[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-langtalks--swe-agent":3,"tool-langtalks--swe-agent":61},[4,18,26,36,44,52],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":17},4358,"openclaw","openclaw\u002Fopenclaw","OpenClaw 是一款专为个人打造的本地化 AI 助手，旨在让你在自己的设备上拥有完全可控的智能伙伴。它打破了传统 AI 助手局限于特定网页或应用的束缚，能够直接接入你日常使用的各类通讯渠道，包括微信、WhatsApp、Telegram、Discord、iMessage 等数十种平台。无论你在哪个聊天软件中发送消息，OpenClaw 都能即时响应，甚至支持在 macOS、iOS 和 Android 设备上进行语音交互，并提供实时的画布渲染功能供你操控。\n\n这款工具主要解决了用户对数据隐私、响应速度以及“始终在线”体验的需求。通过将 AI 部署在本地，用户无需依赖云端服务即可享受快速、私密的智能辅助，真正实现了“你的数据，你做主”。其独特的技术亮点在于强大的网关架构，将控制平面与核心助手分离，确保跨平台通信的流畅性与扩展性。\n\nOpenClaw 非常适合希望构建个性化工作流的技术爱好者、开发者，以及注重隐私保护且不愿被单一生态绑定的普通用户。只要具备基础的终端操作能力（支持 macOS、Linux 及 Windows WSL2），即可通过简单的命令行引导完成部署。如果你渴望拥有一个懂你",349277,3,"2026-04-06T06:32:30",[13,14,15,16],"Agent","开发框架","图像","数据工具","ready",{"id":19,"name":20,"github_repo":21,"description_zh":22,"stars":23,"difficulty_score":10,"last_commit_at":24,"category_tags":25,"status":17},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,"2026-04-05T11:01:52",[14,15,13],{"id":27,"name":28,"github_repo":29,"description_zh":30,"stars":31,"difficulty_score":32,"last_commit_at":33,"category_tags":34,"status":17},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",141543,2,"2026-04-06T11:32:54",[14,13,35],"语言模型",{"id":37,"name":38,"github_repo":39,"description_zh":40,"stars":41,"difficulty_score":32,"last_commit_at":42,"category_tags":43,"status":17},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",107888,"2026-04-06T11:32:50",[14,15,13],{"id":45,"name":46,"github_repo":47,"description_zh":48,"stars":49,"difficulty_score":10,"last_commit_at":50,"category_tags":51,"status":17},4487,"LLMs-from-scratch","rasbt\u002FLLMs-from-scratch","LLMs-from-scratch 是一个基于 PyTorch 的开源教育项目，旨在引导用户从零开始一步步构建一个类似 ChatGPT 的大型语言模型（LLM）。它不仅是同名技术著作的官方代码库，更提供了一套完整的实践方案，涵盖模型开发、预训练及微调的全过程。\n\n该项目主要解决了大模型领域“黑盒化”的学习痛点。许多开发者虽能调用现成模型，却难以深入理解其内部架构与训练机制。通过亲手编写每一行核心代码，用户能够透彻掌握 Transformer 架构、注意力机制等关键原理，从而真正理解大模型是如何“思考”的。此外，项目还包含了加载大型预训练权重进行微调的代码，帮助用户将理论知识延伸至实际应用。\n\nLLMs-from-scratch 特别适合希望深入底层原理的 AI 开发者、研究人员以及计算机专业的学生。对于不满足于仅使用 API，而是渴望探究模型构建细节的技术人员而言，这是极佳的学习资源。其独特的技术亮点在于“循序渐进”的教学设计：将复杂的系统工程拆解为清晰的步骤，配合详细的图表与示例，让构建一个虽小但功能完备的大模型变得触手可及。无论你是想夯实理论基础，还是为未来研发更大规模的模型做准备",90106,"2026-04-06T11:19:32",[35,15,13,14],{"id":53,"name":54,"github_repo":55,"description_zh":56,"stars":57,"difficulty_score":10,"last_commit_at":58,"category_tags":59,"status":17},4292,"Deep-Live-Cam","hacksider\u002FDeep-Live-Cam","Deep-Live-Cam 是一款专注于实时换脸与视频生成的开源工具，用户仅需一张静态照片，即可通过“一键操作”实现摄像头画面的即时变脸或制作深度伪造视频。它有效解决了传统换脸技术流程繁琐、对硬件配置要求极高以及难以实时预览的痛点，让高质量的数字内容创作变得触手可及。\n\n这款工具不仅适合开发者和技术研究人员探索算法边界，更因其极简的操作逻辑（仅需三步：选脸、选摄像头、启动），广泛适用于普通用户、内容创作者、设计师及直播主播。无论是为了动画角色定制、服装展示模特替换，还是制作趣味短视频和直播互动，Deep-Live-Cam 都能提供流畅的支持。\n\n其核心技术亮点在于强大的实时处理能力，支持口型遮罩（Mouth Mask）以保留使用者原始的嘴部动作，确保表情自然精准；同时具备“人脸映射”功能，可同时对画面中的多个主体应用不同面孔。此外，项目内置了严格的内容安全过滤机制，自动拦截涉及裸露、暴力等不当素材，并倡导用户在获得授权及明确标注的前提下合规使用，体现了技术发展与伦理责任的平衡。",88924,"2026-04-06T03:28:53",[14,15,13,60],"视频",{"id":62,"github_repo":63,"name":64,"description_en":65,"description_zh":66,"ai_summary_zh":66,"readme_en":67,"readme_zh":68,"quickstart_zh":69,"use_case_zh":70,"hero_image_url":71,"owner_login":72,"owner_name":73,"owner_avatar_url":74,"owner_bio":75,"owner_company":76,"owner_location":76,"owner_email":76,"owner_twitter":76,"owner_website":77,"owner_url":78,"languages":79,"stars":84,"forks":85,"last_commit_at":86,"license":87,"difficulty_score":32,"env_os":88,"env_gpu":88,"env_ram":88,"env_deps":89,"category_tags":98,"github_topics":76,"view_count":32,"oss_zip_url":76,"oss_zip_packed_at":76,"status":17,"created_at":99,"updated_at":100,"faqs":101,"releases":102},4505,"langtalks\u002Fswe-agent","swe-agent","🤖 AI-powered software engineering multi-agent system with researcher and developer agents that automate code implementation through intelligent planning and execution. Built with LangGraph multi-agent workflows","swe-agent 是一款基于 LangGraph 构建的 AI 驱动软件工程多智能体系统，旨在通过智能化的规划与执行自动化完成代码实现任务。它主要解决了传统 AI 编程工具在复杂任务中缺乏整体规划、容易出错或难以理解大型代码库结构的问题。\n\n该系统采用独特的双阶段工作流：首先由“架构师智能体”深入研究代码库结构，分析需求并制定包含原子任务的详细实施计划；随后由“开发者智能体”严格按计划执行，精准地进行文件修改或新建，确保每一步变更都安全可靠。这种将“思考规划”与“动手执行”分离的机制，显著提升了代码生成的准确性和可维护性。\n\n技术亮点方面，swe-agent 结合了 tree-sitter 进行高级代码分析，并利用语义搜索深入理解项目上下文，同时通过 Pydantic 模型实现了严格的层级状态管理，保障了多智能体协作时的数据一致性与类型安全。\n\n这款工具特别适合希望提升开发效率的软件工程师、需要处理复杂重构任务的技术团队，以及对 AI 辅助编程工作流感兴趣的研究人员。