[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-magnitudedev--browser-agent":3,"tool-magnitudedev--browser-agent":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",140436,2,"2026-04-05T23:32:43",[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":80,"owner_website":81,"owner_url":82,"languages":83,"stars":104,"forks":105,"last_commit_at":106,"license":107,"difficulty_score":23,"env_os":108,"env_gpu":109,"env_ram":108,"env_deps":110,"category_tags":117,"github_topics":118,"view_count":23,"oss_zip_url":79,"oss_zip_packed_at":79,"status":16,"created_at":129,"updated_at":130,"faqs":131,"releases":162},2658,"magnitudedev\u002Fbrowser-agent","browser-agent","Open-source, vision-first browser agent","browser-agent（Magnitude）是一款开源的“视觉优先”浏览器智能体，旨在让用户通过自然语言直接控制浏览器。它能像人类一样“看”懂网页界面，自主规划并执行点击、拖拽等操作，还能智能提取结构化数据或运行可视化测试，广泛应用于网页自动化、无 API 应用集成及端到端测试场景。\n\n传统浏览器自动化工具常依赖给页面元素标注数字框来定位，这在结构复杂的现代网站中往往失效。browser-agent 独创性地采用视觉优先架构，利用强大的多模态大模型直接分析屏幕像素坐标进行操作。这种机制不依赖底层 DOM 结构，具备极强的通用性和未来适应性，能更稳定地应对各类复杂界面。\n\n该工具特别适合开发者、测试工程师及研究人员使用。开发者可将其作为构建自定义智能体的基础模块，快速实现高阶任务编排；测试人员则能利用其内置的测试运行器，编写具备强大视觉断言能力的自动化测试脚本。虽然普通用户也能受益，但其当前的代码集成方式更偏向技术人群。需要注意的是，为了获得最佳效果，建议搭配 Claude Sonnet 4 等具备强视觉理解能力的大模型使用。只需几条命令即可初始化项目，让浏览器操作变得既直观又高效。","\u003Cdiv align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmagnitudedev_browser-agent_readme_6e520ed16e52.png\" alt=\"Magnitude Text Logo\"\u002F>\n\u003C\u002Fdiv>\n\n\u003Cbr\u002F>\n\n\u003Cdiv align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fdocs.magnitude.run\u002Fgetting-started\u002Fintroduction\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F📕-Docs-0369a1?style=flat-square&labelColor=0369a1&color=gray\" alt=\"Documentation\" \u002F>\u003C\u002Fa> \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-Apache%202.0-0369a1?style=flat-square&labelColor=0369a1&color=gray\" alt=\"License\" \u002F> \u003Ca href=\"https:\u002F\u002Fdiscord.gg\u002FVcdpMh9tTy\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fdiscord\u002F1305570963206836295?style=flat-square&logo=discord&logoColor=white&label=Discord&labelColor=5865F2&color=gray\" alt=\"Discord\" \u002F>\u003C\u002Fa> \u003Ca href=\"https:\u002F\u002Fx.com\u002Ftgrnwld\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F-Follow%20Tom!-000000?style=flat-square&labelColor=000000&color=gray&logo=x&logoColor=white\" alt=\"Follow Tom\" \u002F>\u003C\u002Fa> \u003Ca href=\"https:\u002F\u002Fx.com\u002Fndrsrkl\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F-Follow%20Anders!-000000?style=flat-square&labelColor=000000&color=gray&logo=x&logoColor=white\" alt=\"Follow Anders\" \u002F>\u003C\u002Fa>\n\u003C\u002Fdiv>\n\n\u003Chr\u002F>\n\n> 🚀 New: Magnitude is state-of-the-art, scoring [94% on WebVoyager](https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fwebvoyager)!\n\nMagnitude uses vision AI to enable you to control your browser with natural language.\n- 🧭 **Navigate** - Sees and understands any interface to plan out actions\n- 🖱️ **Interact** - Executes precise actions using mouse and keyboard\n- 🔍 **Extract** - Intelligently extracts useful structured data\n- ✅ **Verify** - Built-in test runner with powerful visual assertions\n\nYou can use it to automate tasks on the web, integrate between apps without APIs, extract data, test your web apps, or as a building block for your own browser agents.