns, architectural decisions, and your preferences across tasks using local vector search (LanceDB).\n* **📂 Advanced Task Management**: Organize work with tasks that include chat history, context files, cost tracking, todos, and optional Git worktree isolation for safe experimentation.\n* **🌿 Git Worktrees**: Create isolated development environments for safe, parallel feature development. Multiple merge strategies with built-in revert support.\n* **🔌 Effortless IDE Integration**: Automatically sync context files with your active editor in:\n  * IntelliJ IDEA ([Plugin](https:\u002F\u002Fplugins.jetbrains.com\u002Fplugin\u002F26313-aiderdesk-connector) | [GitHub](https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk-connector-intellij-plugin))\n  * VSCode ([Extension](https:\u002F\u002Fmarketplace.visualstudio.com\u002Fitems?itemName=hotovo-sk.aider-desk-connector) | [GitHub](https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk-connector-vscode-extension))\n* **🧩 MCP Server Support**: Connect to any Model Context Protocol (MCP) server to add external capabilities (web browsing, database access, custom tools).\n* **⚙️ Hooks & Automation**: JavaScript hooks for workflow automation—respond to events, modify behavior, and integrate with external tools.\n* **📊 Usage Dashboard**: Visualize token usage, costs, and model distribution with interactive charts and tables.\n* **💰 Cost Tracking**: Monitor token usage and associated costs per task for both Aider and Agent.\n* **🌐 REST API**: Integrate AiderDesk with external tools and workflows.\n* **🌐 AiderDesk as MCP Server**: Expose AiderDesk's core functionality to other MCP-compatible clients (Claude Desktop, Cursor, etc.).\n\n[**Learn more about AiderDesk →**](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs)\n\n---\n\n### 🤖 Agent Mode: Autonomous AI Assistant\n\nAiderDesk's Agent Mode transforms the application into a powerful, autonomous coding companion. Instead of just generating code snippets, the agent can:\n\n- **Plan Multi-Step Work**: Break down complex requests like \"implement OAuth authentication\" into actionable steps\n- **Intelligently Explore**: Use semantic search and grep to understand your codebase structure\n- **Execute Safely**: Operate files, run commands, and fetch documentation with your approval\n- **Learn from Context**: Retrieve relevant memories about project patterns and your preferences\n- **Delegate to Specialists**: Automatically invoke specialized subagents for code review, testing, or documentation\n\n\u003Cdiv align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fyoutu.be\u002FLsd7QReXfy4\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fhotovo_aider-desk_readme_c384c6fdd5d8.jpg\" alt=\"Agent Mode Demo Video\" width=400>\n  \u003C\u002Fa>\n\u003C\u002Fdiv>\n\n#### Key Capabilities:\n\n- **Autonomous Planning**: The agent breaks down high-level requests into executable steps\n- **Tool-Driven**: Functionality is defined by connected tools (Power Tools, Aider, MCP, Skills, Memory, etc.)\n- **Seamless Integration**: Uses Aider for core coding tasks like generation and modification\n- **Multi-Provider Support**: Works with various LLM providers (OpenAI, Anthropic, Gemini, Bedrock, Deepseek, OpenAI-compatible)\n- **Transparent Operation**: See the agent's reasoning, plans, and tool usage in the chat\n- **Cost Optimization**: Use different models for different tasks—premium for complex work, cost-effective for routine operations\n\n#### Pre-Built Agent Profiles:\n\n- **Power Tools**: Direct file manipulation and codebase analysis\n- **Aider**: AI-powered code generation and refactoring\n- **Aider with Power Search**: Hybrid approach combining code generation with advanced search\n\n[**Learn more about Agent Mode →**](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Fagent-mode)\n\n---\n\n### 🌿 Git Worktrees: Isolated Development Environments\n\nAiderDesk's Git Worktrees feature provides professional-grade isolation for your development work, enabling safe experimentation and parallel development without affecting your main project branch.\n\n#### Key Benefits:\n\n- **🔒 Complete Isolation**: Work on features in separate environments without risking your main codebase\n- **🔄 Flexible Integration**: Choose between standard merge, squash & merge, or uncommitted changes only\n- **⚡ Parallel Development**: Work on multiple features simultaneously in isolated worktrees\n- **🛡️ Safety Features**: Built-in conflict detection, merge revert, and state preservation\n- **🤖 AI Integration**: AI assistance works seamlessly within isolated worktree environments\n\n#### Workflow Overview:\n\n1. **Create Worktree Task**: Start a new task in worktree mode for isolated development\n2. **Work Safely**: All file edits, commits, and AI interactions happen in the isolated environment\n3. **Integrate When Ready**: Choose your preferred merge strategy when work is complete\n4. **Revert if Needed**: Built-in revert functionality to undo merges if issues arise\n\n#### Merge Options:\n\n- **Standard Merge**: Preserves commit history with fast-forward merge\n- **Squash & Merge**: Creates a single clean commit with AI-generated message\n- **Uncommitted Only**: Transfers work-in-progress changes without merging commits\n\n[**Learn more about Git Worktrees →**](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Ffeatures\u002Fgit-worktrees)\n\n---\n\n### 🧠 Memory: AI That Learns Your Project\n\nAiderDesk's Memory system enables the agent to **store and retrieve durable, project-scoped knowledge** across tasks. This helps the agent remember:\n\n- **Your Preferences**: Formatting rules, frameworks you prefer, naming conventions\n- **Architectural Decisions**: \"We use Zod for validation\", \"REST client lives in src\u002Frenderer\u002Fsrc\u002Fapi\u002F\"\n- **Reusable Patterns**: Error handling conventions, logging standards, common abstractions\n\n#### How It Works:\n\n- **Local Vector Search**: Uses LanceDB with local embedding models—no data leaves your machine\n- **Project-Scoped**: Memories are stored per project, keeping knowledge organized\n- **Agent Integration**: Automatically retrieves relevant memories at task start and stores outcomes after completion\n- **Manual Management**: Browse, filter, and delete memories from Settings\n\n[**Learn more about Memory →**](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Ffeatures\u002Fmemory)\n\n---\n\n### 📋 Skills: Extend Agent with Expertise\n\nSkills let you package **reusable, on-demand expertise** that the agent can load when relevant. Instead of pasting large guides into chat, use skills with progressive disclosure:\n\n- **On Startup**: Only loads skill metadata (name + description)—token-efficient\n- **When Relevant**: Agent loads skill's `SKILL.md` and follows its instructions\n- **With Helpers**: Skills can include executable scripts and reference materials\n\n#### Use Skills For:\n\n- **Repeatable Workflows**: \"How we do releases\", \"how we write PR descriptions\"\n- **Domain-Specific Playbooks**: Internal processes, project conventions, brand guidelines\n- **Project-Specific Knowledge**: Keep `.aider-desk\u002Fskills\u002F` in your repo for team sharing\n- **Global Skills**: Personal skills in `~\u002F.aider-desk\u002Fskills\u002F` for all your projects\n\n[**Learn more about Skills →**](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Ffeatures\u002Fskills)\n\n---\n\n### 🎛️ Hooks: Automate Workflows\n\nHooks allow you to **extend AiderDesk's behavior** by executing custom JavaScript code in response to system events. Automate workflows, enforce rules, or integrate with external tools.\n\n#### Hook Events:\n\n- **Task Events**: `onTaskCreated`, `onTaskClosed`, `onPromptSubmitted`\n- **Agent Events**: `onAgentStarted`, `onAgentFinished`, `onAgentStepFinished`\n- **Tool Events**: `onToolCalled`, `onToolFinished`\n- **File Events**: `onFileAdded`, `onFileDropped`\n- **Approval Events**: `onQuestionAsked`, `onHandleApproval` (auto-answer questions)\n- **Modification Events**: Marked with (M)—return modified data to change behavior\n\n#### Locations:\n\n- **Global Hooks**: `~\u002F.aider-desk\u002Fhooks\u002F` (all projects)\n- **Project Hooks**: `.aider-desk\u002Fhooks\u002F` (project-specific)\n\n[**Learn more about Hooks →**](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Ffeatures\u002Fhooks)\n\n---\n\n### 📂 Task System: Organize Your Work\n\nAiderDesk's task management system provides a comprehensive way to organize work with:\n\n- **Complete State**: Chat history, context files, todos, costs, and metadata\n- **Working Modes**: Local development or isolated Git worktree\n- **Cost Tracking**: Per-task costs for both Aider and Agent usage\n- **Todo System**: Built-in checklists that agents can manage automatically\n- **Task Operations**: Create, rename, duplicate, delete, export (Markdown\u002FImage)\n\nTasks are stored per project in `.aider-desk\u002Ftasks\u002F` with full state persistence.\n\n[**Learn more about Task System →**](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Ffeatures\u002Ftasks)\n\n---\n\n### 📄 Comprehensive Context File Management\n\nKeep the AI focused on the relevant code with flexible context management options.\n\n\u003Cdiv align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fyoutu.be\u002F_hA1_NJDK3s\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fhotovo_aider-desk_readme_64582172f011.jpg\" alt=\"Context Files Demo Video\" width=400>\n  \u003C\u002Fa>\n\u003C\u002Fdiv>\n\n1. **Automatic IDE Sync**: Use the IntelliJ IDEA or VSCode plugins to automatically add\u002Fremove the currently active file(s) in your editor to\u002Ffrom the AiderDesk context.\n2. **Manual Control**: Utilize the \"Context Files\" sidebar in AiderDesk, which displays your project's file tree. Click files to manually add or remove them from the context, giving you precise control.\n\n[**Learn more about Context Files →**](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Ffeatures\u002Fide-integration)\n\n---\n\n### 💰 Cost Optimization\n\nAiderDesk provides multiple ways to optimize your AI spending:\n\n- **Per-Task Cost Tracking**: Separate tracking for Aider and Agent operations\n- **Usage Dashboard**: Visualize token usage, costs, and model distribution\n- **Subagent Cost Management**: Use different models per task type:\n  - **Claude Opus**: High-level planning and architecture\n  - **Claude 3.5 Sonnet \u002F GPT-4**: Complex development and debugging\n  - **Gemini Flash \u002F Claude Haiku**: Routine code reviews and testing\n- **Context Optimization**: Agent only loads what's needed (skills, memories)\n- **Tool Approval Control**: Set tools to \"Always\", \"Ask\", or \"Never\" to prevent unnecessary operations\n\n[**Learn more about Cost Optimization →**](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Ffeatures\u002Fusage-dashboard)\n\n---\n\n## 📥 Installation\n\n### Quick Start\n\n1. Download the latest release for your OS from [Releases](https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk\u002Freleases).\n2. Run the executable.