目前项目处于 Alpha 阶段，非常适合愿意尝试前沿技术并参与共建的早期采用者。","\n# SWE Agent with LangGraph - by [LangTalks](https:\u002F\u002Flangtalks.ai)\n\n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flangtalks_swe-agent_readme_6a0f10a75f10.png\" width=\"400\" alt=\"Cover\">\n\n\n![Alpha](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fstatus-alpha-orange) ![Python](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fpython-3.12+-blue) ![License](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Flicense-MIT-green)\n\nA sophisticated AI-powered software engineering agent that automates code implementation through intelligent planning and execution. Built with LangGraph for reliable multi-agent workflows.\n\n> ⚠️ **Alpha Status**: This project is in active development. Features may change and some functionality is experimental. Perfect for early adopters and contributors who want to shape the future of AI-powered development.\n\n![Main Agent](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flangtalks_swe-agent_readme_9c69867ad957.png)\n\n[end to end showcase](https:\u002F\u002Fyoutu.be\u002FvJNqAgLzOSg)\n\n## 🚀 Features\n\n- **Intelligent Code Planning**: AI architect analyzes requirements and creates detailed implementation plans\n- **Automated Code Generation**: Developer agent executes plans with precise file modifications\n- **Multi-Agent Workflow**: Separate planning and implementation phases for better reliability\n- **Codebase Understanding**: Advanced code analysis using tree-sitter and semantic search\n- **Incremental Development**: Atomic task breakdown for safer, more manageable changes\n\n## 🏗️ Architecture\n\nThe system uses a two-stage LangGraph workflow:\n\n### 1. Architect Agent - Research & Planning\n![Architect Agent](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flangtalks_swe-agent_readme_442ca8984c6b.png)\n\nThe architect agent:\n- Researches the codebase structure and patterns\n- Analyzes requirements and creates hypotheses\n- Generates detailed implementation plans with atomic tasks\n- Uses tools for code search and semantic understanding\n\n### 2. Developer Agent - Implementation\n![Developer Agent](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flangtalks_swe-agent_readme_be182bdb545d.png)\n\nThe developer agent:\n- Executes implementation plans step by step\n- Performs atomic code modifications with precision\n- Creates new files and modifies existing ones\n- Validates changes against the original requirements\n\n### Workflow Overview\n```\nUser Request → Architect (Research & Plan) → Developer (Implement) → Results\n```\n\n**Key Components:**\n- **State Management**: Structured data flow between agents using Pydantic models\n- **Tool Integration**: File system operations, code search, and structure analysis\n- **Research Pipeline**: Hypothesis-driven exploration of codebases\n- **Atomic Execution**: Granular tasks for reliable implementation\n\n## 🔄 Agent State Management\n\nThe system uses a hierarchical state management approach with Pydantic models for type safety and validation. Each agent maintains its own state while sharing common entities for seamless data flow.\n\n### Main Agent State (`AgentState`)\n\nThe top-level orchestrator state that manages the overall workflow:\n\n```python\nclass AgentState(BaseModel):\n    implementation_research_scratchpad: Annotated[list[AnyMessage], add_messages]\n    implementation_plan: Optional[ImplementationPlan] = None\n```\n\n**Fields:**\n- `implementation_research_scratchpad`: Message history from research and planning phase\n- `implementation_plan`: Structured plan created by architect agent for developer execution\n\n### Architect Agent State (`SoftwareArchitectState`)\n\nManages the research and planning phase with hypothesis-driven exploration:\n\n```python\nclass SoftwareArchitectState(BaseModel):\n    research_next_step: Optional[str] = None\n    implementation_plan: Optional[ImplementationPlan] = None  \n    implementation_research_scratchpad: Annotated[list[AnyMessage], add_messages] = []\n    is_valid_research_step: Optional[bool] = None\n```\n\n**Fields:**\n- `research_next_step`: Current hypothesis or research direction being explored\n- `implementation_plan`: Generated structured plan with atomic tasks\n- `implementation_research_scratchpad`: Research conversation history and tool outputs\n- `is_valid_research_step`: Validation flag for research hypothesis quality\n\n**Workflow:**\n1. Generate research hypothesis → Validate hypothesis → Conduct research → Extract implementation plan\n\n### Developer Agent State (`SoftwareDeveloperState`)\n\nHandles the step-by-step implementation of the architect's plan:\n\n```python\nclass SoftwareDeveloperState(BaseModel):\n    implementation_plan: Optional[ImplementationPlan] = None\n    current_task_idx: Optional[int] = 0\n    current_atomic_task_idx: Optional[int] = 0\n    diffs: Optional[Diffs] = None\n    atomic_implementation_research: Annotated[list[AnyMessage], add_messages_with_clear]\n    codebase_structure: Optional[str] = None\n    current_file_content: Optional[str] = None\n```\n\n**Fields:**\n- `implementation_plan`: Plan received from architect agent\n- `current_task_idx`: Index of current file-level task being implemented\n- `current_atomic_task_idx`: Index of current atomic change within the task\n- `diffs`: Generated code differences for precise file modifications\n- `atomic_implementation_research`: Research specific to current implementation step\n- `codebase_structure`: Current snapshot of target codebase structure\n- `current_file_content`: Contents of file being modified\n\n**Workflow:**\n1. Iterate through tasks → Research specific implementation → Generate diffs → Apply changes\n\n### Shared Entities\n\nThese Pydantic models provide the data contracts between agents:\n\n#### `ImplementationPlan`\n```python\nclass ImplementationPlan(BaseModel):\n    tasks: List[ImplementationTask]\n```\nThe top-level plan containing all file-level implementation tasks.\n\n#### `ImplementationTask`  \n```python\nclass ImplementationTask(BaseModel):\n    file_path: str                    # Target file for modifications\n    logical_task: str                 # High-level description of changes\n    atomic_tasks: List[AtomicTask]    # Granular modification steps\n```\nRepresents changes needed for a specific file.\n\n#### `AtomicTask`\n```python\nclass AtomicTask(BaseModel):\n    atomic_task: str            # Specific code modification instruction\n    additional_context: str     # Research context for this change\n```\nThe smallest unit of implementation - a single code change.\n\n#### `DiffTask` (Developer-specific)\n```python\nclass DiffTask(BaseModel):\n    original_code_snippet: str    # Exact code being replaced\n    task_description: str         # Detailed change instructions\n```\nPrecise diff instructions for code modifications.\n\n### State Flow Example\n\n```\nUser Request\n    ↓\nAgentState {research_scratchpad: [HumanMessage(\"Add auth\")]}\n    ↓\nSoftwareArchitectState {\n    research_next_step: \"Find existing auth patterns\",\n    implementation_research_scratchpad: [research_messages...]\n}\n    ↓ (after research)\nSoftwareArchitectState {\n    implementation_plan: ImplementationPlan([\n        ImplementationTask(\n            file_path: \"auth.py\",\n            atomic_tasks: [AtomicTask(\"Add User model\")]\n        )\n    ])\n}\n    ↓\nSoftwareDeveloperState {\n    implementation_plan: \u003Creceived_plan>,\n    current_task_idx: 0,\n    current_atomic_task_idx: 0,\n    current_file_content: \"# auth.py content\"\n}\n    ↓ (after implementation)\nFinal Result: Modified codebase\n```\n\n### State Management Benefits\n\n- **Type Safety**: Pydantic validation prevents state corruption\n- **Traceability**: Complete message history for debugging\n- **Resumability**: State can be persisted and resumed\n- **Modularity**: Each agent manages its own concerns\n- **Atomicity**: Granular task breakdown enables precise control\n\n## 📋 Prerequisites\n\n- Python 3.12+\n- uv (Python package manager)\n- Anthropic API key (Claude Sonnet 4)\n\n## ⚡ Quick Start\n\n1. **Clone the repository**\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Flangtalks\u002Fswe-agent-langgraph.git\ncd swe-agent-langgraph\n```\n\n2. **Set up environment**\n```bash\n# Install dependencies with uv\nuv sync\n\n# Create environment file\ncp .env.example .env.local\n# Add your Anthropic API key and langsmith to .env\n```\n\n3. **Clone a repo to .\u002Fworkspace_repo**\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fbrowser-use\u002Fbrowser-use .