\n\n![Video showing Magnitude tests running in a terminal and agent taking actions in the browser](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmagnitudedev_browser-agent_readme_a495d3c7d484.gif)\n\n↕️ Magnitude in action! ↕️\n```ts\n\u002F\u002F Magnitude can handle high-level tasks\nawait agent.act('Create a task', {\n    \u002F\u002F Optionally pass data that the agent will use where appropriate\n    data: {\n        title: 'Use Magnitude',\n        description: 'Run \"npx create-magnitude-app\" and follow the instructions',\n    },\n});\n\n\u002F\u002F It can also handle low-level actions\nawait agent.act('Drag \"Use Magnitude\" to the top of the in progress column');\n\n\u002F\u002F Intelligently extract data based on the DOM content matching a provided zod schema\nconst tasks = await agent.extract(\n    'List in progress tasks',\n    z.array(z.object({\n        title: z.string(),\n        description: z.string(),\n        \u002F\u002F Agent can extract existing data or new insights\n        difficulty: z.number().describe('Rate the difficulty between 1-5')\n    })),\n);\n```\n\n## Get started\n\n### Running your first browser automation\n```bash\nnpx create-magnitude-app\n```\n\nThis will create a new project and walk you through the steps for setting up Magnitude. It will also create an example script that you can run right away!\n\n### Using the test runner\nTo install the test runner for use in an **existing** web app, please run:\n```bash\nnpm i --save-dev magnitude-test && npx magnitude init\n```\n\nThis will create a basic tests directory `tests\u002Fmagnitude` with:\n- `magnitude.config.ts`: Magnitude test configuration file\n- `example.mag.ts`: An example test file\n\nFor information on how to run tests and integrate into CI\u002FCD see [here](https:\u002F\u002Fdocs.magnitude.run\u002Fcore-concepts\u002Frunning-tests).\n\n> [!NOTE]\n> Magnitude requires a large **visually grounded** model. We recommend Claude Sonnet 4 for the best performance, but are also compatible with Qwen-2.5VL 72B. See [docs](https:\u002F\u002Fdocs.magnitude.run\u002Fcustomizing\u002Fllm-configuration) for more information.\n\n\n## Why Magnitude?\n❌ **Problem #1:** Most browser agents draw numbered boxes around page elements - doesn't generalize well due to complex modern sites\n\n✅ **Solution: Vision-first architecture**\n* Visually grounded LLM specifies pixel coordinates\n* True generalization independent of DOM structure\n* Future-proof architecture for desktop apps, VMs, etc.\n\n❌ **Problem #2:** Most browser agents follow \"high-level prompt + tools = work until done\" - works for demos, not production\n\n✅ **Solution: Controllable & repeatable automation**\n* Flexible abstraction levels (granular actions vs. flows)\n* Custom actions + prompts at agent and action level\n* Deterministic runs via native caching system *(in progress)*\n\n## Additional info\n\nPlease see [our docs](https:\u002F\u002Fdocs.magnitude.run) for more information on how to best build Magnitude automations and test cases.\n\n## Contact\nIf you are an enterprise and want more features or support, feel free to reach out to us at founders@magnitude.run or schedule a call [here](https:\u002F\u002Fcal.com\u002Ftom-greenwald\u002F30min) to discuss your needs.