\n\n### Recommended First Steps\n\n1. **Try Agent Mode**: Switch to Agent mode (`\u002Fagent`) and ask to agent to explore your codebase or implement a feature\n2. **Create a Skill**: Add a skill to `.aider-desk\u002Fskills\u002F` to encode your team's conventions\n3. **Set Up Memory**: Enable Memory tools in Agent Settings to let to agent learn your project patterns\n\n### Troubleshooting\n\n#### Disabling Auto Updates\n\nTo prevent automatic updates, set the `AIDER_DESK_NO_AUTO_UPDATE` environment variable:\n\n- **macOS\u002FLinux:** `export AIDER_DESK_NO_AUTO_UPDATE=true`\n- **Windows:** `$env:AIDER_DESK_NO_AUTO_UPDATE = \"true\"`\n\n### Custom Aider Version\n\nBy default, AiderDesk installs the latest version of the `aider-chat` Python package. If you need to use a specific version of Aider, you can set the `AIDER_DESK_AIDER_VERSION` environment variable.\n\nFor example, to use Aider version 0.83.1:\n\n```bash\n# macOS\u002FLinux\nexport AIDER_DESK_AIDER_VERSION=0.83.1\n\n# Windows (PowerShell)\n$env:AIDER_DESK_AIDER_VERSION = \"0.83.1\"\n```\n\nYou can also specify a git URL for installing a development version of Aider:\n\n```bash\n# macOS\u002FLinux\nexport AIDER_DESK_AIDER_VERSION=git+https:\u002F\u002Fgithub.com\u002Fuser\u002Faider.git@branch-name\n```\n\nThis variable will be used during the initial setup and when AiderDesk checks for updates. For more detailed information, please refer to [our docs](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Fadvanced\u002Fcustom-aider-version).\n\n## 👨‍💻 Development Setup\n\nIf you want to run from source, you can follow these steps:\n\n```bash\n# Clone the repository\n$ git clone https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk.git\n$ cd aider-desk\n\n# Install dependencies\n$ npm install\n\n# Run in development mode\n$ npm run dev\n\n# Build executables\n# For Windows\n$ npm run build:win\n\n# For macOS\n$ npm run build:mac\n\n# For Linux\n$ npm run build:linux\n```\n\n## 🤝 Contributing\n\nWe welcome contributions from the community! Here's how you can help improve aider-desk:\n\n1. **Fork the repository** on GitHub\n2. **Create a new branch** for your feature or bugfix:\n   ```bash\n   git checkout -b my-feature-branch\n   ```\n3. **Commit your changes** with clear, descriptive messages\n4. **Push your branch** to your fork\n5. **Create a Pull Request** against the main branch of the original repository\n\nPlease follow these guidelines:\n\n- Keep PRs focused on a single feature or bugfix\n- Update documentation when adding new features\n- Follow the existing code style and conventions\n- Write clear commit messages and PR descriptions\n\nFor major changes, please open an issue first to discuss what you would like to change.\n\n## ⭐ Star History\n\n[![Star History\nChart](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fhotovo_aider-desk_readme_ed5efa7c9978.png)](https:\u002F\u002Fstar-history.com\u002F#hotovo\u002Faider-desk&Date)\n\nThank you ❤️\n","# ✨ AiderDesk：AI 驱动的开发平台\n\n[![文档](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDocs-blue?logo=ReadTheDocs&logoColor=white)](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs)\n[![Discord](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDiscord-5865F2?logo=discord&logoColor=white)](https:\u002F\u002Fdiscord.com\u002Finvite\u002FdyM3G9nTe4)\n[![询问 DeepWiki](https:\u002F\u002Fdeepwiki.com\u002Fbadge.svg)](https:\u002F\u002Fdeepwiki.com\u002Fhotovo\u002Faider-desk)\n[![GitMCP](https:\u002F\u002Fimg.shields.io\u002Fendpoint?url=https:\u002F\u002Fgitmcp.io\u002Fbadge\u002Fhotovo\u002Faider-desk)](https:\u002F\u002Fgitmcp.io\u002Fhotovo\u002Faider-desk)\n\n**AiderDesk 改变了你的软件开发方式**，通过将自主 AI 代理 (Agent)、强大的代码生成能力和综合工具集整合到一个单一的桌面应用程序中。无论你是构建复杂功能、重构大型代码库，还是自动化重复任务，AiderDesk 都能为你提供一位智能的编程伴侣，以符合你工作方式的方式协同工作。\n\n## 🎬 概述\n\n\u003Cdiv align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=KSWlhB-O2SE\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fhotovo_aider-desk_readme_d5a787e6da33.jpg\" alt=\"AiderDesk 概览视频\" width=400>\n  \u003C\u002Fa>\n\u003C\u002Fdiv>\n\n## ✨ 主要功能\n\nAiderDesk 专为现代软件开发而打造，功能丰富：\n\n* **🤖 自主代理模式**：一个由 Vercel AI SDK 驱动的智能化 AI 代理 (Agent)，能够自主规划并执行复杂任务——从实现功能到重构整个模块。只需描述你的需求，代理就会将其分解，探索你的代码库，使用合适的工具，并交付结果。\n* **🧩 可扩展的工具生态系统**：\n  * **强力工具**：直接文件操作、语义搜索、grep、Shell 命令、网页获取\n  * **Aider 工具**：与 Aider 的智能代码生成和修改深度集成\n  * **记忆工具**：基于向量的持久化知识存储，学习项目模式和偏好\n  * **技能**：渐进式披露系统，用于添加领域专业知识和自定义工作流\n  * **任务工具**：完整任务管理，包含成本跟踪和待办事项列表\n  * **待办工具**：内置清单管理，代理可用于跟踪进度\n* **👥 专用子代理**：为特定任务（代码审查、测试、文档）创建成本优化的 AI 专家。支持自动或按需调用，具有可配置的权限和上下文记忆。\n* **🎛️ 可定制的代理配置文件**：使用系统提示词、自定义指令、工具审批和规则文件配置不同的代理行为。包含常见工作流的预构建配置文件。\n* **🧠 持久化记忆**：代理使用本地向量搜索 (LanceDB) 跨任务记住项目模式、架构决策和你的偏好。\n* **📂 高级任务管理**：组织工作，任务包含聊天历史、上下文文件、成本跟踪、待办事项，以及可选的 Git 工作树隔离以确保安全实验。\n* **🌿 Git 工作树**：创建隔离的开发环境，用于安全、并行的功能开发。支持多种合并策略及内置的回滚支持。\n* **🔌 无缝集成开发环境 (IDE) 集成**：自动将上下文文件与你的活动编辑器同步，支持：\n  * IntelliJ IDEA ([插件](https:\u002F\u002Fplugins.jetbrains.com\u002Fplugin\u002F26313-aiderdesk-connector) | [GitHub](https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk-connector-intellij-plugin))\n  * VSCode ([扩展](https:\u002F\u002Fmarketplace.visualstudio.com\u002Fitems?itemName=hotovo-sk.aider-desk-connector) | [GitHub](https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk-connector-vscode-extension))\n* **🧩 MCP 服务器支持**：连接到任何模型上下文协议 (MCP) 服务器，以添加外部功能（网页浏览、数据库访问、自定义工具）。\n* **⚙️ 钩子与自动化**：用于工作流自动化的 JavaScript 钩子——响应事件、修改行为并与外部工具集成。\n* **📊 使用仪表盘**：通过交互式图表和表格可视化 Token 使用情况、成本和模型分布。\n* **💰 成本跟踪**：监控 Aider 和代理每个任务的 Token 使用量及相关成本。\n* **🌐 REST API**：将 AiderDesk 与外部工具和流程集成。\n* **🌐 AiderDesk 作为 MCP 服务器**：向其他兼容 MCP 的客户端（Claude Desktop、Cursor 等）暴露 AiderDesk 的核心功能。\n\n[**了解更多关于 AiderDesk →**](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs)\n\n---\n\n### 🤖 代理模式：自主 AI 助手\n\nAiderDesk 的代理模式将应用程序转变为强大的自主编程伴侣。不仅仅是生成代码片段，代理还可以：\n\n- **规划多步工作**：将复杂的请求（如“实现 OAuth 认证”）分解为可操作的步骤\n- **智能探索**：使用语义搜索和 grep 理解你的代码库结构\n- **安全执行**：在获得你批准的情况下操作文件、运行命令和获取文档\n- **从上下文中学习**：检索有关项目模式和你偏好的相关记忆\n- **委托给专家**：自动调用专用子代理进行代码审查、测试或文档编写\n\n\u003Cdiv align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fyoutu.be\u002FLsd7QReXfy4\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fhotovo_aider-desk_readme_c384c6fdd5d8.jpg\" alt=\"代理模式演示视频\" width=400>\n  \u003C\u002Fa>\n\u003C\u002Fdiv>\n\n#### 关键能力：\n\n- **自主规划**：代理将高层请求分解为可执行的步骤\n- **工具驱动**：功能由连接的工具定义（强力工具、Aider、MCP、技能、记忆等）\n- **无缝集成**：使用 Aider 处理核心编码任务，如生成和修改\n- **多提供商支持**：适用于各种大语言模型 (LLM) 提供商（OpenAI、Anthropic、Gemini、Bedrock、Deepseek、OpenAI 兼容）\n- **透明操作**：在聊天中查看代理的推理、计划和工具使用情况\n- **成本优化**：针对不同任务使用不同模型——复杂工作使用高端模型，常规操作使用高性价比模型\n\n#### 预构建代理配置文件：\n\n- **强力工具**：直接文件操作和代码库分析\n- **Aider**：AI 驱动的代码生成和重构\n- **带强力搜索的 Aider**：结合代码生成与高级搜索的混合方法\n\n[**了解更多关于代理模式 →**](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Fagent-mode)\n\n### 🌿 Git Worktrees：隔离的开发环境\n\nAiderDesk 的 Git Worktrees（工作树）功能为您的开发工作提供专业级的隔离，使您能够在不影响主项目分支的情况下进行安全的实验和并行开发。\n\n#### 主要优势：\n\n- **🔒 完全隔离**：在独立环境中开发功能，无需担心影响主代码库\n- **🔄 灵活集成**：可选择标准合并、压缩并合并 (Squash & Merge)，或仅传输未提交的更改\n- **⚡ 并行开发**：在隔离的工作树中同时处理多个功能\n- **🛡️ 安全功能**：内置冲突检测、合并回退 (merge revert) 和状态保存\n- **🤖 AI 集成**：AI 辅助功能可在隔离的工作树环境中无缝工作\n\n#### 工作流程概览：\n\n1. **创建工作树任务**：以工作树模式启动新任务，进行隔离开发\n2. **安全工作**：所有文件编辑、提交 (commit) 和 AI 交互均在隔离环境中进行\n3. **准备就绪后集成**：工作完成后选择您首选的合并策略\n4. **必要时回退**：内置回退功能，若出现问题可撤销合并\n\n#### 合并选项：\n\n- **标准合并**：通过快进合并 (fast-forward merge) 保留提交历史\n- **压缩并合并**：生成单个干净的提交，附带 AI 生成的消息\n- **仅未提交更改**：传输进行中的更改，不合并提交记录\n\n[**了解更多关于 Git Worktrees →**](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Ffeatures\u002Fgit-worktrees)\n\n---\n\n### 🧠 Memory：学习您项目的 AI\n\nAiderDesk 的 Memory（记忆）系统使智能体 (Agent) 能够**跨任务存储和检索持久的、项目范围的知识**。这有助于智能体记住：\n\n- **您的偏好**：格式化规则、偏好的框架、命名约定\n- **架构决策**：“我们使用 Zod 进行验证”，\"REST 客户端位于 src\u002Frenderer\u002Fsrc\u002Fapi\u002F\"\n- **可复用模式**：错误处理约定、日志标准、常见抽象\n\n#### 工作原理：\n\n- **本地向量搜索 (Local Vector Search)**：使用 LanceDB 和本地嵌入模型 (embedding models)——数据不会离开您的机器\n- **项目范围**：记忆按项目存储，保持知识有序\n- **智能体集成**：在任务开始时自动检索相关记忆，并在完成后存储结果\n- **手动管理**：从设置中浏览、过滤和删除记忆\n\n[**了解更多关于 Memory →**](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Ffeatures\u002Fmemory)\n\n---\n\n### 📋 Skills：用专业知识扩展智能体\n\nSkills（技能）允许您打包**可复用的、按需的专业知识**，以便智能体在相关时加载。与其将大型指南粘贴到聊天中，不如使用具有渐进式披露功能的技能：\n\n- **启动时**：仅加载技能元数据（名称 + 描述）——节省 Token\n- **相关时**：智能体加载技能的 `SKILL.md` 并遵循其指令\n- **配合助手**：技能可以包含可执行脚本和参考资料\n\n#### 使用技能进行：\n\n- **可重复工作流**：“我们如何发布”，“我们如何编写 PR 描述”\n- **特定领域手册**：内部流程、项目规范、品牌指南\n- **项目特定知识**：将 `.aider-desk\u002Fskills\u002F` 保存在仓库中以供团队共享\n- **全局技能**：个人技能存储在 `~\u002F.aider-desk\u002Fskills\u002F` 以供所有项目使用\n\n[**了解更多关于 Skills →**](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Ffeatures\u002Fskills)\n\n---\n\n### 🎛️ Hooks：自动化工作流\n\nHooks（钩子）允许您通过在响应系统事件时执行自定义 JavaScript 代码来**扩展 AiderDesk 的行为**。自动化工作流、强制执行规则或与外部工具集成。\n\n#### 钩子事件：\n\n- **任务事件**：`onTaskCreated`, `onTaskClosed`, `onPromptSubmitted`\n- **智能体事件**：`onAgentStarted`, `onAgentFinished`, `onAgentStepFinished`\n- **工具事件**：`onToolCalled`, `onToolFinished`\n- **文件事件**：`onFileAdded`, `onFileDropped`\n- **审批事件**：`onQuestionAsked`, `onHandleApproval`（自动回答问题）\n- **修改事件**：标记为 (M)——返回修改后的数据以更改行为\n\n#### 位置：\n\n- **全局钩子**：`~\u002F.aider-desk\u002Fhooks\u002F`（所有项目）\n- **项目钩子**：`.aider-desk\u002Fhooks\u002F`（特定于项目）\n\n[**了解更多关于 Hooks →**](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Ffeatures\u002Fhooks)\n\n---\n\n### 📂 Task System：组织您的工作\n\nAiderDesk 的任务管理系统提供了一种全面的方式来组织工作，包括：\n\n- **完整状态**：聊天记录、上下文文件、待办事项、成本和元数据\n- **工作模式**：本地开发或隔离的 Git Worktree\n- **成本追踪**：分别追踪 Aider 和智能体使用的每任务成本\n- **待办系统**：内置检查清单，智能体可自动管理\n- **任务操作**：创建、重命名、复制、删除、导出（Markdown\u002F图片）\n\n任务按项目存储在 `.aider-desk\u002Ftasks\u002F` 中，具有完整的状态持久化功能。\n\n[**了解更多关于 Task System →**](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Ffeatures\u002Ftasks)\n\n---\n\n### 📄 全面的上下文文件管理\n\n通过灵活的上下文管理选项，让 AI 专注于相关代码。\n\n\u003Cdiv align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fyoutu.be\u002F_hA1_NJDK3s\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fhotovo_aider-desk_readme_64582172f011.jpg\" alt=\"Context Files Demo Video\" width=400>\n  \u003C\u002Fa>\n\u003C\u002Fdiv>\n\n1. **自动 IDE 同步**：使用 IntelliJ IDEA 或 VSCode 插件，自动将编辑器中当前活动的文件添加到\u002F从 AiderDesk 上下文中移除。\n2. **手动控制**：利用 AiderDesk 中的“上下文文件”侧边栏，显示您的项目文件树。点击文件可从上下文中手动添加或删除它们，为您提供精确的控制权。\n\n[**了解更多关于 Context Files →**](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Ffeatures\u002Fide-integration)\n\n---\n\n### 💰 成本优化\n\nAiderDesk 提供了多种优化 AI 支出的方式：\n\n- **按任务成本追踪**：分别追踪 Aider 和智能体操作的成本\n- **使用仪表盘**：可视化 Token 使用情况、成本和模型分布\n- **子智能体成本管理**：针对不同任务类型使用不同模型：\n  - **Claude Opus**：高层规划和架构\n  - **Claude 3.5 Sonnet \u002F GPT-4**：复杂开发和调试\n  - **Gemini Flash \u002F Claude Haiku**：常规代码审查和测试\n- **上下文优化**：智能体仅加载所需内容（技能、记忆）\n- **工具审批控制**：将工具设置为“始终”、“询问”或“从不”，以防止不必要的操作\n\n[**了解更多关于 Cost Optimization →**](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Ffeatures\u002Fusage-dashboard)\n\n---\n\n## 📥 安装\n\n### 快速开始\n\n1. 从 [Releases](https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk\u002Freleases) 下载适用于您操作系统的最新版本。\n2. 运行可执行文件。\n\n### 推荐的第一步\n\n1. **尝试代理模式 (Agent Mode)**：切换到代理模式（`\u002Fagent`）并让代理探索你的代码库或实现功能\n2. **创建技能 (Skill)**：将技能添加到 `.aider-desk\u002Fskills\u002F` 以编码团队的约定\n3. **设置记忆 (Memory)**：在代理设置中启用记忆工具 (Memory tools)，让代理学习你的项目模式\n\n### 故障排除\n\n#### 禁用自动更新\n\n为防止自动更新，请设置 `AIDER_DESK_NO_AUTO_UPDATE` 环境变量 (environment variable)：\n\n- **macOS\u002FLinux：** `export AIDER_DESK_NO_AUTO_UPDATE=true`\n- **Windows：** `$env:AIDER_DESK_NO_AUTO_UPDATE = \"true\"`\n\n### 自定义 Aider 版本\n\n默认情况下，AiderDesk 会安装 `aider-chat` Python 包 (Python package) 的最新版本。