\u002Fworkspace_repo\n```\n\n4. **Run the agent**\n```bash\n# Activate environment\nsource .venv\u002Fbin\u002Factivate\n\n# Start LangGraph server\nlanggraph dev\n\n```\n\n4. **Example usage**\nInput:\n![Input](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flangtalks_swe-agent_readme_396f35b11d4e.png)\n>Enable the browser-use agent to accept multi-modal instructions by supporting image inputs (e.g., step1.png, step2.png) alongside text. This will improve the agent’s ability to interpret and follow ambiguous or unclear textual commands\n\nOutput: (browsing the workspace repo git)\n![Output](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flangtalks_swe-agent_readme_5ce5cf935e5d.png)\n\n\n## 🛠️ Development\n\n### Project Structure\n```\nagent\u002F\n├── architect\u002F          # Planning and research agent\n│   ├── graph.py       # Main architect workflow\n│   ├── state.py       # State definitions\n│   └── prompts\u002F       # Prompt templates\n├── developer\u002F          # Implementation agent  \n│   ├── graph.py       # Main developer workflow\n│   ├── state.py       # State definitions\n│   └── prompts\u002F       # Prompt templates\n├── common\u002F            # Shared entities and state\n│   └── entities.py    # Pydantic models\n└── tools\u002F             # File operations and search tools\n    ├── search.py      # Code search tools\n    ├── codemap.py     # Code analysis tools\n    └── write.py       # File operations\n\nworkspace_repo\u002F        # Target codebase for modifications\nscripts\u002F              # Utility scripts\nhelpers\u002F              # Prompt templates and utilities\nstatic\u002F               # Documentation images\n```\n\n### Core Components\n\n**Entities & State Management:**\n- `ImplementationPlan`: Structured task breakdown\n- `AtomicTask`: Individual code modification units\n- `ImplementationTask`: File-level implementation steps\n\n**Agent Workflows:**\n- Research-driven planning with hypothesis validation\n- Tool-assisted code exploration and analysis\n- Incremental implementation with verification\n\n### Running Tests\n```bash\n# Run all tests\nuv run pytest\n\n# Run with coverage\nuv run pytest --cov=agent\n\n# Run specific test modules\nuv run pytest tests\u002Ftest_architect.py\n```\n\n## 📁 Main Directory Files\n\n| File | Description |\n|------|-------------|\n| `README.md` | Project documentation (this file) |\n| `pyproject.toml` | Python project configuration and dependencies |\n| `langgraph.json` | LangGraph application configuration with graph definitions |\n| `langgraph_debug.py` | Debug configurations for development and testing |\n| `uv.lock` | Locked dependency versions for reproducible builds |\n| `.env` | Environment variables (create from .env.example) |\n| `.env.example` | Template for environment configuration |\n| `.gitignore` | Git ignore patterns for Python and IDE files |\n| `.python-version` | Python version specification for pyenv |\n\n## 🎯 Use Cases\n\n- **Feature Development**: Implement new features based on high-level requirements\n- **Bug Fixes**: Analyze and fix issues with automated code changes\n- **Code Refactoring**: Restructure code while maintaining functionality\n- **Documentation**: Generate and update code documentation\n- **Testing**: Create test cases and fix failing tests\n\n## 🗺️ Roadmap - LangTalks Community Project!\n\nWe're building the future of AI-powered software development together! These are the next major features we're looking for community contributions on:\n\n### 🔄 Core Agent Enhancements\n- [ ] **Multi-step Research & Development Loop**: Iterative refinement of implementation plans with feedback cycles\n- [ ] **Testing Agent**: Dedicated agent for unit testing, functional testing, and test case generation  \n- [ ] **Error Fixer Agent**: Specialized agent for detecting, analyzing, and fixing code errors\n- [ ] **Product Manager Agent**: High-level planning and requirement analysis agent\n\n### 🔧 Development Tools & Quality\n- [ ] **Add Linters**: Integrate code quality tools (ESLint, Black, Pylint) into the workflow\n- [ ] **Components Evaluation Benchmarking**: Performance metrics and quality assessment frameworks\n- [ ] **Code Semantic Indexing**: Advanced code understanding and similarity detection\n\n### 🌐 Integrations & Connectivity  \n- [ ] **GitHub MCP Integration**: Direct integration with GitHub repositories and workflows\n- [ ] **Context7 MCP Integration**: Enhanced context management and code understanding\n- [ ] **Multi-Language Support**: Extend beyond Python to JavaScript, TypeScript, Java, Go, etc.