\n\nYou can also join our \u003Ca href=\"https:\u002F\u002Fdiscord.gg\u002FVcdpMh9tTy\" target=\"_blank\">Discord community\u003C\u002Fa> for help or any suggestions!\n","\u003Cdiv align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmagnitudedev_browser-agent_readme_6e520ed16e52.png\" alt=\"Magnitude 文本标志\"\u002F>\n\u003C\u002Fdiv>\n\n\u003Cbr\u002F>\n\n\u003Cdiv align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fdocs.magnitude.run\u002Fgetting-started\u002Fintroduction\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F📕-文档-0369a1?style=flat-square&labelColor=0369a1&color=gray\" alt=\"文档\" \u002F>\u003C\u002Fa> \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F许可证-Apache%202.0-0369a1?style=flat-square&labelColor=0369a1&color=gray\" alt=\"许可证\" \u002F> \u003Ca href=\"https:\u002F\u002Fdiscord.gg\u002FVcdpMh9tTy\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fdiscord\u002F1305570963206836295?style=flat-square&logo=discord&logoColor=white&label=Discord&labelColor=5865F2&color=gray\" alt=\"Discord\" \u002F>\u003C\u002Fa> \u003Ca href=\"https:\u002F\u002Fx.com\u002Ftgrnwld\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F-关注汤姆！-000000?style=flat-square&labelColor=000000&color=gray&logo=x&logoColor=white\" alt=\"关注汤姆\" \u002F>\u003C\u002Fa> \u003Ca href=\"https:\u002F\u002Fx.com\u002Fndrsrkl\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F-关注安德斯！-000000?style=flat-square&labelColor=000000&color=gray&logo=x&logoColor=white\" alt=\"关注安德斯\" \u002F>\u003C\u002Fa>\n\u003C\u002Fdiv>\n\n\u003Chr\u002F>\n\n> 🚀 新：Magnitude 处于行业领先水平，在 WebVoyager 上取得了 [94% 的成绩](https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fwebvoyager)！\n\nMagnitude 利用视觉 AI，使您能够通过自然语言控制浏览器。\n- 🧭 **导航** - 能够识别并理解任何界面，从而规划操作步骤\n- 🖱️ **交互** - 使用鼠标和键盘执行精确的操作\n- 🔍 **提取** - 智能地提取有用的结构化数据\n- ✅ **验证** - 内置测试运行器，具备强大的视觉断言功能\n\n您可以使用它来自动化网络任务、在没有 API 的情况下实现应用间的集成、提取数据、测试您的 Web 应用程序，或者将其作为构建您自己的浏览器代理的基础组件。\n\n![视频展示了 Magnitude 测试在终端中运行，以及代理在浏览器中执行操作](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmagnitudedev_browser-agent_readme_a495d3c7d484.gif)\n\n↕️ Magnitude 实际运行效果！ ↕️\n```ts\n\u002F\u002F Magnitude 可以处理高层次的任务\nawait agent.act('创建一个任务', {\n    \u002F\u002F 可选地传递代理将在适当位置使用的数据\n    data: {\n        title: '使用 Magnitude',\n        description: '运行 \"npx create-magnitude-app\" 并按照说明操作',\n    },\n});\n\n\u002F\u002F 它也可以处理低层次的操作\nawait agent.act('将“使用 Magnitude”拖动到进行中列的顶部');\n\n\u002F\u002F 根据 DOM 内容，结合提供的 zod 模式智能提取数据\nconst tasks = await agent.extract(\n    '列出进行中的任务',\n    z.array(z.object({\n        title: z.string(),\n        description: z.string(),\n        \u002F\u002F 代理可以提取现有数据或新的见解\n        difficulty: z.number().describe('请评估难度，范围为 1 到 5')\n    })),\n);\n```\n\n## 开始使用\n\n### 运行您的第一个浏览器自动化\n```bash\nnpx create-magnitude-app\n```\n\n这将创建一个新的项目，并引导您完成 Magnitude 的设置步骤。同时还会生成一个示例脚本，您可以立即运行！\n\n### 使用测试运行器\n要在 **现有** Web 应用中安装测试运行器，请运行：\n```bash\nnpm i --save-dev magnitude-test && npx magnitude init\n```\n\n这将创建一个基本的测试目录 `tests\u002Fmagnitude`，其中包含：\n- `magnitude.config.ts`: Magnitude 测试配置文件\n- `example.mag.ts`: 示例测试文件\n\n有关如何运行测试以及如何集成到 CI\u002FCD 中的信息，请参阅 [此处](https:\u002F\u002Fdocs.magnitude.run\u002Fcore-concepts\u002Frunning-tests)。\n\n> [!注意]\n> Magnitude 需要一个大型的 **视觉基础** 模型。我们推荐使用 Claude Sonnet 4 以获得最佳性能，但也兼容 Qwen-2.5VL 72B。更多信息请参阅 [文档](https:\u002F\u002Fdocs.magnitude.run\u002Fcustomizing\u002Fllm-configuration)。\n\n\n## 为什么选择 Magnitude？