如果您需要使用特定版本的 Aider，可以设置 `AIDER_DESK_AIDER_VERSION` 环境变量。\n\n例如，要使用 Aider 版本 0.83.1：\n\n```bash\n# macOS\u002FLinux\nexport AIDER_DESK_AIDER_VERSION=0.83.1\n\n# Windows (PowerShell)\n$env:AIDER_DESK_AIDER_VERSION = \"0.83.1\"\n```\n\n您也可以指定一个 Git URL (git URL) 来安装 Aider 的开发版本：\n\n```bash\n# macOS\u002FLinux\nexport AIDER_DESK_AIDER_VERSION=git+https:\u002F\u002Fgithub.com\u002Fuser\u002Faider.git@branch-name\n```\n\n此变量将在初始设置期间以及 AiderDesk 检查更新时使用。有关更详细的信息，请参阅 [我们的文档](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Fadvanced\u002Fcustom-aider-version)。\n\n## 👨‍💻 开发环境设置\n\n如果您想从源代码运行，可以按照以下步骤操作：\n\n```bash\n# Clone the repository\n$ git clone https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk.git\n$ cd aider-desk\n\n# Install dependencies\n$ npm install\n\n# Run in development mode\n$ npm run dev\n\n# Build executables\n# For Windows\n$ npm run build:win\n\n# For macOS\n$ npm run build:mac\n\n# For Linux\n$ npm run build:linux\n```\n\n## 🤝 贡献\n\n我们欢迎社区的贡献！以下是您帮助改进 aider-desk 的方法：\n\n1. **在 GitHub 上 Fork 仓库 (repository)**\n2. **为您的功能或错误修复创建一个新分支 (branch)**：\n   ```bash\n   git checkout -b my-feature-branch\n   ```\n3. **提交您的更改**，使用清晰、描述性的消息\n4. **将您的分支推送到您的 Fork**\n5. **针对原始仓库的主分支 (main branch) 创建拉取请求 (Pull Request)**\n\n请遵循以下指南：\n\n- 保持 PR 专注于单个功能或错误修复\n- 添加新功能时更新文档\n- 遵循现有的代码风格和约定\n- 编写清晰的提交消息和 PR 描述\n\n对于重大更改，请先开启一个 Issue (Issue) 以讨论您想要更改的内容。\n\n## ⭐ 星标历史\n\n[![Star History\nChart](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fhotovo_aider-desk_readme_ed5efa7c9978.png)](https:\u002F\u002Fstar-history.com\u002F#hotovo\u002Faider-desk&Date)\n\n谢谢 ❤️","# AiderDesk 快速上手指南\n\nAiderDesk 是一款集成了自主 AI 代理、强大代码生成能力和综合工具集的桌面开发平台。它旨在成为你的智能编程伴侣，帮助你构建复杂功能、重构大型代码库或自动化重复任务。\n\n## 1. 环境准备\n\n在开始之前，请确保您的开发环境满足以下要求：\n\n*   **操作系统**：支持 Windows、macOS 或 Linux（具体版本请查阅官方文档）。\n*   **网络环境**：需要稳定的互联网连接以访问 LLM 服务和同步上下文。\n*   **代码编辑器**：推荐使用 **VSCode** 或 **IntelliJ IDEA**，以便利用深度集成功能。\n*   **模型支持**：支持多种 LLM 提供商（OpenAI, Anthropic, Gemini, Bedrock, Deepseek 等）。\n\n## 2. 安装步骤\n\nAiderDesk 提供独立的桌面应用程序及 IDE 插件两种入口。\n\n### 方式一：安装桌面客户端（推荐）\n1.  访问官方文档页面获取最新安装包：[https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs)\n2.  根据您的操作系统下载对应的桌面应用（Desktop Application）。\n3.  运行安装程序并按照提示完成初始化。\n\n### 方式二：安装 IDE 连接器\n如果您希望在现有编辑器中直接使用 AiderDesk 功能，可安装对应插件：\n*   **VSCode**: 安装扩展 `[hotovo-sk.aider-desk-connector](https:\u002F\u002Fmarketplace.visualstudio.com\u002Fitems?itemName=hotovo-sk.aider-desk-connector)`\n*   **IntelliJ IDEA**: 安装插件 `[AiderDesk Connector](https:\u002F\u002Fplugins.jetbrains.com\u002Fplugin\u002F26313-aiderdesk-connector)`\n\n## 3. 基本使用\n\n### 启动与配置\n1.  打开 AiderDesk 应用。\n2.  进入设置界面，配置您首选的大模型提供商（LLM Provider）。\n3.  输入 API Key 完成身份验证。\n\n### 创建任务\n1.  点击 **“新建任务”**（New Task）。\n2.  选择工作模式：\n    *   **本地开发**：直接在当前项目目录操作。\n    *   **Git Worktree**：创建隔离的开发环境，安全并行开发新功能。\n3.  在对话框中输入需求，例如：“实现 OAuth 认证”或“重构用户模块”。\n\n### 启用智能助手 (Agent Mode)\n1.  切换到 **Agent Mode**（代理模式）。\n2.  描述复杂任务，AI 代理将自动拆解步骤、探索代码库并使用相应工具。\n3.  在聊天窗口中查看代理的思考过程、计划及工具使用情况。\n4.  确认关键操作（如文件修改、Shell 命令执行）以获得安全保障。\n\n### 上下文管理\n*   **自动同步**：安装 IDE 插件后，当前编辑器中活跃的文件会自动添加到 AiderDesk 上下文中。\n*   **手动控制**：在 AiderDesk 侧边栏的“Context Files”中，您可以手动添加或删除文件树中的文件，精确控制 AI 关注的范围。\n\n### 成本与监控\n*   在仪表盘（Dashboard）中可视化查看 Token 使用情况、成本分布及模型分布。\n*   每个任务均包含独立的费用追踪，方便管理预算。\n\n---\n*更多高级功能（如记忆系统、技能扩展、Hooks 自动化）请查阅 [完整文档](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs)。*","后端工程师小张正在负责重构一个遗留的订单处理模块，同时需要集成新的第三方支付接口。\n\n### 没有 aider-desk 时\n- 需要手动在多个文件间切换搜索依赖关系，理解复杂的业务逻辑耗时费力，经常需要查阅大量文档\n- 编写单元测试和修改现有代码容易误伤旧功能，缺乏安全的沙箱环境验证，上线风险较高\n- 频繁在不同工具间复制粘贴上下文，导致开发流程割裂且容易遗漏关键信息，影响交付质量\n- 难以追踪每次 AI 辅助生成的代码 Token 消耗，团队预算控制不够清晰，无法评估投入产出比\n\n### 使用 aider-desk 后\n- 自主代理自动探索代码库并拆解任务，无需人工逐行查找依赖即可开始工作，大幅减少上下文切换时间\n- 利用 Git Worktrees 创建隔离环境进行实验，支持一键回滚确保主分支绝对安全，消除重构顾虑\n- IDE 插件自动同步当前编辑内容，直接在编辑器内完成代码生成与实时修改，保持流畅的开发体验\n- 内置持久化记忆记录项目架构偏好，Agent 能更精准地遵循既有编码规范，输出更符合团队标准的代码\n\naider-desk 通过整合自主代理、安全沙箱与持久记忆，让复杂重构任务变得可控且高效。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fhotovo_aider-desk_34dc3162.png","hotovo","Hotovo","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fhotovo_97543c45.png","",null,"https:\u002F\u002Fwww.hotovo com","https:\u002F\u002Fgithub.com\u002Fhotovo",[83,87,91,95,99,103,106,110,114,118],{"name":84,"color":85,"percentage":86},"TypeScript","#3178c6",90.7,{"name":88,"color":89,"percentage":90},"Handlebars","#f7931e",2.5,{"name":92,"color":93,"percentage":94},"JavaScript","#f1e05a",1.8,{"name":96,"color":97,"percentage":98},"Tree-sitter Query","#8ea64c",1.3,{"name":100,"color":101,"percentage":102},"SCSS","#c6538c",1.2,{"name":104,"color":105,"percentage":102},"Python","#3572A5",{"name":107,"color":108,"percentage":109},"CSS","#663399",0.8,{"name":111,"color":112,"percentage":113},"Shell","#89e051",0.3,{"name":115,"color":116,"percentage":117},"HTML","#e34c26",0.1,{"name":119,"color":120,"percentage":117},"Dockerfile","#384d54",1118,92,"2026-04-05T00:54:25","Apache-2.0","未说明",{"notes":127,"python":125,"dependencies":125},"该工具为桌面应用程序，支持本地向量搜索（使用 LanceDB），强调数据隐私（数据不出本地）。需安装 VSCode 或 IntelliJ IDEA 插件以实现代码上下文自动同步。支持通过 JavaScript 钩子脚本扩展工作流。兼容多种大语言模型提供商（如 OpenAI、Anthropic 等）。",[15,13,14],[130,131,132,133,134,135,136],"ai","ai-coding","ai-coding-tools","ai-developer-tools","ai-development","coding","electron-app","2026-03-27T02:49:30.150509","2026-04-06T07:16:06.280975",[140,145,150,155,160],{"id":141,"question_zh":142,"answer_zh":143,"source_url":144},2680,"WSL2\u002FLinux 环境下点击按钮打开文件管理器添加上下文文件时应用卡死怎么办？","这是特定 Ubuntu 版本下的已知问题，可能与 GTK 相关包版本冲突有关。建议将 Ubuntu 从 24.04.2 TLS 更新至 24.04.03 LTS。若仍无法解决，请确保安装了以下关键依赖包：gir1.2-gtk-4.0, libgtk-4-1, nautilus, pipewire 等。","https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk\u002Fissues\u002F316",{"id":146,"question_zh":147,"answer_zh":148,"source_url":149},2681,"遇到 \"summarizer unexpectedly failed\" 错误且没有生成提交信息该如何排查？","该错误通常与 LiteLLM 或模型摘要器有关。首先请在任务文件夹下的 `.aider.chat.history.md` 文件中查看详细异常堆栈。经验证，尝试切换模型（例如从 glm-4.6 切换到 ds v3.2）可显著降低该错误的发生频率。","https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk\u002Fissues\u002F347",{"id":151,"question_zh":152,"answer_zh":153,"source_url":154},2682,"如何开启 API 调用的详细调试日志以便排查连接错误？","此功能已在 v0.25.0 版本中修复并支持。升级后可通过配置标志开启更详细的日志输出，从而能够追踪是 API 端、Prompt 还是上下文导致的错误，而不再仅显示 UI 报错。","https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk\u002Fissues\u002F182",{"id":156,"question_zh":157,"answer_zh":158,"source_url":159},2683,"如何在代码合并到主分支前审查 Agent 的修改内容？","该功能已在 v0.56.0 版本中实现。现在支持在工作区（worktree）模式下查看所有更新文件（包括已提交的），并提供类似 GitHub PR 的并排差异视图及评论功能，方便人工审核后再合并。","https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk\u002Fissues\u002F679",{"id":161,"question_zh":162,"answer_zh":163,"source_url":164},2684,"应用启动后卡在 \"Starting\" 且日志显示 PID 清理错误怎么办？","日志显示 `Error cleaning up old PID file` 且 `taskkill` 返回 code 128。这通常意味着存在残留的旧进程或 PID 文件。建议手动检查任务管理器结束相关进程，或删除项目目录下的残留 PID 文件后重试。","https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk\u002Fissues\u002F273",[166,171,176,181,186,191,196,201,206,211,216,221,226,231,236,241,246,251,256,261],{"id":167,"version":168,"summary_zh":169,"released_at":170},102178,"v0.54.0","## Highlights\r\n\r\n### 📥 Message Queuing: Stay in the Flow\r\nDon't let a running agent slow down your thought process. You can now **queue messages** while a task is being processed. \r\n- **Plan Ahead**: Prepare your next set of instructions or corrections while the agent is busy implementing the current ones.\r\n- **Smart Execution**: Once the agent finishes, it will automatically pick up and execute the next message in your queue.\r\n- **Immediate Correction**: If you spot an error, you can interrupt the agent and have it immediately start working on your queued correction.\r\n\r\n### ♊ Gemini CLI: Direct Subscription Support\r\nWe've added direct support for **Gemini CLI** as a new provider. A big thank you to **@jutaz** for contributing this integration. You can now utilize your official Gemini subscription directly within AiderDesk, giving you even more flexibility in choosing your preferred AI powerhouse.\r\n\r\n### 🪝 Enhanced Hook System & Event Modification\r\nWe’ve increased the power of our **Hook System**. We have updated some events, so you can now not only listen to them but also **modify their data in real-time**. Change tool results on the fly, append an additional message after an agent finishes to keep the momentum going. We've included several new examples in our [documentation](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Ffeatures\u002Fhooks#example-hooks) to help you get started.\r\n*Stay tuned - we are working on a revamped Extension system coming next week that will have much better control and take this customization to a whole new level!*\r\n\r\n### 🧪 BMAD Workflow Polishing\r\nOur **BMAD Method mode** is moving closer to leaving its preview phase with several targeted improvements:\r\n-   **Context Awareness**: Agents are now better aware of their current workflow stage and will suggest more intelligent \"next steps.\"\r\n-   **Story-Driven Naming**: Workflows like `create-story`, `dev-story`, and `code-review` now automatically name your tasks based on the user story being handled and prepares the context so agent does not have to iterate to find out what to do next.\r\n-   **Model Flexibility**: You can now select specific models for each BMAD workflow, allowing you to use more powerful models for planning and cost-efficient ones for implementation.\r\n\r\n### ⚙️ Improved Aider Connector Robustness\r\nWe've resolved several issues regarding the background management of the Aider process. The **Aider connector** now properly starts when needed - whether you're executing Aider-specific slash commands or when an agent utilizes Aider tools in its workflow.\r\n\r\n## Other Changes\r\n\r\n-   **Updated Files UI**: Added a convenient \"View Changes\" button directly on updated context files for faster reviews.\r\n-   **Gemini Fixes**: Ensured user messages are always included in requests to Gemini models to avoid signature errors.\r\n-   **Non-Git Support**: AiderDesk now correctly skips `git add` operations when working in a project that isn't a Git repository.\r\n-   **UI Refinements**: Moved message navigation controls outside the scrollable container for better accessibility and added missing tooltips to icons in the prompt field.\r\n-   **Subagent Efficiency**: Subagents now always use the \"Compact\" context compaction type by default to save tokens and improve performance.\r\n\r\n💡 **We’d love to hear your thoughts!** Your feedback helps us improve AiderDesk - feel free to share any suggestions or issues. Happy coding! 🚀","2026-02-23T06:15:28",{"id":172,"version":173,"summary_zh":174,"released_at":175},102179,"v0.53.0","## Highlights\r\n\r\n### 📝 Updated Files & Inline Code Requests\r\nWe've introduced a new **Updated Files** section that provides a centralized view of all files modified during your current task. This feature is more than just a changelog—it’s a powerful review and refinement tool.\r\n-   **Visual Diff Access**: Click any file to open the diff view with your preferred visualization mode (Side-by-Side, Unified, or Compact).\r\n-   **Inline Code Requests**: The standout feature is the ability to click on any specific line and request a change. AiderDesk creates a focused task to implement your modification with full context, making code review and refinement incredibly precise.