[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-polterguy--magic":3,"tool-polterguy--magic":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":76,"owner_location":79,"owner_email":80,"owner_twitter":81,"owner_website":82,"owner_url":83,"languages":84,"stars":113,"forks":114,"last_commit_at":115,"license":116,"difficulty_score":10,"env_os":117,"env_gpu":118,"env_ram":118,"env_deps":119,"category_tags":123,"github_topics":124,"view_count":23,"oss_zip_url":80,"oss_zip_packed_at":80,"status":16,"created_at":131,"updated_at":132,"faqs":133,"releases":164},2018,"polterguy\u002Fmagic","magic","Fully Autonomous AI-based Software Development Assistant","Magic 是一个开源的全栈AI开发助手，你只需用自然语言描述需求，它就能自动生成完整的应用程序——包括数据库、后端接口和前端界面，全程无需手动编写代码。它内置了Hyperlambda语言作为AI编程核心，结合SQLite等数据库，实现了零依赖、零锁仓的开发体验，所有代码可直接在本地运行，无需部署或编译，保存即测试。\n\nMagic 解决了传统开发中繁琐的环境配置、多系统集成和重复性编码问题，尤其适合希望快速验证想法的开发者、初创团队和非专业程序员。它不仅能生成应用，还能运行AI代理自动浏览网页、填写表单、对接CRM或ERP系统，真正实现“AI写代码、AI做任务”。\n\n独特之处在于其一体化设计：内置IDE、SQL可视化工具、RBAC权限系统、任务调度器和机器学习模块，所有功能无缝集成。你甚至可以用注释驱动开发，用自然语言描述功能，系统自动补全代码。无论是想搭建一个客户管理系统，还是让AI自动处理日常任务，Magic 都能让你从零开始，几分钟内看到成果。适合喜欢高效、低门槛开发的用户，无需深厚技术背景也能上手。","\n# Magic Cloud - Fully Autonomous AI-based Software Development Assistant\n\n> Magic is an open-source, self-hostable AI software development platform that generates and runs full-stack business applications and AI agents from natural language.\n\nMagic is built on top of [OpenAI](https:\u002F\u002Fopenai.com) and [Hyperlambda](https:\u002F\u002Fainiro.io\u002Fhyperlambda\u002F), a DSL specifically created to solve anything related to backend software development, and to be _\"The AI agent programming language\"_. Create full stack apps, in an open source environment, resembling Lovable, Bolt, or Replit. Use natural language as input, and host it on your own hardware if you wish.\n\n**No additional \"backend connectors\" or \"database connectors\" required**!\n\n**Hence, ZERO lockin!!**\n\nEverything is 100% integrated, thx to SQLite, with optional MySQL, PostgreSQL, and Microsoft SQL Server capabilities. Basically, run the whole shebang on _your own hardware_ if you wish ...\n\n## Open Source _\"Vibe Coding\"_ platform\n\nBelow is an app that was created with the following prompt; \n\n> Create me a full stack app to manage VIP customer for a car dealership\n\nThe whole process took about 30 minutes in total, with less than a handful of errors, correcting the LLM or giving feedback some 5 to 10 times during the process. All bugs were easily tracked down and eliminated by a seasoned software developer during the process.\n\n![CRM system for car dealership](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fpolterguy_magic_readme_9add79b1428b.png)\n\nMagic asked a handful of control questions, before it automatically generated the database, created the backend code based upon the integrated Hyperlambda Generator, before finally assembling the frontend based upon the API - Complete with authentication and authorization, 100% secure (of course!) Everything deployed locally, on the integrated and built-in webserver.\n\n**Once you save the code, you can test it! No _\"deployment\"_ or _\"publish\"_ required to test code!**\n\n## Bad ass AI Agent!\n\nBelow is the AI agent in Magic 100% autonomously browsing the web and filling out a _\"contact us\"_ form. This particular example is using the integrated headless browser, that allows your AI agent to _\"see\"_ the web, autonomously browse it, and solve tasks.\n\n![Headless browser in Magic filling out a form](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fpolterguy_magic_readme_1d6f0ed9f5b5.png)\n\nHowever, you can also vibe code AI agents integrated with your CRM system, ERP system, legacy databases, _\"whatever\"_. Magic fundamentally _is_ an AI agent, for building software and AI agents. What you use it for, is up to you.\n\nIn addition to the AI agent in its dashboard, that generates entire full stack apps using nothing but natural language input - There's a whole range of additional components in the system allowing you to automate software development, such as for instance;\n\n* CRUD generator, creating API endpoint using database meta information\n* SQL Studio, allowing you to visually design and manage your SQL databases\n* Built-in RBAC\n* Hyper IDE, for manually edit code in a VS code like environment\n* Task manager for administrating and scheduling tasks\n* Machine Learning component allowing you to manage AI agents and chatbots\n* Plugin repository for installing both frontend types of websites, and backend code\n* Plus many more ...