\n❌ **问题 #1：** 大多数浏览器代理会在页面元素周围绘制编号方框——由于现代网站结构复杂，这种方法难以泛化。\n\n✅ **解决方案：视觉优先架构**\n* 基于视觉的 LLM 指定像素坐标\n* 真正的泛化能力，不受 DOM 结构限制\n* 面向未来的架构，适用于桌面应用、虚拟机等\n\n❌ **问题 #2：** 大多数浏览器代理采用“高层级提示 + 工具 = 直到完成”的模式——这种模式适合演示，但不适合生产环境。\n\n✅ **解决方案：可控且可重复的自动化**\n* 灵活的抽象层级（细粒度操作与流程）\n* 在代理和操作层面自定义动作和提示\n* 通过原生缓存系统实现确定性运行 *(正在进行中)*\n\n## 更多信息\n\n有关如何更好地构建 Magnitude 自动化和测试用例的更多信息，请参阅我们的 [文档](https:\u002F\u002Fdocs.magnitude.run)。\n\n## 联系方式\n如果您是企业用户，并希望获得更多功能或支持，请随时通过 founders@magnitude.run 与我们联系，或在此 [预约通话](https:\u002F\u002Fcal.com\u002Ftom-greenwald\u002F30min)，讨论您的需求。\n\n您也可以加入我们的 \u003Ca href=\"https:\u002F\u002Fdiscord.gg\u002FVcdpMh9tTy\" target=\"_blank\">Discord 社区\u003C\u002Fa>,获取帮助或提出建议！","# Browser-Agent (Magnitude) 快速上手指南\n\nMagnitude 是一款基于视觉 AI 的浏览器自动化工具，能够通过自然语言控制浏览器执行导航、交互、数据提取和验证任务。它采用“视觉优先”架构，不依赖 DOM 结构，适用于复杂现代网站及桌面应用自动化。\n\n## 环境准备\n\n- **系统要求**：支持 Node.js 的操作系统（Windows \u002F macOS \u002F Linux）\n- **前置依赖**：\n  - Node.js >= 18.x\n  - npm 或 pnpm 包管理器\n  - 一个支持视觉理解的 LLM 模型（推荐 **Claude Sonnet 4**，兼容 **Qwen-2.5VL 72B**）\n- **网络建议**：由于模型服务可能涉及海外 API，请确保网络通畅或配置合适的代理。\n\n## 安装步骤\n\n### 方式一：创建新项目（推荐新手）\n\n运行以下命令创建一个新的 Magnitude 项目并跟随向导完成初始化：\n\n```bash\nnpx create-magnitude-app\n```\n\n该命令会自动生成项目结构、配置文件及示例脚本，可直接运行体验。\n\n### 方式二：在现有项目中集成测试功能\n\n若你已在开发 Web 应用，希望加入自动化测试：\n\n```bash\nnpm i --save-dev magnitude-test && npx magnitude init\n```\n\n执行后将生成 `tests\u002Fmagnitude` 目录，包含：\n- `magnitude.config.ts`：配置文件\n- `example.mag.ts`：示例测试文件\n\n> 💡 提示：目前暂无官方中国镜像源，如遇下载缓慢，可配置 npm 国内镜像（如淘宝镜像）加速依赖安装：\n> ```bash\n> npm config set registry https:\u002F\u002Fregistry.npmmirror.com\n> ```\n\n## 基本使用\n\n### 1. 执行高层任务（自然语言驱动）\n\n```ts\nawait agent.act('Create a task', {\n    data: {\n        title: 'Use Magnitude',\n        description: 'Run \"npx create-magnitude-app\" and follow the instructions',\n    },\n});\n```\n\n### 2. 执行底层操作（精确控制）\n\n```ts\nawait agent.act('Drag \"Use Magnitude\" to the top of the in progress column');\n```\n\n### 3. 智能数据提取（结合 Zod  schema）\n\n```ts\nconst tasks = await agent.extract(\n    'List in progress tasks',\n    z.array(z.object({\n        title: z.string(),\n        description: z.string(),\n        difficulty: z.number().describe('Rate the difficulty between 1-5')\n    })),\n);\n```\n\n### 4. 运行测试（集成 CI\u002FCD）\n\n在项目根目录执行：\n\n```bash\nnpx magnitude test\n```\n\n详细测试运行与集成文档请参阅：[https:\u002F\u002Fdocs.magnitude.run\u002Fcore-concepts\u002Frunning-tests](https:\u002F\u002Fdocs.magnitude.run\u002Fcore-concepts\u002Frunning-tests)\n\n---\n\n开始构建你的视觉驱动浏览器自动化流程吧！","某电商运营团队需要每日从多个竞争对手的动态网页中采集商品价格、库存状态及促销标签，并自动验证自家网站在促销活动中的前端展示是否正确。\n\n### 没有 browser-agent 时\n- **维护成本极高**：传统自动化脚本依赖固定的 DOM 结构（如 XPath 或 CSS 选择器），一旦竞品网站更新布局或混淆类名，脚本立即失效，需人工反复修复。\n- **无法处理复杂交互**：面对需要拖拽排序、悬停显示详情或非标准弹窗的场景，基于代码的脚本难以模拟精确的鼠标轨迹和视觉判断。\n- **数据提取僵化**：只能抓取预设字段，无法像人类一样理解“打折力度”或“热销指数”等隐含信息，更无法根据上下文动态评估数据质量。\n- **视觉验证缺失**：测试环节仅能核对数据是否存在，无法确认按钮颜色、排版错位等直接影响用户体验的视觉回归问题。\n\n### 使用 browser-agent 后\n- **自适应页面变化**：browser-agent 采用“视觉优先”架构，直接通过屏幕像素坐标操作，即使网站重构 UI，只要视觉元素不变，任务即可稳定运行。\n- **拟人化精准操作**：利用多模态大模型理解界面，能自然执行“将商品拖至置顶”或“点击浮动广告关闭键”等复杂动作，无需硬编码坐标。\n- **智能结构化提取**：只需定义 Zod  schema，browser-agent 即可理解页面语义，自动提取价格并智能推断“难度系数”或“推荐指数”等新洞察。\n- **内置视觉断言测试**：在 CI\u002FCD 流程中直接运行视觉测试，自动截图比对，确保促销页面的按钮位置、颜色样式与设计要求完全一致。\n\nbrowser-agent 通过将浏览器控制从“代码驱动”升级为“视觉驱动”，让自动化脚本具备了像人类一样的观察与决策能力，彻底解决了现代动态网页难以自动化的痛点。