\n\n### 📈 Advanced Features (Future)\n- [ ] **Interactive Planning UI**: Web interface for plan review and modification\n- [ ] **Collaborative Workflows**: Multi-developer coordination and conflict resolution\n- [ ] **Performance Optimization**: Faster research and implementation cycles\n- [ ] **Plugin System**: Extensible tool and agent architecture\n\n> **Want to contribute?** Pick any feature above and join our LangTalks community! Each feature is designed to be tackled by individual contributors or small teams.\n\n## 🤝 Contributing\n\nWe welcome contributions! This project aims to push the boundaries of AI-powered software development. Areas where we need help:\n\n### Priority Areas\n- **Agent Improvements**: Better reasoning and planning strategies\n- **Tool Development**: New code analysis and modification tools\n- **Testing**: Comprehensive test coverage and validation frameworks\n- **Documentation**: Examples, tutorials, and use cases\n- **Performance**: Optimization and benchmarking\n\n### How to Contribute\n\n1. **Fork the repository**\n2. **Create a feature branch** (`git checkout -b feature\u002Famazing-feature`)\n3. **Make your changes** following the existing code patterns\n4. **Add tests** for new functionality\n5. **Ensure tests pass** (`uv run pytest`)\n6. **Update documentation** if needed\n7. **Commit your changes** (`git commit -m 'Add amazing feature'`)\n8. **Push to the branch** (`git push origin feature\u002Famazing-feature`)\n9. **Open a Pull Request** with a clear description\n\n### Development Setup\n\n```bash\n# Clone your fork\ngit clone https:\u002F\u002Fgithub.com\u002Flangtalks\u002Fswe-agent-langgraph.git\ncd swe-agent-langgraph\n\n# Set up development environment\nuv sync --dev\n\n# Install pre-commit hooks (optional but recommended)\npre-commit install\n\n# Run tests to ensure everything works\nuv run pytest\n```\n\n## 📊 Technical Details\n\n### Dependencies\n- **LangGraph**: Multi-agent workflow orchestration\n- **LangChain**: AI integration and tool management\n- **Anthropic**: Claude Sonnet 4 for intelligent reasoning\n- **Tree-sitter**: Robust code parsing and analysis\n- **Pydantic**: Type-safe data validation and serialization\n\n### Performance Considerations\n- Atomic task execution for reliability\n- Efficient code analysis with tree-sitter\n- Structured state management for scalability\n- Tool-based architecture for extensibility\n\n## 🔧 Configuration\n\nKey configuration files:\n- `langgraph.json`: Defines agent graphs and dependencies\n- `.env`: API keys and environment variables\n- `pyproject.toml`: Python dependencies and project metadata\n\n## 📄 License\n\nThis project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.\n\n## 🙏 Acknowledgments\n\n- Built with [LangGraph](https:\u002F\u002Flangchain-ai.github.io\u002Flanggraph\u002F) for reliable agent workflows\n- Powered by [Anthropic Claude](https:\u002F\u002Fwww.anthropic.com\u002F) for intelligent reasoning\n- Uses [tree-sitter](https:\u002F\u002Ftree-sitter.github.io\u002F) for robust code parsing\n- See our [deepwiki](https:\u002F\u002Fdeepwiki.com\u002Flangtalks\u002Fswe-agent\u002F1-overview)\n\n## 📞 Support & Community\n\n- **LangTalks Homepage**: Visit [www.langtalks.ai](https:\u002F\u002Fwww.langtalks.ai) for community resources and support\n- **Issues**: Report bugs and request features via [GitHub Issues](https:\u002F\u002Fgithub.com\u002Flangtalks\u002Fswe-agent-langgraph\u002Fissues)\n- **Discussions**: Join conversations in [GitHub Discussions](https:\u002F\u002Fgithub.com\u002Flangtalks\u002Fswe-agent-langgraph\u002Fdiscussions)\n- **Documentation**: Complete documentation is available in this README\n\n---\n\n**Ready to revolutionize software development with AI? Join us at [LangTalks](https:\u002F\u002Fwww.langtalks.ai) and help build the future of automated coding!** ⚡🤖\n","# 使用 LangGraph 的软件工程代理 - 由 [LangTalks](https:\u002F\u002Flangtalks.ai) 提供\n\n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flangtalks_swe-agent_readme_6a0f10a75f10.png\" width=\"400\" alt=\"封面\">\n\n\n![Alpha](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fstatus-alpha-orange) ![Python](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fpython-3.12+-blue) ![License](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Flicense-MIT-green)\n\n一款基于人工智能的先进软件工程代理，通过智能规划与执行自动化代码实现。采用 LangGraph 构建，确保多代理工作流的可靠性。\n\n> ⚠️ **Alpha 状态**：本项目正处于积极开发中。功能可能会发生变化，部分功能仍处于实验阶段。非常适合希望参与塑造 AI 驱动开发未来的早期采用者和贡献者。