\r\n*Learn more about this workflow in our [Updated Files documentation](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Ffeatures\u002Fupdated-files).*\r\n\r\n### 📊 Mermaid Diagrams in Messages\r\nAiderDesk now natively renders **Mermaid diagrams** within chat messages. Whether your agent is generating flowcharts, sequence diagrams, or architectural maps to explain its reasoning, they will now display beautifully and interactively right in the conversation thread.\r\n\r\n### 🧪 Enhanced BMAD Mode & Workflows\r\nWe've polished the BMAD Method mode with several improvements based on early feedback:\r\n-   **Brainstorming Workflow**: Added a new Brainstorming workflow option to the BMAD toolkit for better ideation support.\r\n-   **Quick Flow Detection**: Improved state detection for Quick Flow workflows to ensure smoother transitions.\r\n-   **Browser Reliability**: Consecutive prompts in BMAD mode now work correctly when running via the browser.\r\n-   **Library Installation**: Updated the installation process to use a command-based approach for better reliability.\r\n\r\n### 📱 Mobile Layout Optimizations\r\nThe mobile experience continues to improve with specific optimizations for complex workflows. In **BMAD mode**, the agent selector now stacks cleanly under the mode selector, and action buttons are absolutely positioned at the top right. We've also fixed the project dropdown to truncate long names with ellipses instead of pushing icons off-screen.\r\n\r\n## Other Changes\r\n\r\n-   **Quick Task Access**: Added a \"Show task info\" button for instant access to task metadata.\r\n-   **Context Management**: Image messages are now sent even when \"Include Context Files\" is disabled.\r\n-   **Performance**: Optimized MCP server initialization and Aider connector start-up management for faster load times.\r\n-   **Bug Fixes**:\r\n    -   Increased timeout for the semantic search tool to prevent errors during parallel execution.\r\n    -   Properly synced task `baseDir` with project `baseDir` to handle migrated tasks.\r\n    -   Fixed display of project files with special characters.\r\n    -   Correctly reading BMAD workflow files when preparing context.\r\n\r\n💡 **We’d love to hear your thoughts!** Your feedback helps us improve AiderDesk - feel free to share any suggestions or issues. Happy coding! 🚀","2026-02-16T06:09:48",{"id":177,"version":178,"summary_zh":179,"released_at":180},102180,"v0.52.0","## Highlights\r\n\r\n### 🧪 Introducing BMAD Method Mode (Preview)\r\nWe are thrilled to introduce a brand-new working mode based on the **BMAD Method** (Breakthrough Method of Agile AI Driven Development). BMAD is an AI-driven framework designed to guide you through the entire software lifecycle - from initial ideation and brainstorming to solutioning and agentic implementation. \r\n- **Guided Workflows**: Leverage specialized agents that follow structured phases like Adversarial Review and Advanced Elicitation to ensure high-quality code and robust planning.\r\n- **Versatile Application**: Whether you are starting a project from scratch or implementing a small feature in an established codebase, BMAD helps prevent agent conflicts and provides a clear path forward.\r\n- **Preview Release**: Please note that this is a **Preview mode** with some rough edges. We are actively refining the experience, and your feedback is incredibly valuable to us as we continue to improve it.\r\nLearn more about the methodology at [docs.bmad-method.org](https:\u002F\u002Fdocs.bmad-method.org).\r\n\r\n### 📱 New Compact Message View\r\nFor those who prefer a more streamlined workspace, we’ve added a new **Compact View** for messages. This mode optimizes the chat interface by combining assistant responses with tool execution blocks into a more cohesive and dense layout. It’s perfect for power users who want to see more of the conversation at once without sacrificing detail.\r\n*You can enable this in **Settings -> General -> Message Display**.*\r\n\r\n### 🌐 New Providers: Kimi Coding Plan & Anthropic Compatibility\r\nWe are expanding our model ecosystem with direct support for the **Kimi Coding Plan** provider, giving you access to the powerful new **Kimi K2.5** model. Additionally, we’ve added a generic **Anthropic Compatible** provider. If you are using a third-party service or a local proxy that utilizes the Anthropic API format, AiderDesk now has you covered.\r\n\r\n## Other Changes\r\n\r\n- **Improved Context Management**: Decreased the default max tokens for semantic search to ensure better context window utilization.\r\n- **Context File Details**: Added more detailed information about files currently in the context for better transparency.\r\n- **Rule Handling**: Fixed the handling of the `rules\u002F` directory when using Aider modes in combination with Git Worktrees.\r\n- **Message Pruning**: The \"Remove messages up to here\" action now properly cleans up associated tool calls from assistant messages.\r\n- **Default Profiles**: Updated default agent profiles to not include context files by default for a cleaner start.\r\n- **UX Refinements**: Improved the initial user message experience and moved mode tooltips to the right alignment for better visibility.\r\n- **Bug Fixes**: Resolved an issue with Merge actions from worktrees and ensured files are cleared properly on new tasks when the IDE plugin is disconnected.\r\n- **Performance**: Optimized how agents handle files within their context for faster response times.\r\n\r\n💡 **We’d love to hear your thoughts!** Especially on the new BMAD mode - your feedback helps us improve AiderDesk. Happy coding! 🚀","2026-02-09T06:06:16",{"id":182,"version":183,"summary_zh":184,"released_at":185},102181,"v0.51.0","## Highlights\r\n\r\n### 🧠 Advanced Context Management: Auto-Handoff vs. Auto-Compact\r\nWe’ve upgraded how AiderDesk handles long conversations! In your task settings, you can now choose your preferred \"Compaction Type.\" \r\nPreviously, auto-compaction would summarize the thread within the same task. Now, you can opt for **Auto-Handoff**. When your conversation hits the token threshold, AiderDesk will automatically extract the essential context and move the conversation into a new subtask. This keeps your current context perfectly clean while preserving the original conversation history for reference.\r\n\r\n### ✅ TODO Item Handoff: Modularize Your Workflow\r\nSometimes a single TODO item turns out to be a task of its own. You can now **handoff a specific TODO item to a new task** directly from the TODO window. This is perfect for isolating work that can be done in parallel, like generating unit tests or implementing a sub-component, without cluttering your main task's context.\r\n\r\n### 🌐 Browser & Headless Enhancements: Notifications & URL Navigation\r\nFor those running AiderDesk via Docker or as a headless server, we’ve brought the browser experience closer to the desktop app:\r\n*   **Image Support**: You can now paste images directly from your clipboard into the browser prompt.\r\n*   **Push Notifications**: Stay informed on task completion with native browser notifications.\r\n*   **Direct Bookmarking**: We’ve added support for `project` and `task` URL parameters. You can now bookmark specific tasks or share direct links to them for instant navigation.\r\n\r\n### ⚔️ The `\u002Fresolve-conflicts` Command\r\nThe AI-powered conflict resolution that was previously exclusive to our Worktree rebase flow is now available as a standalone command! Simply type `\u002Fresolve-conflicts` in any project to let the agent analyze and resolve merge conflicts for you. It’s like having a senior dev handle the messy parts of a rebase.\r\n\r\n### ✂️ \"Delete Up To Here\": Precise Course Correction\r\nBuilding on our \"Fork\" and \"Delete Message\" features, we’ve added **Delete up to here**. If you notice an agent drifting off-course, you can select any message and delete everything following it. This allows you to \"re-steer\" the agent from a known good point without the overhead of creating an entirely new forked task.\r\n\r\n## Other Changes\r\n\r\n-   **Performance**: Improved UI responsiveness by replacing the underlying tooltip rendering library - big Thanks to @jutaz.\r\n-   **UI Refinements**: Updated the visuals of model selectors and moved project tabs into a dropdown menu on mobile devices for better space management.\r\n-   **Subtasks of Archived Task**: Archiving a parent task now automatically hides its subtasks to keep your sidebar clean.\r\n-   **Context Tree**: Updated how project files not currently in the context are are showing up in the tree, and removed empty tooltips from the folders.\r\n-   **Tool Improvements**: Glob and Grep tool messages now provide more detailed information to the user.\r\n-   **Subtask Inheritance**: New subtasks now correctly inherit the agent profile, provider, and model from their parent.\r\n-   **Claude SDK**: The Claude Agent SDK provider is now only visible if the `claude` executable is detected on your system.\r\n-   **Reliability**: Prompt text is now properly preserved when switching between tasks, and agent profiles are correctly updated after deletion.\r\n-   **Diff Viewer**: Added support for unknown file types, allowing them to be viewed without syntax highlighting rather than failing.\r\n\r\n💡 **We’d love to hear your thoughts!** Your feedback helps us improve AiderDesk - feel free to share any suggestions or issues. Happy coding! 🚀","2026-02-02T06:30:10",{"id":187,"version":188,"summary_zh":189,"released_at":190},102182,"v0.50.1","## Hotfix\r\n\r\n- fixed \"Claude Code process exited with code 1: undefined\" issue with Claude Agent SDK provider\r\n- fixed occasional TypeError issue when receiving the response in Agent mode\r\n\r\nComplete release notes for version 0.50.0 can be found [here](https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk\u002Freleases\u002Ftag\u002Fv0.50.0).","2026-01-27T14:00:10",{"id":192,"version":193,"summary_zh":194,"released_at":195},102183,"v0.50.0","## Highlights\r\n\r\n### 🎭 Claude Agent SDK: Use Your Claude Subscription Directly!\r\nWe’ve made it happen: you can now use your **Claude subscription** within AiderDesk! By integrating the Claude Agent SDK in the officially supported way, we’ve opened a path for you to leverage your existing subscription while enjoying the AiderDesk UI and toolset. \r\n*   **How it works**: This integration utilizes the official SDK, meaning it requires **Claude Code** to be installed on your system.\r\n*   **Agent Mode Exclusive**: Due to the nature of the SDK, this provider is currently available exclusively in **Agent mode**.\r\n*   **Community Shout-out**: This was made possible thanks to the fantastic work of **Ben Vargas** and his [ai-sdk-provider-claude-code](https:\u002F\u002Fgithub.com\u002Fben-vargas\u002Fai-sdk-provider-claude-code) library. If you find this useful, please head over and give his repository a star!\r\n\r\nFor setup details, check out our [Provider Documentation](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Fconfiguration\u002Fproviders#claude-agent-sdk).