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmagnitudedev_browser-agent_6e520ed1.png","magnitudedev","Magnitude","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fmagnitudedev_44b7181e.png","Open source coding agent",null,"usemagnitude","https:\u002F\u002Fmagnitude.dev","https:\u002F\u002Fgithub.com\u002Fmagnitudedev",[84,88,92,96,100],{"name":85,"color":86,"percentage":87},"TypeScript","#3178c6",85,{"name":89,"color":90,"percentage":91},"JavaScript","#f1e05a",9.5,{"name":93,"color":94,"percentage":95},"HTML","#e34c26",5,{"name":97,"color":98,"percentage":99},"Python","#3572A5",0.4,{"name":101,"color":102,"percentage":103},"Shell","#89e051",0,4016,224,"2026-04-02T17:00:44","Apache-2.0","未说明","非必需本地 GPU。工具基于视觉大模型（Vision AI），推荐使用云端 API（如 Claude Sonnet 4 或 Qwen-2.5VL 72B），因此对本地显卡无强制要求，具体取决于所选模型的托管方式。",{"notes":111,"python":112,"dependencies":113},"该工具主要通过 npx create-magnitude-app 安装，依赖 Node.js 环境而非传统的 Python 深度学习栈。核心功能需要接入大型视觉语言模型（VLM），官方推荐 Claude Sonnet 4 以获得最佳性能，同时也兼容 Qwen-2.5VL 72B。用户需自行配置相应的 LLM API 密钥。","未说明 (通过 npx\u002Fnpm 运行，通常依赖 Node.js 环境)",[114,115,116],"Node.js","npm\u002Fnpx","magnitude-test (可选，用于测试运行器)",[13,15,14,53],[119,120,121,122,123,124,125,126,127,128],"test","typescript","automation","browser","framework","javascipt","ai","playwright","selenium","rpa","2026-03-27T02:49:30.150509","2026-04-06T09:46:08.067534",[132,137,142,147,152,157],{"id":133,"question_zh":134,"answer_zh":135,"source_url":136},12310,"遇到 'Misalignment: Could not align click\u002Ftype action: Invalid URL' 错误怎么办？","这通常是由于版本过旧或配置问题导致的。请运行 'npm i magnitude-test@latest' 更新到最新版本，并查阅官方文档 (https:\u002F\u002Fdocs.magnitude.run) 了解新的语法规范。如果问题依旧，请检查 executor 的设置是否正确。","https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fbrowser-agent\u002Fissues\u002F25",{"id":138,"question_zh":139,"answer_zh":140,"source_url":141},12311,"Magnitude 是否支持等待动态内容加载（Wait 功能）？","是的，该功能已在 magnitude-core v0.2.16 及 magnitude-test 0.1.2 版本中添加。现在代理可以自动检测并等待网络请求或动态内容。此外，建议在使用此功能时尝试 Gemini 2.5 Pro 模型，因为测试发现 OpenAI 模型在此类任务上表现不佳。","https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fbrowser-agent\u002Fissues\u002F7",{"id":143,"question_zh":144,"answer_zh":145,"source_url":146},12312,"运行测试时浏览器打开后无操作、只消耗 Token 且报错 'Expected positive integer for width' 是怎么回事？","这是一个已知的视口宽度舍入问题（接收到了浮点数而非整数）。该问题已在 magnitude-test 0.2.11 版本中修复。请升级依赖包：'npm i magnitude-test@latest' 即可解决。","https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fbrowser-agent\u002Fissues\u002F52",{"id":148,"question_zh":149,"answer_zh":150,"source_url":151},12313,"运行 'npx magnitude' 时报错 'Cannot read properties of undefined (reading 'exports')' 如何解决？","这通常是因为缺少必要的环境变量配置。请确保在运行测试前，已将 Anthropic API Key 设置为环境变量（例如 ANTHROPIC_API_KEY）。如果密钥缺失，工具可能无法正确初始化从而导致模块加载错误。","https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fbrowser-agent\u002Fissues\u002F98",{"id":153,"question_zh":154,"answer_zh":155,"source_url":156},12314,"如何在测试中切换到新打开的浏览器标签页（Tab）？","切换标签页的功能已在 version 0.1.2 中发布。更新到该版本后，Agent 将能够识别并切换到新打开的标签页继续执行操作，无需手动编写复杂的切换逻辑。","https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fbrowser-agent\u002Fissues\u002F21",{"id":158,"question_zh":159,"answer_zh":160,"source_url":161},12315,"如何让测试在单个用例失败后继续执行，而不是直接退出？","目前开发团队正在重构测试运行器（增加进程隔离），该功能的具体实现方式（配置项或命令行参数）可能会随之调整。