\n\n![主代理](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flangtalks_swe-agent_readme_9c69867ad957.png)\n\n[端到端展示](https:\u002F\u002Fyoutu.be\u002FvJNqAgLzOSg)\n\n## 🚀 功能特性\n\n- **智能代码规划**：AI 架构师分析需求并创建详细的实施方案\n- **自动化代码生成**：开发者代理根据计划精确地进行文件修改\n- **多代理工作流**：分离规划与实施阶段，提升可靠性\n- **代码库理解**：利用 tree-sitter 和语义搜索进行高级代码分析\n- **增量式开发**：将任务分解为原子级，使更改更安全、更易管理\n\n## 🏗️ 架构设计\n\n系统采用两阶段的 LangGraph 工作流：\n\n### 1. 架构师代理 - 研究与规划\n![架构师代理](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flangtalks_swe-agent_readme_442ca8984c6b.png)\n\n架构师代理负责：\n- 研究代码库结构与模式\n- 分析需求并提出假设\n- 生成包含原子任务的详细实施方案\n- 利用工具进行代码搜索和语义理解\n\n### 2. 开发者代理 - 实施\n![开发者代理](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flangtalks_swe-agent_readme_be182bdb545d.png)\n\n开发者代理负责：\n- 按步骤执行实施方案\n- 精确地进行原子级别的代码修改\n- 创建新文件或修改现有文件\n- 根据原始需求验证更改结果\n\n### 工作流概述\n```\n用户请求 → 架构师（研究与规划）→ 开发者（实施）→ 结果\n```\n\n**关键组件：**\n- **状态管理**：使用 Pydantic 模型在各代理间实现结构化的数据流转\n- **工具集成**：文件系统操作、代码搜索及结构分析\n- **研究流程**：基于假设驱动的代码库探索\n- **原子级执行**：细粒度任务确保可靠的实施过程\n\n## 🔄 代理状态管理\n\n系统采用分层的状态管理方法，借助 Pydantic 模型实现类型安全与验证。每个代理维护自己的状态，同时共享公共实体以实现无缝的数据流动。\n\n### 主代理状态 (`AgentState`)\n\n顶层协调器状态，用于管理整个工作流：\n\n```python\nclass AgentState(BaseModel):\n    implementation_research_scratchpad: Annotated[list[AnyMessage], add_messages]\n    implementation_plan: Optional[ImplementationPlan] = None\n```\n\n**字段：**\n- `implementation_research_scratchpad`：研究与规划阶段的消息记录\n- `implementation_plan`：由架构师代理创建的待开发者执行的结构化方案\n\n### 架构师代理状态 (`SoftwareArchitectState`)\n\n管理基于假设驱动的研究与规划阶段：\n\n```python\nclass SoftwareArchitectState(BaseModel):\n    research_next_step: Optional[str] = None\n    implementation_plan: Optional[ImplementationPlan] = 0  \n    implementation_research_scratchpad: Annotated[list[AnyMessage], add_messages] = []\n    is_valid_research_step: Optional[bool] = None\n```\n\n**字段：**\n- `research_next_step`：当前正在探索的假设或研究方向\n- `implementation_plan`：生成的包含原子任务的结构化方案\n- `implementation_research_scratchpad`：研究对话历史及工具输出\n- `is_valid_research_step`：用于验证研究假设质量的标志位\n\n**工作流：**\n1. 生成研究假设 → 验证假设 → 进行研究 → 提取实施方案\n\n### 开发者代理状态 (`SoftwareDeveloperState`)\n\n负责逐步执行架构师的方案：\n\n```python\nclass SoftwareDeveloperState(BaseModel):\n    implementation_plan: Optional[ImplementationPlan] = None\n    current_task_idx: Optional[int] = 0\n    current_atomic_task_idx: Optional[int] = 0\n    diffs: Optional[Diffs] = None\n    atomic_implementation_research: Annotated[list[AnyMessage], add_messages_with_clear]\n    codebase_structure: Optional[str] = None\n    current_file_content: Optional[str] = None\n```\n\n**字段：**\n- `implementation_plan`：从架构师代理接收到的方案\n- `current_task_idx`：当前正在实施的文件级任务索引\n- `current_atomic_task_idx`：该任务中的原子级变更索引\n- `diffs`：用于精确文件修改的代码差异\n- `atomic_implementation_research`：针对当前实施步骤的研究记录\n- `codebase_structure`：目标代码库的当前结构快照\n- `current_file_content`：正在修改的文件内容\n\n**工作流：**\n1. 遍历任务列表 → 针对具体实施进行研究 → 生成差异 → 应用更改\n\n### 共享实体\n\n这些 Pydantic 模型定义了各代理之间的数据契约：\n\n#### `ImplementationPlan`\n```python\nclass ImplementationPlan(BaseModel):\n    tasks: List[ImplementationTask]\n```\n顶层方案，包含所有文件级的实施任务。\n\n#### `ImplementationTask`\n```python\nclass ImplementationTask(BaseModel):\n    file_path: str                    # 待修改的目标文件\n    logical_task: str                 # 更改的高层次描述\n    atomic_tasks: List[AtomicTask]    # 细粒度的修改步骤\n```\n表示针对特定文件所需的更改。\n\n#### `AtomicTask`\n```python\nclass AtomicTask(BaseModel):\n    atomic_task: str            # 具体的代码修改指令\n    additional_context: str     # 此次更改的研究背景\n```\n最小的实施单元——单个代码变更。\n\n#### `DiffTask`（开发者专用）\n```python\nclass DiffTask(BaseModel):\n    original_code_snippet: str    # 被替换的确切代码片段\n    task_description: str         # 详细的变更说明\n```\n用于代码修改的精确差异指令。\n\n### 状态流转示例\n\n```\n用户请求\n    ↓\nAgentState {research_scratchpad: [HumanMessage(\"添加认证功能\")]}\n    ↓\nSoftwareArchitectState {\n    research_next_step: \"查找现有的认证模式\",\n    implementation_research_scratchpad: [research_messages...]\n}\n    ↓ （研究完成后）\nSoftwareArchitectState {\n    implementation_plan: ImplementationPlan([\n        ImplementationTask(\n            file_path: \"auth.py\",\n            atomic_tasks: [AtomicTask(\"添加User模型\")]\n        )\n    ])\n}\n    ↓\nSoftwareDeveloperState {\n    implementation_plan: \u003C接收到的计划>,\n    current_task_idx: 0,\n    current_atomic_task_idx: 0,\n    current_file_content: \"# auth.py 的内容\"\n}\n    ↓ （实现完成后）\n最终结果：修改后的代码库\n```\n\n### 状态管理的优势\n\n- **类型安全**：通过 Pydantic 验证防止状态被破坏\n- **可追溯性**：完整的消息历史便于调试\n- **可恢复性**：状态可以持久化并恢复\n- **模块化**：每个代理管理自己的职责范围\n- **原子性**：细粒度的任务分解实现精确控制\n\n## 📋 前置条件\n\n- Python 3.12+\n- uv（Python 包管理器）\n- Anthropic API 密钥（Claude Sonnet 4）\n\n## ⚡ 快速入门\n\n1. **克隆仓库**\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Flangtalks\u002Fswe-agent-langgraph.git\ncd swe-agent-langgraph\n```\n\n2. **设置环境**\n```bash\n# 使用 uv 安装依赖\nuv sync\n\n# 创建环境文件\ncp .env.example .env.local\n# 在 .env 中添加你的 Anthropic API 密钥和 langsmith 信息\n```\n\n3. **克隆一个仓库到 .