\r\n\r\n### 🚀 Peak Performance & \"Less Experimental\" Rendering\r\nTogether with @pcfreak30, we’ve performed a deep refactor of the UI core to make AiderDesk feel snappier than ever. If you’re a power user who frequently switches between dozens of tasks or manages conversations with hundreds of messages, you’ll notice an immediate difference in responsiveness. \r\nFurthermore, our **Virtualized Message Rendering** (found in Settings) has received significant updates. It now handles auto-scrolling much more consistently and has officially graduated to being **\"less experimental.\"** If you want the absolute smoothest experience in long threads, now is the perfect time to toggle it on.\r\n\r\n### 🍴 Task Forking: The Ultimate \"What If?\" Tool\r\nEver noticed your agent heading down a rabbit hole? Now you can stop it in its tracks and **Fork a task from any message**. This creates a new branch of your conversation from that specific point, allowing you to:\r\n*   Steer the agent in a different direction if it makes a mistake.\r\n*   Experiment with different models from the same starting point.\r\n*   Preserve a \"known good\" state while trying out risky implementations.\r\n\r\n### 🖥️ Browser-Based Terminals\r\nFor those using AiderDesk via the browser client, we are excited to bring **full terminal support** to your web sessions. Whether you’re checking logs from a tablet or running a quick build command from your phone, you now have a real terminal at your fingertips wherever you are.\r\n\r\n## Other Changes\r\n\r\n-   **Visual Progress**: Added a visual indicator to project tabs so you can see at a glance which project is currently processing.\r\n-   **Optimistic UI**: User messages now appear instantly upon submission for a more fluid chat experience.\r\n-   **Smart Notifications**: System notifications now include the specific task name and state, providing better context when you're tabbed away.\r\n-   **Z.AI Control**: Added an option to disable \"thinking\" blocks for Z.AI Coding Plan models. Thanks to @aguyinatx for the initial work on this.\r\n-   **Mobile Refinements**: On mobile devices, the \"Enter\" key now adds a new line by default instead of submitting the prompt, preventing accidental sends.\r\n-   **Message Search**: The CTRL+F message search now correctly scopes results to the active project only.\r\n-   **Cost Tracking**: Fixed the the cost reporting of Aider.\r\n-   **Handoff Reliability**: Fixed an issue to ensure handoff task context is properly stored before execution.\r\n\r\n💡 **We’d love to hear your thoughts!** Your feedback helps us improve AiderDesk - feel free to share any suggestions or issues. Happy coding! 🚀","2026-01-26T06:27:57",{"id":197,"version":198,"summary_zh":199,"released_at":200},102184,"v0.49.0","## Highlights\r\n\r\n### 🌳 Hierarchical Task Management: Introducing Subtasks\r\nTake your project organization to the next level with **Subtasks**! You can now break down complex goals into a structured tree of parent and child tasks. \r\n-   **Worktree Inheritance**: Subtasks automatically inherit the Git worktree of their parent, ensuring all related work stays in the same isolated environment.\r\n-   **Agent Integration**: We've updated the task tools so agents can now programmatically create and manage subtasks, allowing them to autonomously break down large features into manageable units.\r\n\r\n### 🎯 Focused Handoff: From Discussion to Action\r\nIntroducing the `\u002Fhandoff` command—a smarter way to transition between project phases. Unlike simple compaction, **Handoff** intelligently extracts only the relevant context, decisions, and progress from your current conversation and moves them into a brand-new, focused task. It’s perfect for shifting focus from \"planning\" to \"implementation\" without carrying over the noise of the entire brainstorm.\r\n\r\n### 🔍 Semantic Task Search & Advanced Task Tools\r\nFinding information across your project history is now faster and more intuitive. We've added new tools that allow agents to perform **semantic searches** within the current task or its parent. This enables agents to \"remember\" architectural decisions or implementation patterns discussed previously, ensuring consistency as you move from high-level planning to subtask implementation.\r\n\r\n### 🧹 Full Context Control: Message Removal\r\nYou now have absolute control over your conversation history. We've added the ability to **remove any individual message** from a task. This is a powerful tool for pruning irrelevant information, correcting the agent's \"memory\" after a wrong turn, or simply keeping the context window clean and focused on the current problem.\r\n\r\n### 📱 Mobile-First Responsiveness & Performance\r\nAiderDesk is now much more capable on the go! We've overhauled the layout with a **mobile-first approach**:\r\n-   **Responsive Layouts**: The task sidebar, prompt field, and model selectors now adjust fluidly to small screens.\r\n-   **Mobile Sidebar**: Enjoy a full-screen task sidebar on mobile devices with a dedicated toggle button and proper scroll support.\r\n-   **Rendering Optimization**: We've improved general rendering performance, making the UI feel snappier during long conversations.\r\n\r\n### ⚙️ Updated Default Max Output Tokens Behavior\r\nWe've changed the default behavior for model configurations: **Max output tokens are now undefined by default**. \r\nPreviously, some models (like those from Z.AI) defaulted to a very high max token limit which could prematurely \"choke\" the context window in Agent mode. \r\n*Note: This change applies to new models. If you have existing models with high token limits that you'd like to clear, you should manually reset them in the Model Library.*\r\n\r\n## Other Changes\r\n\r\n-   **LiteLLM Improvements**: Usage data is now correctly included when streaming with LiteLLM models. Thanks to @jutaz for this contribution!\r\n-   **Select All**: Added a \"Select all\" action to the menu. Another thank you to @jutaz!\r\n-   **MDX Support**: Changes made in `.mdx` files are now properly rendered in the diff viewer.\r\n-   **Optimistic UI**: Added optimistic updates when redoing user messages or interrupting responses for a more \"instant\" feel.\r\n-   **Context Tooltips**: Hovering over the \"Read File\" or \"Edit File\" power tool messages now shows a tooltip with the full file path.\r\n-   **UI Refinements**: Set a max width for the prompt autocomplete popup and fixed layout overflows for model dropdowns.\r\n-   **Workflow Fixes**: Properly interrupting the \"determine task state\" action if an agent is stopped early.\r\n\r\n💡 **We’d love to hear your thoughts!** Your feedback helps us improve AiderDesk - feel free to share any suggestions or issues. Happy coding! 🚀","2026-01-19T06:13:02",{"id":202,"version":203,"summary_zh":204,"released_at":205},102185,"v0.48.1","## Hotfix\r\n\r\n- properly passing the model info (max output tokens, input tokens...) to Aider in the connector\r\n\r\nComplete release notes for version 0.48.0 can be found [here](https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk\u002Freleases\u002Ftag\u002Fv0.48.0).","2026-01-12T13:49:07",{"id":207,"version":208,"summary_zh":209,"released_at":210},102186,"v0.48.0","## Highlights\r\n\r\n### 🌐 Expanded Provider Support: LiteLLM, Synthetic, & OpenCode ZEN\r\nWe are continuing to broaden your choice of LLMs with direct support for three new providers:\r\n-   **LiteLLM**: A huge thank you to **@jutaz** for contributing this integration, allowing for even more flexible model routing.\r\n-   **Synthetic Provider**: Access a new range of powerful models directly.\r\n-   **OpenCode ZEN**: Take advantage of their [free models](https:\u002F\u002Fopencode.ai\u002Fdocs\u002Fzen\u002F#models) currently available for a limited time!\r\n\r\n### ⚡ Z.AI One-Click MCP Setup\r\nFor users on the **Z.AI Coding Plan**, we've made it incredibly easy to supercharge your agent. You can now set up Z.AI's specialized MCP servers (including [Vision](https:\u002F\u002Fdocs.z.ai\u002Fdevpack\u002Fmcp\u002Fvision-mcp-server) for images and [Web Search](https:\u002F\u002Fdocs.z.ai\u002Fdevpack\u002Fmcp\u002Fsearch-mcp-server) for web results) with a single click directly from the provider settings.\r\n\r\n### 📺 Real-time Bash Tool Streaming\r\nNo more waiting in the dark! The **bash power tool** now supports live output streaming. If your agent is running a long-duration script or a complex build command, you can see the terminal output in real-time as it happens, making it much easier to monitor progress.\r\n\r\n### 🌳 Professional Worktree & Task Management\r\nWe've added several features to make working with Git worktrees and complex tasks even smoother:\r\n-   **Worktree Symlinks**: You can now configure folders (like `node_modules` or `vendor`) to be automatically symlinked into your worktrees via **Settings -> Tasks -> Worktree**. This ensures your agent has access to all dependencies without duplicating data.\r\n-   **Enhanced Sidebar Visibility**: A new worktree icon appears on task items in the sidebar, so you can immediately see which tasks are running in isolated environments.\r\n-   **Manual State Control**: You can now manually override a task's state via a new \"Change State\" submenu in the task context menu.\r\n-   **Task Info Command**: Use the new `\u002Ftask-info` command to see detailed metadata about your current task, with quick options to copy the path or open the worktree folder.\r\n\r\n### 🛠️ Smart MCP Placeholders\r\nTo provide better control over MCP server configurations, we've refined our directory placeholders:\r\n-   `${taskDir}`: Points to the current working directory (either the project root or the specific worktree directory).\r\n-   `${projectDir}`: Now **always** points to the main project root, even when working inside a worktree.\r\n\r\n## Other Changes\r\n\r\n-   **Performance**: Optimized performance when switching between different tasks and projects.\r\n-   **Agent Creator Skill**: Added a new [agent-creator skill](https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk\u002Ftree\u002Fmain\u002F.aider-desk\u002Fskills\u002Fagent-creator) to help you design and create custom agent profiles via interactive Q&A. Just copy the folder into `~\u002F.aider-desk\u002Fskills\u002F` folder and ask agent that has access to the skills to help you create a specific agent.\r\n-   **UI Enhancements**: Added missing menu actions to the last user message, added a favicon to the browser client, and updated the empty task message.\r\n-   **Task Configuration**: Added an option to hide task state actions for a cleaner sidebar experience.\r\n-   **Power Tool Fixes**: Power tools now properly resolve the `~` character to the user's HOME directory path.\r\n-   **Platform Fixes**: Improved copy-paste handling for MacOS and mobile devices, and fixed Markdown export functionality in browser environments.\r\n-   **Aider Integration**: Properly setting default `max_output_tokens` in Aider modes and respecting compacting thresholds when updated within a task.\r\n-   **Context & Memory**: Removed a duplicate `AGENTS.md` entry from the system prompt knowledge section and improved the `retrieve_memories` tool description for better LLM utilization.