建议关注项目后续更新或参考其他主流测试运行器（如 Jest, Mocha）的 '--bail' 或类似参数的命名惯例，目前暂无稳定的独立配置开关。","https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fbrowser-agent\u002Fissues\u002F41",[163,168,173,178,183,188,193,197,202,206,210,215,220,225,230,235,240,245,250,255],{"id":164,"version":165,"summary_zh":166,"released_at":167},62693,"magnitude-core@0.3.1","### 补丁变更\n\n-   [`b1771a4`](https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fbrowser-agent\u002Fcommit\u002Fb1771a4ee6e0d5064e67fd3de52be13eabbe023b) 感谢 [@ddwang](https:\u002F\u002Fgithub.com\u002Fddwang)! - 暂停\u002F恢复 API\n\n-   [`dedb7c6`](https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fbrowser-agent\u002Fcommit\u002Fdedb7c644ffd629f08896bac205b22152075050b) 感谢 [@anerli](https:\u002F\u002Fgithub.com\u002Fanerli)! - 将默认模型更改为 haiku 4.5\n\n-   [`2ef1102`](https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fbrowser-agent\u002Fcommit\u002F2ef110264e65e06f2dff0d60ffb5cd36ae636a6e) 感谢 [@DmitryGvozd12](https:\u002F\u002Fgithub.com\u002FDmitryGvozd12)! - 修复累积令牌计数膨胀问题\n","2026-02-08T20:51:15",{"id":169,"version":170,"summary_zh":171,"released_at":172},62694,"magnitude-test@0.3.13","### 补丁变更\n\n-   更新了依赖项 \\[[`b1771a4`](https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fbrowser-agent\u002Fcommit\u002Fb1771a4ee6e0d5064e67fd3de52be13eabbe023b), [`dedb7c6`](https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fbrowser-agent\u002Fcommit\u002Fdedb7c644ffd629f08896bac205b22152075050b), [`2ef1102`](https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fbrowser-agent\u002Fcommit\u002F2ef110264e65e06f2dff0d60ffb5cd36ae636a6e)]:\n    -   magnitude-core@0.3.1\n","2026-02-08T20:51:12",{"id":174,"version":175,"summary_zh":176,"released_at":177},62695,"magnitude-mcp@0.1.3","### 补丁变更\n\n-   [`a8f4ee0`](https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fbrowser-agent\u002Fcommit\u002Fa8f4ee05b1269c3a2513497f085593d81992969c) 感谢 [@anerli](https:\u002F\u002Fgithub.com\u002Fanerli)! - 在 mcp 动作集执行后增加稳定性等待\n\n-   更新了依赖项 \\[[`b1771a4`](https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fbrowser-agent\u002Fcommit\u002Fb1771a4ee6e0d5064e67fd3de52be13eabbe023b), [`dedb7c6`](https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fbrowser-agent\u002Fcommit\u002Fdedb7c644ffd629f08896bac205b22152075050b), [`2ef1102`](https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fbrowser-agent\u002Fcommit\u002F2ef110264e65e06f2dff0d60ffb5cd36ae636a6e)]:\n    -   magnitude-core@0.3.1\n","2026-02-08T20:51:09",{"id":179,"version":180,"summary_zh":181,"released_at":182},62696,"create-magnitude-app@0.0.8","### 补丁变更\n\n-   [`dedb7c6`](https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fbrowser-agent\u002Fcommit\u002Fdedb7c644ffd629f08896bac205b22152075050b) 感谢 [@anerli](https:\u002F\u002Fgithub.com\u002Fanerli)! - 将默认模型更改为 haiku 4.5\n","2026-02-08T20:51:06",{"id":184,"version":185,"summary_zh":186,"released_at":187},62697,"magnitude-core@0.3.0","### 小改动\n\n-   [`265a2ab`](https:\u002F\u002Fgithub.com\u002Fsagekit\u002Fmagnitude\u002Fcommit\u002F265a2ab348ec916f6382d3ad1dfe572b55e33090) 感谢 [@anerli](https:\u002F\u002Fgithub.com\u002Fanerli)! - 移除对分离接地的支持，以减少依赖问题并简化代码\n","2025-09-25T11:28:35",{"id":189,"version":190,"summary_zh":191,"released_at":192},62698,"magnitude-test@0.3.12","### 补丁变更\n\n-   [`265a2ab`](https:\u002F\u002Fgithub.com\u002Fsagekit\u002Fmagnitude\u002Fcommit\u002F265a2ab348ec916f6382d3ad1dfe572b55e33090) 感谢 [@anerli](https:\u002F\u002Fgithub.com\u002Fanerli)! - 移除对分离接地的支持，以减少依赖问题并简化代码\n\n-   