\u002Fworkspace_repo**\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fbrowser-use\u002Fbrowser-use .\u002Fworkspace_repo\n```\n\n4. **运行代理**\n```bash\n# 激活环境\nsource .venv\u002Fbin\u002Factivate\n\n# 启动 LangGraph 服务器\nlanggraph dev\n\n```\n\n4. **示例用法**\n输入：\n![Input](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flangtalks_swe-agent_readme_396f35b11d4e.png)\n>使 browser-use 代理能够接受多模态指令，支持图像输入（如 step1.png、step2.png）与文本一起使用。这将提高代理对模糊或不明确文本命令的理解和执行能力。\n\n输出：（浏览工作区仓库的 git）\n![Output](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flangtalks_swe-agent_readme_5ce5cf935e5d.png)\n\n\n## 🛠️ 开发\n\n### 项目结构\n```\nagent\u002F\n├── architect\u002F          # 规划与研究代理\n│   ├── graph.py       # 主架构师工作流\n│   ├── state.py       # 状态定义\n│   └── prompts\u002F       # 提示模板\n├── developer\u002F          # 实现代理  \n│   ├── graph.py       # 主开发者工作流\n│   ├── state.py       # 状态定义\n│   └── prompts\u002F       # 提示模板\n├── common\u002F            # 共享实体和状态\n│   └── entities.py    # Pydantic 模型\n└── tools\u002F             # 文件操作和搜索工具\n    ├── search.py      # 代码搜索工具\n    ├── codemap.py     # 代码分析工具\n    └── write.py       # 文件操作工具\n\nworkspace_repo\u002F        # 目标代码库\nscripts\u002F              # 工具脚本\nhelpers\u002F              # 提示模板和实用工具\nstatic\u002F               # 文档图片\n```\n\n### 核心组件\n\n**实体与状态管理：**\n- `ImplementationPlan`：结构化的任务分解\n- `AtomicTask`：单个代码修改单元\n- `ImplementationTask`：文件级别的实现步骤\n\n**代理工作流：**\n- 基于研究的规划，并验证假设\n- 使用工具辅助进行代码探索和分析\n- 逐步实施并进行验证\n\n### 运行测试\n```bash\n# 运行所有测试\nuv run pytest\n\n# 带覆盖率运行\nuv run pytest --cov=agent\n\n# 运行特定测试模块\nuv run pytest tests\u002Ftest_architect.py\n```\n\n## 📁 主目录文件\n\n| 文件 | 描述 |\n|------|-------------|\n| `README.md` | 项目文档（本文档） |\n| `pyproject.toml` | Python 项目配置及依赖 |\n| `langgraph.json` | LangGraph 应用程序配置及图定义 |\n| `langgraph_debug.py` | 用于开发和测试的调试配置 |\n| `uv.lock` | 锁定的依赖版本，确保构建可重复性 |\n| `.env` | 环境变量（根据 .env.example 创建） |\n| `.env.example` | 环境配置模板 |\n| `.gitignore` | Git 忽略规则，用于 Python 和 IDE 文件 |\n| `.python-version` | pyenv 的 Python 版本规范 |\n\n## 🎯 使用场景\n\n- **功能开发**：根据高层次需求实现新功能\n- **Bug 修复**：分析并修复问题，自动完成代码更改\n- **代码重构**：在保持功能不变的情况下重组代码\n- **文档编写**：生成和更新代码文档\n- **测试**：创建测试用例并修复失败的测试\n\n## 🗺️ 路线图 - LangTalks 社区项目！\n\n我们正在共同构建 AI 驱动软件开发的未来！以下是接下来的主要功能，期待社区贡献：\n\n### 🔄 核心代理增强\n- [ ] **多步研究与开发循环**：通过反馈循环迭代优化实施计划\n- [ ] **测试代理**：专门用于单元测试、功能测试和测试用例生成的代理\n- [ ] **错误修复代理**：专门用于检测、分析和修复代码错误的代理\n- [ ] **产品经理代理**：负责高层次规划和需求分析的代理\n\n### 🔧 开发工具与质量\n- [ ] **添加 linter 工具**：将代码质量工具（ESLint、Black、Pylint）集成到工作流中\n- [ ] **组件评估基准测试**：性能指标和质量评估框架\n- [ ] **代码语义索引**：高级代码理解和相似度检测\n\n### 🌐 集成与连接\n- [ ] **GitHub MCP 集成**：直接与 GitHub 仓库和工作流集成\n- [ ] **Context7 MCP 集成**：增强上下文管理和代码理解能力\n- [ ] **多语言支持**：从 Python 扩展到 JavaScript、TypeScript、Java、Go 等\n\n### 📈 高级功能（未来）\n- [ ] **交互式规划 UI**：用于查看和修改计划的 Web 界面\n- [ ] **协作工作流**：多开发者协调与冲突解决\n- [ ] **性能优化**：加快研究和实施周期\n- [ ] **插件系统**：可扩展的工具和代理架构\n\n> **想参与贡献吗？** 选择上述任意功能加入我们的 LangTalks 社区吧！每个功能都适合个人贡献者或小型团队完成。\n\n## 🤝 贡献方式\n\n我们欢迎任何形式的贡献！该项目旨在推动 AI 驱动软件开发的边界。我们需要帮助的领域包括：\n\n### 优先领域\n- **代理改进**：更好的推理和规划策略\n- **工具开发**：新的代码分析和修改工具\n- **测试**：全面的测试覆盖和验证框架\n- **文档**：示例、教程和使用案例\n- **性能**：优化和基准测试\n\n### 如何贡献\n\n1. **Fork 仓库**\n2. **创建功能分支**（`git checkout -b feature\u002Famazing-feature`）\n3. **按照现有代码风格进行修改** \n4. **为新功能添加测试**\n5. **确保测试通过**（`uv run pytest`）\n6. **如有需要，更新文档**\n7. **提交更改**（`git commit -m '添加超赞功能'`）\n8. **推送到分支**（`git push origin feature\u002Famazing-feature`）\n9. **打开带有清晰描述的 Pull Request**\n\n### 开发环境搭建\n\n```bash\n# 克隆你的 Fork\ngit clone https:\u002F\u002Fgithub.com\u002Flangtalks\u002Fswe-agent-langgraph.git\ncd swe-agent-langgraph\n\n# 设置开发环境\nuv sync --dev\n\n# 安装 pre-commit 钩子（可选但推荐）\npre-commit install\n\n# 运行测试以确保一切正常\nuv run pytest\n```\n\n## 📊 技术细节\n\n### 依赖项\n- **LangGraph**: 多智能体工作流编排\n- **LangChain**: AI 集成与工具管理\n- **Anthropic**: Claude Sonnet 4 提供智能推理能力\n- **Tree-sitter**: 强大的代码解析与分析工具\n- **Pydantic**: 类型安全的数据验证与序列化\n\n### 性能考量\n- 原子化任务执行以保证可靠性\n- 使用 tree-sitter 进行高效代码分析\n- 结构化状态管理以提升可扩展性\n- 基于工具的架构设计以增强可扩展性\n\n## 🔧 配置\n\n关键配置文件：\n- `langgraph.json`: 定义智能体图及依赖关系\n- `.env`: API 密钥及其他环境变量\n- `pyproject.toml`: Python 依赖与项目元数据\n\n## 📄 许可证\n\n本项目采用 MIT 许可证授权——详情请参阅 [LICENSE](LICENSE) 文件。