\r\n-   **Defaults**: Agent mode is now set as the default for all new projects.\r\n\r\n💡 **We’d love to hear your thoughts!** Your feedback helps us improve AiderDesk - feel free to share any suggestions or issues. Happy coding! 🚀","2026-01-12T06:20:11",{"id":212,"version":213,"summary_zh":214,"released_at":215},102187,"v0.47.0","## Highlights\r\n\r\n### 📋 Smart Task Management & State Tracking\r\nWe've introduced a comprehensive task lifecycle system to help you track your development workflow more intuitively than ever!\r\n-   **Intelligent Task States**: Tasks now move through meaningful states like *Todo*, *In Progress*, *More Info Needed*, *Ready for Review*, and *Done*.\r\n-   **Contextual Actions**: The UI now presents the most relevant actions based on a task's current state - such as *Execute*, *Resume*, *Proceed*, or *Mark as Done* - streamlining your daily operations.\r\n-   **AI-Powered Organization**: AiderDesk can now automatically generate descriptive task names and determine task states using AI.\r\n-   **Task Pinning**: Keep your most important or recurring tasks at the very top of your sidebar for instant access.\r\nYou can customize these features (including disabling auto-naming or the advanced \"Smart\" states) in the new **Settings -> Tasks** page.\r\n\r\n### 🧠 Streamlined Memory Management\r\nManaging your agent's long-term memory just got easier. You can now **delete specific memories directly from the \"Retrieve memory\" tool message** in the chat. This allows for instant cleanup of outdated or irrelevant information without ever leaving your conversation flow.\r\n\r\n### 🌐 Enhanced Fetch Tool: Markdown, HTML, & Raw Support\r\nThe `fetch` power tool is now much more versatile! We've added an optional `format` parameter that allows agents to retrieve web content as:\r\n-   **Markdown** (Default): Optimized for LLM consumption.\r\n-   **HTML**: For raw structure analysis.\r\n-   **Raw**: For non-HTML content or API responses.\r\nThis improvement ensures better reliability when agents need to gather information from a wider variety of web sources and fixes issues where certain URLs would previously fail to load.\r\n\r\n## Other Changes\r\n\r\n-   **Agent Tools**: The `list_tasks` tool now supports filtering by state, giving agents better oversight of your project's progress.\r\n-   **Task Multiselect**: Added `Shift + Click` support for selecting ranges of tasks in the sidebar, making batch archiving or deletion even faster.\r\n-   **Model Library**: Improved the cost input field to prevent annoying reformatting while you are actively typing.\r\n-   **Agent Reliability**: Added sanitization for agent profiles to ensure all required configuration fields are always present.\r\n-   **Context Accuracy**: Improved path matching when checking if a file is already in the context before prompting to add it.\r\n-   **Z.AI Optimization**: Updated the default temperature for Z.AI models to 0.7 for more balanced and creative results.\r\n\r\n💡 **We’d love to hear your thoughts!** Your feedback helps us improve AiderDesk - feel free to share any suggestions or issues. Happy coding! 🚀","2026-01-05T06:38:54",{"id":217,"version":218,"summary_zh":219,"released_at":220},102188,"v0.46.0","## Highlights\r\n\r\n### 📝 Powerful Prompt Templating & Custom System Prompts\r\nTake full control over how your AI thinks and acts! Thanks to a major contribution from **@pcfreak30**, we've implemented a modular **templating system for agent prompts**. This allows power users to override:\r\n-   **System Prompts**: Define the core personality and rules for your main agent.\r\n-   **Workflow Prompts**: Customize how the agent handles conversation compaction, project initialization, conflict resolution, and specific workflow steps.\r\nThis level of customization allows you to fine-tune AiderDesk for your specific tech stack or coding style. Check out the [Custom Prompts documentation](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Fconfiguration\u002Fcustom-prompts) to get started!\r\n\r\n### ⌨️ Boost Your Speed with Keyboard Shortcuts\r\nNavigation just got a whole lot faster! Huge thanks to **@ohnum** for implementing a comprehensive **Keyboard Shortcuts** system. You can now quickly:\r\n-   Switch between projects and tasks.\r\n-   Jump straight to the prompt field, Settings, or Model Library.\r\n-   Navigate the UI without touching your mouse.\r\nView and customize all shortcuts in **Settings -> Hotkeys**.\r\n\r\n### 🧠 Improved Thinking Blocks & Message Navigation\r\nWe've overhauled how AI \"thinking\" and user history are displayed to provide better clarity during complex tasks:\r\n-   **Distinct Thinking Blocks**: Model reasoning is now visually distinct from the final response and visible by default, as it often contains crucial planning details.\r\n-   **User Message Navigation**: User messages are now more prominent, and we've added **navigation buttons** to let you jump quickly between your prompts in a long conversation.\r\n\r\n### 🛠️ Advanced Agent Tools: Task Execution & Memory Updates\r\nWe are continuing to expand the capabilities of our autonomous agents:\r\n-   **Create & Execute Tasks**: The `create_task` tool now includes an optional `execute` argument. An agent can now create a sub-task, trigger its execution, and wait for the result before proceeding.\r\n-   **Memory Updates**: A new `update_memory` tool has been added to the memory management suite for more precise control over stored information (disabled by default for safety).\r\n\r\n## Other Changes\r\n\r\n-   **Testing Framework**: Added Vitest and a specialized \"Writing Tests\" skill for developers contributing to AiderDesk.\r\n-   **Provider Management**: Added the ability to disable entire provider profiles to hide their models from the UI.\r\n-   **UI Enhancements**: Added icons to worktree actions, improved cursor styling for context headers, and added a copy button to bash tool commands.\r\n-   **Multi-select Suggestions**: Use Ctrl\u002FCmd + Click to select multiple files from the suggestion list.\r\n-   **Mobile\u002FTouch Fixes**: Improved task selection reliability on touch devices.\r\n-   **Simultaneous Tasks**: The browser client now correctly supports running multiple tasks at once.\r\n\r\n💡 **We’d love to hear your thoughts!** Your feedback helps us improve AiderDesk - feel free to share any suggestions or issues. Happy coding! 🚀","2025-12-29T07:46:36",{"id":222,"version":223,"summary_zh":224,"released_at":225},102189,"v0.45.1","## Hotfix\r\n\r\n- fixed \"failed to parse chunk\" issue with Ollama `gpt-oss` models when using Aider modes\r\n\r\nComplete release notes for version 0.45.0 can be found [here](https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk\u002Freleases\u002Ftag\u002Fv0.45.0).","2025-12-23T10:50:38",{"id":227,"version":228,"summary_zh":229,"released_at":230},102190,"v0.45.0","## Highlights\r\n\r\n### 🌳 Advanced Git Worktree Management: Seamlessly Multi-task & Merge\r\nWe've significantly leveled up our Git Worktree support to make parallel task management more efficient than ever!\r\n-   **Enhanced Merging & Rebasing**: You can now merge worktrees into custom target branches with specific commit messages or rebase them against the latest changes.\r\n-   **Visual Status Indicators**: Stay informed with real-time UI indicators showing ahead commits, uncommitted changes, and conflict predictions.\r\n-   **AI-Assisted Conflict Resolution**: Encounter a conflict? We've added a specialized **Conflict Resolution agent** that can automatically analyze and fix merge conflicts for you.\r\n-   **Improved UI Flow**: Worktree safety confirmations have been moved directly into the UI for a smoother, less disruptive experience.\r\n\r\n![rebase](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F530355f6-e5c4-4aeb-8648-885bc98e7ae1)\r\n\r\n### 🪝 New Hook System: Customize & Automate AiderDesk\r\nIntroducing a powerful **Hook System** that allows you to extend AiderDesk using JavaScript! You can now create script files to intercept and react to application events like tasks, prompts, and tool executions.\r\n-   **Prompt Control**: Automatically scan prompts for sensitive information or append custom instructions.\r\n-   **Tool Governance**: Monitor or deny tool executions based on security rules or input arguments.\r\n-   **Real-world Example**: Check out our [Wakatime hook](https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk\u002Fblob\u002Fmain\u002F.aider-desk\u002Fhooks\u002Fwakatime-hooks.js) proof-of-concept to track your coding activity.\r\nRead the full documentation on how to build your own hooks [here](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Ffeatures\u002Fhooks).\r\n\r\n### 🔍 Versatile Diff Viewer: Side-by-Side, Unified, & Compact Views\r\nWe know everyone has a different preference for reviewing code changes. You can now choose between three distinct diff view types:\r\n-   **Side-by-Side**: The classic view for clear comparison.\r\n-   **Unified**: A streamlined, single-column view favored by many developers.\r\n-   **Compact**: A brand-new mode that chunks changes by both lines and words for the most efficient review possible.\r\n\r\n![diff-views](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F19dac32e-685f-47a9-ad69-da5208623266)\r\n\r\n### 📂 Intelligent File Detection in Prompts\r\nSave time and tokens with automatic file detection! When working in Aider modes, AiderDesk now scans your prompts for file names. If it detects a file that isn't currently in your context, it will interactively ask if you'd like to add it before sending the request to the LLM. Thanks to **@codycwiseman** for this excellent suggestion!\r\n\r\n## Other Changes\r\n\r\n-   Improved project file listing to correctly respect `.gitignore` patterns even in non-git mode.\r\n-   Added an auto-restart mechanism for Gemini voice sessions to mitigate the 30-second timeout limitation.\r\n-   The \"Retrying\" message from Aider is now correctly propagated to the UI.\r\n-   The paste action in the \"Add File\" dialog now supports non-file text.\r\n-   Fixed the height of the MCP server configuration form.\r\n-   Customized the message for the skill activation tool for better clarity.\r\n-   Browser client now maintains the current mode when switching other parameters.\r\n\r\n💡 **We’d love to hear your thoughts!** Your feedback helps us improve AiderDesk - feel free to share any suggestions or issues. Happy coding! 🚀","2025-12-22T07:26:16",{"id":232,"version":233,"summary_zh":234,"released_at":235},102172,"v0.59.0","## Highlights\r\n\r\n### 🪟 Multiple Windows & Customizable Titles\r\nYou can now open **multiple AiderDesk windows** simultaneously. Need to keep an eye on one task while actively working on another? Simply open a new window and switch to the project or task you need. Each window operates independently, making multi-tasking across your projects more natural than ever. You can also **customize window titles**, so you can quickly identify which window is which at a glance. Use File -> New menu action (CTRL+N) or for Open in New Window action from a project's tab menu.