更新了依赖项 \\[[`265a2ab`](https:\u002F\u002Fgithub.com\u002Fsagekit\u002Fmagnitude\u002Fcommit\u002F265a2ab348ec916f6382d3ad1dfe572b55e33090)]:\n    -   magnitude-core@0.3.0\n","2025-09-25T11:28:32",{"id":194,"version":195,"summary_zh":191,"released_at":196},62699,"magnitude-mcp@0.1.2","2025-09-25T11:28:29",{"id":198,"version":199,"summary_zh":200,"released_at":201},62700,"magnitude-mcp@0.1.2-beta.0","### 补丁变更\n\n-   [`265a2ab`](https:\u002F\u002Fgithub.com\u002Fsagekit\u002Fmagnitude\u002Fcommit\u002F265a2ab348ec916f6382d3ad1dfe572b55e33090) 感谢 [@anerli](https:\u002F\u002Fgithub.com\u002Fanerli)! - 移除对分离接地的支持，以减少依赖问题并简化代码\n\n-   更新了依赖项 \\[[`265a2ab`](https:\u002F\u002Fgithub.com\u002Fsagekit\u002Fmagnitude\u002Fcommit\u002F265a2ab348ec916f6382d3ad1dfe572b55e33090)]:\n    -   magnitude-core@0.3.0-beta.0\n","2025-09-25T11:17:34",{"id":203,"version":204,"summary_zh":200,"released_at":205},62701,"magnitude-test@0.3.12-beta.0","2025-09-25T11:17:31",{"id":207,"version":208,"summary_zh":186,"released_at":209},62702,"magnitude-core@0.3.0-beta.0","2025-09-25T11:17:28",{"id":211,"version":212,"summary_zh":213,"released_at":214},62703,"magnitude-mcp@0.1.1","### Patch Changes\n\n-   [`24a00b1`](https:\u002F\u002Fgithub.com\u002Fsagekit\u002Fmagnitude\u002Fcommit\u002F24a00b1884e18b4ec88f3c591efab732352ab0a1) Thanks [@anerli](https:\u002F\u002Fgithub.com\u002Fanerli)! - fix bundling issue\n","2025-09-25T10:31:01",{"id":216,"version":217,"summary_zh":218,"released_at":219},62704,"magnitude-mcp@0.1.0","### Minor Changes\n\n-   [`b1ee0c2`](https:\u002F\u002Fgithub.com\u002Fsagekit\u002Fmagnitude\u002Fcommit\u002Fb1ee0c225ee33ac22f96a8c1828d6101921e57fc) Thanks [@anerli](https:\u002F\u002Fgithub.com\u002Fanerli)! - - new MCP server for interacting with browser with persistent sessions\n    -   fix innerHTML assignment bugs causing issues for cursor visual and data extraction on google sites\n    -   fix cursor visual not having fixed positioning when scrolling main document\n    -   fix screenshot not breaking out of retry loop on success\n\n### Patch Changes\n\n-   Updated dependencies \\[[`b1ee0c2`](https:\u002F\u002Fgithub.com\u002Fsagekit\u002Fmagnitude\u002Fcommit\u002Fb1ee0c225ee33ac22f96a8c1828d6101921e57fc)]:\n    -   magnitude-core@0.2.32\n","2025-09-25T10:08:26",{"id":221,"version":222,"summary_zh":223,"released_at":224},62705,"magnitude-core@0.2.32","### Patch Changes\n\n-   [`b1ee0c2`](https:\u002F\u002Fgithub.com\u002Fsagekit\u002Fmagnitude\u002Fcommit\u002Fb1ee0c225ee33ac22f96a8c1828d6101921e57fc) Thanks [@anerli](https:\u002F\u002Fgithub.com\u002Fanerli)! - - new MCP server for interacting with browser with persistent sessions\n    -   fix innerHTML assignment bugs causing issues for cursor visual and data extraction on google sites\n    -   fix cursor visual not having fixed positioning when scrolling main document\n    -   fix screenshot not breaking out of retry loop on success\n","2025-09-25T10:08:23",{"id":226,"version":227,"summary_zh":228,"released_at":229},62706,"magnitude-test@0.3.11","### Patch Changes\n\n-   [#131](https:\u002F\u002Fgithub.com\u002Fsagekit\u002Fmagnitude\u002Fpull\u002F131) [`9d4b8a5`](https:\u002F\u002Fgithub.com\u002Fsagekit\u002Fmagnitude\u002Fcommit\u002F9d4b8a52badd6d5c39d3357b8a1713bd875154d6) Thanks [@ewired](https:\u002F\u002Fgithub.com\u002Fewired)! - Code structure change to improve IPC for future reworking and fix a build issue with a circular dependency.