\n\n## 🙏 致谢\n\n- 基于 [LangGraph](https:\u002F\u002Flangchain-ai.github.io\u002Flanggraph\u002F) 构建，提供可靠的任务流支持\n- 由 [Anthropic Claude](https:\u002F\u002Fwww.anthropic.com\u002F) 提供智能推理能力\n- 使用 [tree-sitter](https:\u002F\u002Ftree-sitter.github.io\u002F) 实现稳健的代码解析\n- 更多信息请参阅我们的 [deepwiki](https:\u002F\u002Fdeepwiki.com\u002Flangtalks\u002Fswe-agent\u002F1-overview)\n\n## 📞 支持与社区\n\n- **LangTalks 主页**: 访问 [www.langtalks.ai](https:\u002F\u002Fwww.langtalks.ai) 获取社区资源与支持\n- **问题**: 通过 [GitHub Issues](https:\u002F\u002Fgithub.com\u002Flangtalks\u002Fswe-agent-langgraph\u002Fissues) 报告 bug 或请求功能\n- **讨论**: 加入 [GitHub Discussions](https:\u002F\u002Fgithub.com\u002Flangtalks\u002Fswe-agent-langgraph\u002Fdiscussions) 参与交流\n- **文档**: 完整文档见本 README\n\n---\n\n**准备好用 AI 彻底革新软件开发了吗？欢迎加入 [LangTalks](https:\u002F\u002Fwww.langtalks.ai)，共同构建自动化编码的未来！** ⚡🤖","# SWE-Agent 快速上手指南\n\nSWE-Agent 是一款基于 LangGraph 构建的高级 AI 软件工程代理，能够通过智能规划和执行自动化代码实现。它采用“架构师 + 开发者”的双代理工作流，将需求分析、代码规划与具体实施分离，以提高代码生成的准确性和可靠性。\n\n## 环境准备\n\n在开始之前，请确保您的系统满足以下要求：\n\n*   **操作系统**: Linux, macOS 或 Windows (WSL2 推荐)\n*   **Python 版本**: 3.12 或更高版本\n*   **包管理器**: `uv` (推荐的现代 Python 包管理器)\n*   **API 密钥**: Anthropic API Key (需支持 Claude Sonnet 模型)\n*   **Git**: 用于克隆目标代码库\n\n> **注意**：本项目目前处于 **Alpha** 阶段，功能可能随时变更，适合早期采用者和贡献者体验。\n\n## 安装步骤\n\n### 1. 克隆项目仓库\n首先获取 SWE-Agent 的源代码：\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Flangtalks\u002Fswe-agent-langgraph.git\ncd swe-agent-langgraph\n```\n\n### 2. 安装依赖\n使用 `uv` 同步安装项目所需的所有依赖包：\n\n```bash\nuv sync\n```\n\n*(国内用户若遇到网络问题，可尝试配置 uv 使用国内镜像源，例如：`export UV_INDEX_URL=https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple`)*\n\n### 3. 配置环境变量\n创建本地环境配置文件并填入必要的 API 密钥：\n\n```bash\ncp .env.example .env.local\n```\n\n编辑 `.env.local` 文件，填入您的 Anthropic API Key 和 LangSmith 配置（如需）：\n\n```bash\n# .env.local 内容示例\nANTHROPIC_API_KEY=your_anthropic_api_key_here\nLANGSMITH_API_KEY=your_langsmith_api_key_here # 可选\nLANGSMITH_TRACING=true\n```\n\n### 4. 准备目标代码库\nSWE-Agent 需要在一个具体的代码库中工作。创建一个工作目录并克隆您想要修改的项目（示例中使用 `browser-use`）：\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fbrowser-use\u002Fbrowser-use .\u002Fworkspace_repo\n```\n\n## 基本使用\n\n### 1. 激活环境并启动服务\n激活虚拟环境并启动 LangGraph 开发服务器：\n\n```bash\nsource .venv\u002Fbin\u002Factivate\nlanggraph dev\n```\n\n启动成功后，终端会显示本地服务地址（通常为 `http:\u002F\u002Flocalhost:8123`），LangGraph Studio 界面将自动在浏览器中打开。\n\n### 2. 提交任务\n在 LangGraph Studio 界面中，输入您的需求指令。SWE-Agent 会自动协调“架构师”进行研究和规划，然后指挥“开发者”执行代码修改。\n\n**输入示例：**\n> Enable the browser-use agent to accept multi-modal instructions by supporting image inputs (e.g., step1.png, step2.png) alongside text. This will improve the agent's ability to interpret and follow ambiguous or unclear textual commands.\n\n### 3. 查看结果\n代理执行完成后，您可以在界面中查看：\n*   **Architect Agent** 生成的详细实施计划。\n*   **Developer Agent** 生成的代码差异（Diffs）。\n*   最终修改后的文件内容已自动应用到 `.\u002Fworkspace_repo` 目录中。\n\n您可以直接在本地 IDE 中检查 `.\u002Fworkspace_repo` 下的代码变更，验证运行结果。","某电商初创团队的后端工程师急需在现有的 Python 订单系统中集成“超时自动取消”功能，但面对数万行缺乏文档的遗留代码，他担心手动修改会引发未知的连锁故障。\n\n### 没有 swe-agent 时\n- **理解成本极高**：工程师需花费数天人工梳理复杂的调用链和数据库事务逻辑，才能确定修改位置。\n- **实施风险巨大**：手动编辑核心文件时，极易因遗漏边缘情况导致订单状态不一致或数据丢失。\n- **测试覆盖不足**：由于害怕破坏现有功能，往往只能进行简单的局部测试，难以验证全局影响。\n- **迭代效率低下**：从需求分析到最终上线，整个流程耗时漫长，严重拖慢产品迭代节奏。\n\n### 使用 swe-agent 后\n- **智能规划先行**：Architect Agent 自动扫描代码库，利用语义搜索快速定位关键模块，并生成包含原子任务的详细实施计划。\n- **精准原子执行**：Developer Agent 严格按照计划逐步执行，对文件进行细粒度的精确修改，避免了对无关代码的误触。\n- **闭环验证保障**：系统在每次修改后自动对照原始需求进行逻辑校验，确保新功能与旧逻辑完美兼容。\n- **交付速度倍增**：将原本需要数天的分析与编码工作压缩至小时级，让工程师能专注于更高阶的架构优化。\n\nswe-agent 通过“先规划后执行”的双代理协作模式，将高风险的遗留系统改造转化为安全、可控且高效的自动化流程。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flangtalks_swe-agent_d105236e.png","langtalks","LangTalks","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Flangtalks_a7743aef.png","",null,"www.langtalks.ai","https:\u002F\u002Fgithub.com\u002Flangtalks",[80],{"name":81,"color":82,"percentage":83},"Python","#3572A5",100,623,123,"2026-04-01T04:41:07","MIT","未说明",{"notes":90,"python":91,"dependencies":92},"需要使用 uv 作为 Python 包管理器；必须配置 Anthropic API Key（使用 Claude Sonnet 4 模型）；可选配置 LangSmith 用于追踪；运行前需克隆一个目标代码库到 .\u002Fworkspace_repo 目录。","3.12+",[93,94,95,96,97],"uv","langgraph","pydantic","tree-sitter","pytest",[35,13],"2026-03-27T02:49:30.150509","2026-04-07T00:48:23.997992",[],[]]