\r\n\r\n### 📋 System Logs Viewer with Extension Filtering\r\nWe've added a built-in **System Logs Viewer**, giving you direct access to application logs without hunting through log files on disk. What makes it especially useful is the ability to **quickly filter logs by extension** - if you're developing or debugging your own extensions, you can isolate just the relevant log entries in seconds.\r\n\r\n### 🧹 Cleaner Interruption Handling\r\nInterrupting an agent mid-response has been improved for a clearer experience. **Unfinished messages are now automatically removed** when you interrupt, making it unambiguous that they are not part of the conversation context going forward. This prevents confusion where partial or incomplete responses could previously linger and affect the agent's subsequent behavior.\r\n\r\n### 💻 AiderDesk npm CLI: New Way to Run the Backend\r\nFor power users, we've added the **`aiderdesk` npm binary** with support for various subcommands. This provides an alternative to running AiderDesk via Docker - it starts just the \"backend\" part of the application, allowing you to access the UI through your browser. It's an option for those who prefer a quick setup. Learn more in our [npm CLI documentation](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Fadvanced\u002Fnpm-cli).\r\n\r\n## Other Changes\r\n\r\n-   **Updated Files Improvements**: The \"view all files\" mode state is now remembered between sessions, and updated files are properly refreshed after git actions (including rebase).\r\n-   **Power Edit Tool Fix**: The edit tool now correctly handles files with `\\r\\n` (Windows-style) line separators.\r\n-   **Compact View Fixes**: Correctly deleting messages and \"messages up to here\" actions when in compact view mode.\r\n-   **Terminal Focus**: The terminal is now properly focused when activated.\r\n-   **Startup Reliability**: Removed the timeout when installing Python packages on start-up to prevent premature failures on slower connections.\r\n-   **Dependency Fixes**: Added missing dependency for Auggie SDK provider in the Electron app.\r\n-   **New Extensions**: Added `fast-grep` extension based on trigram decomposition (experimental) and `\u002Fplannotator-review` command to the Plannotator extension.\r\n\r\n💡 **We'd love to hear your thoughts!** Your feedback helps us improve AiderDesk - feel free to share any suggestions or issues. Happy coding! 🚀","2026-03-30T05:18:56",{"id":237,"version":238,"summary_zh":239,"released_at":240},102173,"v0.58.1","## Hotfix\r\n\r\n- fixed auto-compaction not working when hitting the threshold\r\n\r\nComplete release notes for version 0.58.0 can be found [here](https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk\u002Freleases\u002Ftag\u002Fv0.58.0).","2026-03-24T06:25:43",{"id":242,"version":243,"summary_zh":244,"released_at":245},102174,"v0.58.0","## Highlights\r\n\r\n### 🛠️ Enhanced Extension Capabilities & Visual Refresh\r\nWe've continued to expand the Extension API with powerful new capabilities:\r\n-   **New Extension Hooks**: Added `onAgentStepStarted` hook to monitor agent execution at a granular level.\r\n-   **New Context Functions**: Extensions can now use `openPath` to open files\u002Ffolders in the system and `runPromptInAgent` to programmatically execute prompts.\r\n-   **Visual Refresh**: The Extensions settings page got a visual overhaul - extensions are now more distinct when browsing, and we've added support for **icons** for even better visibility.\r\n\r\nTo help you get started with creating extensions, check out our [extension-creator skill](https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk\u002Ftree\u002Fmain\u002F.aider-desk\u002Fskills\u002Fextension-creator), browse the [documentation](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Fextensions), or join our [Discord](https:\u002F\u002Fdiscord.com\u002Finvite\u002FdyM3G9nTe4) for support.\r\n\r\n### 🧩 BMAD Mode is Now an Extension\r\nAs part of our commitment to making AiderDesk as extensible as possible, we've extracted the **BMAD Method mode** from the core application into a standalone extension. This was the original vision for BMAD in AiderDesk - it gives you full flexibility to tune the mode to your liking, inspect its implementation, and even contribute improvements.\r\n-   **Easy Installation**: Head to **Settings -> Extensions -> Available** and install it with one click.\r\n-   **Version Note**: The extension currently uses BMAD version 6.0.4. The latest BMAD versions (6.1.0, 6.2.0) introduced breaking changes, but we're actively working on updating the extension to benefit from those improvements in upcoming releases.\r\n-   **Full Functionality**: The extension maintains all the same features you're used to - Brainstorming, Quick Spec, User Stories, Code Review, and more.\r\n\r\n### ⚙️ Configurable Auxiliary Models\r\nYou can now configure **auxiliary models** for background tasks like task name generation, task state determination, and commit message creation. This lets you use faster (and cheaper) models for these simple operations while reserving your primary model for the actual work.\r\n*Configure this in **Settings -> Tasks** and pick the models that fit your workflow and budget.*\r\n\r\n### 🔄 Quick Subagents Toggle\r\nA small but handy addition: we've added a **quick toggle button for subagents** directly in the Agent selector, right next to the existing toggles. If your workflow sometimes requires your main agent to work solo without spawning subagents, you can now switch this on or off without diving into Settings.\r\n\r\n## Other Changes\r\n\r\n-   **Performance Optimizations**:\r\n    -   Implemented virtualization for the task sidebar for smoother scrolling with many tasks.\r\n    -   Optimized prompt field and Files sidebar rendering for better responsiveness.\r\n-   **Subagent Control**: Added support for **cancelling running subagents** mid-execution.\r\n-   **Bug Fixes**:\r\n    -   Properly triggering `onTaskUpdated` on empty task updates.\r\n    -   Correctly handling added extension repositories.\r\n    -   Properly saving task updates when working mode changes.\r\n    -   Available extension cards now correctly show remote info.\r\n\r\n💡 **We'd love to hear your thoughts!** Especially on the BMAD extension transition - your feedback helps us improve AiderDesk. Happy coding! 🚀","2026-03-23T06:18:39",{"id":247,"version":248,"summary_zh":249,"released_at":250},102175,"v0.57.0","## Highlights\r\n\r\n### 🎨 Extension UI Components System: Build Interactive Interfaces\r\nWe've introduced a powerful **UI Components** system that allows extensions to render custom React components directly within the AiderDesk interface. This opens up entirely new possibilities for extending the app - whether you want to add buttons to the task status bar, custom selectors in the header, or interactive panels in the messages area.\r\n-   **Rich Component Library**: Access ready-to-use components like `Button`, `ModelSelector`, `MultiSelect`, `Chip`, `Input`, and more.\r\n-   **Multiple Placements**: Add UI elements to various locations - task status bar (left\u002Fright), header (left\u002Fright), task input area, and messages section (top\u002Fbottom).\r\n-   **Interactive Actions**: Handle user interactions with full access to the Extension API, including data loading and state management.\r\n\r\nThis is the UI placement demo extension in action, showcasing where you can position your own UI extensions:\r\n\u003Cimg width=\"1393\" height=\"830\" alt=\"image\" src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F273f041d-eac3-4676-941b-2cf0b3112f4e\" \u002F>\r\n\r\nTo demonstrate the power of this new system, we've built two showcase extensions:\r\n-   **[providers-quota-extension](https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk\u002Ftree\u002Fmain\u002Fpackages\u002Fextensions\u002Fextensions\u002Fproviders-quota-extension)**: Displays real-time quota information for Z.ai Coding Plan and Synthetic providers directly in the UI, so you always know how much usage you have left.\r\n-   **[multi-model-run](https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk\u002Ftree\u002Fmain\u002Fpackages\u002Fextensions\u002Fextensions\u002Fmulti-model-run)**: A powerful tool that lets you run the same prompt across multiple models simultaneously, each in its own worktree. This is perfect for comparing how different models approach the same problem and choosing the best outcome.\r\n\r\nWe've also added a new `openUrl` function to the Extension context, giving extensions the ability to open URLs in three different ways: in a new app window, in a modal overlay (the same style used for Settings and Model Library), or in an external browser. The **[plannotator](https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk\u002Ftree\u002Fmain\u002Fpackages\u002Fextensions\u002Fextensions\u002Fplannotator)** extension has already been updated to use this - it now opens the plan review interface directly within the app for a seamless experience. If you're using Plannotator, make sure to reinstall it to benefit from this change!\r\n\r\nThe UI extensions give you incredible power to enhance your workflow. We're excited to see what you come up with - make sure to share your creations in our **[Discord](https:\u002F\u002Fdiscord.gg\u002FMRqqmr3MAs)**, where we're building a growing library of extensions for everyone!\r\n\r\n### 🛠️ Programmatic Tool Calling Extension\r\nAs another notable addition to the extensions ecosystem, we've added the **[programmatic-tool-calls](https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk\u002Ftree\u002Fmain\u002Fpackages\u002Fextensions\u002Fextensions\u002Fprogrammatic-tool-calls)** extension. This implements a pattern recently introduced by Anthropic that helps the LLM better organize context and tool calls. Instead of making multiple separate tool invocations, the LLM can write JavaScript code that executes multiple tools programmatically and receives a combined result - ideal for complex, multi-step operations.\r\n\r\n### 📄 File Preview Modal: Quick File Viewing\r\nYou can now click on any file in the **Project Files** sidebar to open a code viewer directly in AiderDesk. No more switching to your IDE just to quickly check a file's contents - we've got you covered. We've also added the ability to expand\u002Fcollapse folders by clicking directly on the folder name for easier navigation.\r\n\r\n### 🌐 New Provider: Auggie SDK\r\nWe've added support for the **Auggie SDK** provider! If you have an [Augment Code](https:\u002F\u002Fwww.augmentcode.com) subscription, you can now use it within AiderDesk at no additional cost.\r\n\r\n### 🔧 Enhanced Task Tools for Agents\r\nThe `create_task` tool in our [Task Tools](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Fagent-mode\u002Ftask-tools) has been updated with additional input parameters. Agents can now specify `mode`, `auto-approve`, `worktree` vs `local` mode, and create tasks specifically as subtasks. This gives your agents much more flexibility in how they organize and execute work autonomously.\r\n\r\n### 🐘 Alibaba Coding Plan Improvements\r\nWe've updated the underlying library for the **Alibaba Coding Plan** provider for better stability. We've also added the ability to configure **thinking** settings - including enabled state and token budget - for supported models.