\n\n-   [#134](https:\u002F\u002Fgithub.com\u002Fsagekit\u002Fmagnitude\u002Fpull\u002F134) [`510c24e`](https:\u002F\u002Fgithub.com\u002Fsagekit\u002Fmagnitude\u002Fcommit\u002F510c24e9ae85b9497a133f6ae93c9ccde3da56f5) Thanks [@ewired](https:\u002F\u002Fgithub.com\u002Fewired)! - magnitude-test now shows rich data in --debug logs including agent thoughts, token usage and known costs. --plain is now available for easily readable plain text logs. display.thoughts is now available in magnitude.config.ts to show thoughts in the TUI.\n\n-   [`9123917`](https:\u002F\u002Fgithub.com\u002Fsagekit\u002Fmagnitude\u002Fcommit\u002F9123917e4dd8be30687b43be6797d152fdffc571) Thanks [@anerli](https:\u002F\u002Fgithub.com\u002Fanerli)! - fix: add array llm client type to test config\n\n-   Updated dependencies \\[[`b1ee0c2`](https:\u002F\u002Fgithub.com\u002Fsagekit\u002Fmagnitude\u002Fcommit\u002Fb1ee0c225ee33ac22f96a8c1828d6101921e57fc)]:\n    -   magnitude-core@0.2.32\n","2025-09-25T10:08:20",{"id":231,"version":232,"summary_zh":233,"released_at":234},62707,"magnitude-core@0.2.31","### Patch Changes\n\n-   [`e54e4e1`](https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fmagnitude\u002Fcommit\u002Fe54e4e10ab05b8de593ba97eedf89121a3235971) Thanks [@anerli](https:\u002F\u002Fgithub.com\u002Fanerli)! - add anthropic version config option to vertex\n","2025-08-06T18:45:30",{"id":236,"version":237,"summary_zh":238,"released_at":239},62708,"magnitude-test@0.3.10","### Patch Changes\n\n-   Updated dependencies \\[[`e54e4e1`](https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fmagnitude\u002Fcommit\u002Fe54e4e10ab05b8de593ba97eedf89121a3235971)]:\n    -   magnitude-core@0.2.31\n","2025-08-06T18:45:27",{"id":241,"version":242,"summary_zh":243,"released_at":244},62709,"magnitude-core@0.2.29","### Patch Changes\n\n-   [`870d225`](https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fmagnitude\u002Fcommit\u002F870d2257c21cef24b4c14938fc20ae23fb369a80) Thanks [@anerli](https:\u002F\u002Fgithub.com\u002Fanerli)! - disable temp config in vertex for now, seems format does not align with expected\n","2025-08-06T18:35:03",{"id":246,"version":247,"summary_zh":248,"released_at":249},62710,"magnitude-test@0.3.8","### Patch Changes\n\n-   Updated dependencies \\[[`870d225`](https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fmagnitude\u002Fcommit\u002F870d2257c21cef24b4c14938fc20ae23fb369a80)]:\n    -   magnitude-core@0.2.29\n","2025-08-06T18:35:00",{"id":251,"version":252,"summary_zh":253,"released_at":254},62711,"magnitude-core@0.2.28","### Patch Changes\n\n-   [`aa3dadb`](https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fmagnitude\u002Fcommit\u002Faa3dadb9b4662d610809191b84aa59cb017981f6) Thanks [@anerli](https:\u002F\u002Fgithub.com\u002Fanerli)! - throw on failure to get conn observations - prevents potential infinite wait loop if browser closed\n","2025-07-30T19:34:24",{"id":256,"version":257,"summary_zh":258,"released_at":259},62712,"magnitude-test@0.3.7","### Patch Changes\n\n-   Updated dependencies \\[[`aa3dadb`](https:\u002F\u002Fgithub.com\u002Fmagnitudedev\u002Fmagnitude\u002Fcommit\u002Faa3dadb9b4662d610809191b84aa59cb017981f6)]:\n    -   magnitude-core@0.2.28\n","2025-07-30T19:34:21"]