\r\n\r\n## Other Changes\r\n\r\nThanks to **@jutaz** for several contributions:\r\n-   Suppress Mermaid rendering errors for cleaner output\r\n-   Improved process killing reliability in the bash tool\r\n-   Allow reverting untracked (new) files in the Updated Files sidebar\r\n\r\nAnd also to our new contributor **@zerone0x** for fixing an issue with incorrect shell used in the Bash tool.\r\n\r\nAdditional improvements:\r\n-   **UI Polish","2026-03-16T06:32:03",{"id":252,"version":253,"summary_zh":254,"released_at":255},102176,"v0.56.0","## Highlights\r\n\r\n### 🧩 Extension Management & New Powerful Extensions\r\nWe’ve introduced a dedicated **Extension Settings** page to give you full control over your AiderDesk customizations. You can now easily enable or disable installed extensions, define custom repositories, and install new extensions directly from the UI. \r\n\r\nTo showcase the power of this new system, we’ve added several game-changing extensions:\r\n-   **[LSP Integration](https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk\u002Ftree\u002Fmain\u002Fpackages\u002Fextensions\u002Fextensions\u002Flsp)**: Automatically checks your code for errors after the agent uses the edit or write tools, ensuring higher code quality.\r\n-   **[Chunkhound Semantic Search](https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk\u002Ftree\u002Fmain\u002Fpackages\u002Fextensions\u002Fextensions\u002Fchunkhound-on-semantic-search-tool)**: Replaces the internal semantic search with the [Chunkhound](https:\u002F\u002Fchunkhound.github.io) alternative for even better context retrieval.\r\n-   **[Plannotator](https:\u002F\u002Fgithub.com\u002Fhotovo\u002Faider-desk\u002Ftree\u002Fmain\u002Fpackages\u002Fextensions\u002Fextensions\u002Fplannotator)**: Adds a new mode that enhances your workflow with a dedicated planning phase and a beautiful GUI for reviewing and refining agent plans. Based on the awesome [Plannotator](https:\u002F\u002Fgithub.com\u002Fbacknotprop\u002Fplannotator) project!\r\n-   **Community Contributions**: A huge thanks to **@klocus** for contributing **[Dynamic Context Pruning](https:\u002F\u002Fgithub.com\u002Fklocus\u002Faider-desk-dcp)** (for managing long-running contexts) and **[Conductor](https:\u002F\u002Fgithub.com\u002Fklocus\u002Faider-desk-conductor)** (for multi-agent orchestration).\r\n\r\nJoin our **[Discord](https:\u002F\u002Fdiscord.gg\u002FMRqqmr3MAs)** to discover these and other extensions.\r\n\r\n### 🌳 Enhanced Worktree & Updated Files UX\r\nReviewing changes in **Worktree mode** is now more comprehensive. The **Updated Files** list now includes files that have been locally committed within the worktree, giving you a complete picture of all changes compared to the main branch. This is particularly useful in \"Code\" mode where changes are committed automatically.\r\nAdditionally, **Inline Code Requests** (triggered by clicking a line in a diff) now utilize your current task and mode by default, keeping your workflow focused. You still have the option to spin up a brand-new task if the change requires a fresh context.\r\n\r\n### 📅 Organized Task Sidebar: Day Grouping\r\nKeeping track of your work history is now much easier. The task sidebar now groups your tasks by date (e.g., *Today*, *Yesterday*), allowing you to quickly locate recent work or dive back into older projects with a glance.\r\n\r\n## Other Changes\r\n\r\n-   **Bash Tool Improvements**: Added timeout information to bash tool messages and fixed an issue where `stdout` was not properly captured for piped commands.\r\n-   **BMAD Method**: Starting an empty \"Quick Spec\" flow now correctly initializes the starting context.\r\n-   **UI Refinements**: Added a full-width toggle to the Updated Files page, refined the welcome message layout, and removed unnecessary tooltips on overlay pages.\r\n-   **Reliability**:\r\n    -   Added connection timeouts during startup and support for the `AIDER_DESK_NO_UPDATES` environment variable.\r\n    -   Aider now correctly initializes when a repo map is included in an Agent profile.\r\n    -   Increased LLM processing timeouts in Agent mode for more complex tasks.\r\n    -   Fixed an issue where multiple notifications would trigger for queued prompts.\r\n-   **Extension API**: Exposed the `allTools` argument to the extension `tool.execute` function and added the `onPromptTemplate` event for managing system prompts.\r\n-   **UX Tweaks**: Your prompt field text is now preserved when switching between different projects, and \"Activate Skill\" messages are preserved during conversation compaction.\r\n\r\n**We’d love to hear your thoughts!** Especially on the new extensions - your feedback helps us improve AiderDesk. Happy coding! 🚀","2026-03-09T06:21:13",{"id":257,"version":258,"summary_zh":259,"released_at":260},102177,"v0.55.0","## Highlights\r\n\r\n### 🔌 Extension System: A New Era of Customization\r\nWe are thrilled to introduce a brand-new **Extension System** that replaces our previous hook system with vastly expanded capabilities. Extensions now give you great control over AiderDesk's behavior:\r\n-   **Register New Capabilities**: Add custom **tools**, create slash **commands** like `\u002Fgenerate-tests`, define custom **modes**, and provide specialized **agent profiles**.\r\n-   **Hook into Events**: Listen to and react to virtually any event in AiderDesk - from task creation and agent execution to tool calls and file context changes.\r\n-   **Modify Behavior in Real-Time**: Many events allow you to modify data before it's processed. You can block dangerous operations, transform prompts, filter context files, auto-answer approvals, or even modify AI responses.\r\nCheck out our [Extension Documentation](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Fextensions) to get started. We've included several examples like WakaTime integration, sound notifications, and more. We're excited to see what you come up with!\r\n\r\n> ⚠️ **Depprecation Notice**: The previous hook system is now considered **deprecated** and will be removed in **version 0.60.0**. Please migrate your hooks to the new Extension system.\r\n\r\n### ✅ Commit Changes Directly from Updated Files\r\nThe **Updated Files** page has been leveled up! You can now **commit your changes directly from the dialog** without switching to a terminal or your IDE with an option to also generate commit message. We've also added a new **\"View All Files\"** mode that lets you see the diffs for all updated files on a single page, making code review faster and more convenient.\r\n\r\n### 🌐 New Provider: Alibaba Coding Plan\r\nWe've added direct support for the **Alibaba Coding Plan** provider. This subscription offers access to a suite of powerful models - including **glm-5**, **Kimi-K2.5**, and **MiniMax-M2.5** - all in one place at a great price. If you've been looking to diversify your model options, this is a fantastic addition to your toolkit.\r\n\r\n### 🎓 BMAD Mode: Out of Preview\r\nWe're happy to announce that **BMAD Method mode** is now officially out of preview! After several iterations based on your valuable feedback, the workflows for brainstorming, story creation, and code review are stable and production-ready. Thank you for helping us shape this feature!\r\n\r\n## Other Changes\r\n\r\n-   **Enhanced Reasoning**: Added \"ultrathink\" (xhigh) reasoning effort option for OpenAI and OpenAI-compatible providers.\r\n-   **File Locking**: Added automatic file locking when using the edit power tool to prevent race conditions during concurrent edits.\r\n-   **Performance**: Sped up application startup by loading long-running managers in the background.\r\n-   **Memory Stability**: Improved handling of the memory embedding process when switching between models and removed `BAAI\u002Fbge-large-en-v1.5` due to stability issues.\r\n-   **Task Export**: Added \"Copy as Markdown\" action to the task actions menu in the sidebar.\r\n-   **Bug Fixes**: Omitted deleted files from the Updated Files page to prevent errors.\r\n\r\n💡 **We'd love to hear your thoughts!** Especially on the new Extension system - your feedback helps us improve AiderDesk. Happy coding! 🚀","2026-03-02T06:26:22",{"id":262,"version":263,"summary_zh":264,"released_at":265},102191,"v0.44.0","## Highlights\r\n\r\n### 🧠 Agent Memory & 🛠️ Skill System: Supercharge Your Agent's Intelligence!\r\nToday, we're introducing two groundbreaking features to make your agents more capable than ever:\r\n-   **Agent Memory**: Your agents can now remember past actions and decisions, leading to more informed and efficient behavior in future tasks. This enhances their awareness of the project beyond just rule files. Learn more about Memory [here](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Ffeatures\u002Fmemory).\r\n-   **Skill System**: Extend your agent's capabilities with reusable, on-demand expertise! Skills provide agents with specific contexts and guided workflows, enabling them to tackle complex tasks in a repeatable and predictable manner. Discover more about Skills [here](https:\u002F\u002Faiderdesk.hotovo.com\u002Fdocs\u002Ffeatures\u002Fskills).\r\nTo help you leverage these new features, we've also added \"Use memory tools\" and \"Use task tools\" quick actions to the Agent selector.\r\n\r\n### 🎤 Enhanced Voice Control: Browser Support & Advanced Settings\r\nWe've significantly upgraded the Voice Control feature:\r\n-   **Browser Client Support**: You can now use voice control directly from your browser client, expanding accessibility.\r\n-   **Dedicated Voice Settings**: A new configuration page allows you to select your preferred microphone, choose providers, and fine-tune provider-specific settings like language for OpenAI.\r\n\r\n### ✅ Streamlined Workflow with Auto Approve Enhancements\r\nTake control of your agent's interactions with new Auto Approve features:\r\n-   **Project-Level Auto Approve Lock**: Enable Auto Approve by default for all new tasks within a project, streamlining your workflow.\r\n-   **Auto Approve for Aider Modes**: Enjoy a continuous flow in Aider's Code, Ask, and Architect modes by automatically approving agent questions, ideal for users who trust their LLM's judgment.\r\n\r\n## Other Changes\r\n\r\n-   Fixed an issue with the missing `reasoning_content` field when using `deepseek-reasoner`.\r\n-   Optimized agent profile updates and related event handling for improved performance.\r\n-   `max_tokens` are now properly passed when using Aider modes.\r\n-   The architect model is now correctly set for each task.\r\n-   API_KEY environment variables are now properly read for some providers when using Aider modes.\r\n-   Updated the link to telemetry documentation.\r\n-   User messages are now correctly stored even if Aider returns a response in case of an interrupt or error.\r\n-   Fixed microphone permission issues on macOS.\r\n-   Fixed project name display in Agent settings for Windows paths.\r\n\r\n💡 **We’d love to hear your thoughts!** Your feedback helps us improve AiderDesk - feel free to share any suggestions or issues. Happy coding! 🚀","2025-12-15T06:49:17"]