\n\nBelow is a screenshot from Hyper IDE.\n\n![Hyper IDE](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fpolterguy_magic_readme_599bf006cd00.png)\n\nThe above illustrates how Magic facilitates for _\"comment driven development\"_, as in provide _\"natural language instructions\"_ with a declarative comment, and have the system implement the code using the built-in AI code generator.\n\n## Also a web server\n\nMagic is also a web server, allowing you to _instantly deploy_ everything, without compilation, build processes, complex pipeline connectors, etc. So the process is as follows;\n\n1. Create your prompt\n2. Press enter\n3. Test!\n\n... or use the integrated headless browser to automatically generate AI workflows that _tests your system automatically once done_!\n\nThis comes in stark contrast to other less sophisticated tools, such as Lovable and Bolt44 that requires you to deploy into _2 different 3rd party providers before you can even test your code_. Hence, the development model in Magic is probably for most practical concerns roughly 10x faster and more optimised ...\n\nIn addition to having the ability to generate pure JS, CSS, and HTML frontends, that's immediately being served, without any deployment pipelines - The system also comes with several pre-built frontend systems out of the box, such as the [AI Expert System](https:\u002F\u002Fainiro.io\u002Fai-expert-system), which allows you to serve password protected AI agents, and\u002For for that matter deliver entire SaaS AI solutions.\n\nThe system is particularly well suited for creating AI agents.\n\n## Headless Browser\n\nMagic contains a headless browser, PuppeteerSharp specifically, that allows you to browse the web as a human being, fill out forms, click buttons, etc.\n\n* _\"Go to xyz website, identify their contact us form and change URLs if required, and fill out their contact us form\"_\n\nYou can see an example of that prompt in the screenshot below.\n\n![Headless browser in Magic filling out a form](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fpolterguy_magic_readme_1d6f0ed9f5b5.png)\n\n## Git integration\n\nContrary to other vibe coding tools, Magic Cloud was built for software developers from day 1. That means among other things it's got Git integrated as an integral part of the platform. This allows you to setup any amount of pipelines you wish, using Git for code, or GitHub workflows for deployments.\n\n1. Create a new project\n2. Vibe code all the tools and even your GitHub workflows if you wish\n3. Commit and push\n\nBelow is how the integrated AI agents objectively compares Magic Cloud to Lovable and Bolt44.\n\n![A comparison between Lovable, Bolt44 and Magic Cloud](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fpolterguy_magic_readme_0cfebcf908cc.png)\n\n## Python, Terminal, and C# integration\n\nGenerate and execute Python scripts on the fly, and have the LLM use these as _\"tools\"_. In addition, you can use BASH and the underlying terminal, and you can create Hyperlambda extension keywords using C#.\n\nSince Magic is running in a protected service account by default, this is actually quite safe - However, obviously do *not* open up endpoints allowing 3rd party users to generate and execute arbitrary Python code.\n\nYou can also persist Python scripts, and reference these later as _\"tools\"_, permanently widening the capabilities of AI agents, or for that matter integrate Python execution into your endpoints and services.\n\n![Executing Python from Magic Cloud](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fpolterguy_magic_readme_2b2eb0d7f8c4.png)\n\n**NOTICE** - You _obviously_ have to be logged in as root to generate and execute both Python scripts, Terminal scripts, and to create C# extensions. Magic has a unique security model however, that eliminates entire _axioms_ of security-related _\"holes\"_. But you still need to _keep_ your brain. Magic is not (pun!) a _\"magic pill\"_.\n\n## Deploy anywhere\n\nIf you choose to create AI agents instead of full stack app, something the system is particularly well suited for, you can choose to deliver these as password protected AI expert systems, or embeddable AI chatbots, embedded on some website. Below is our AI chatbot. You can try it [here](https:\u002F\u002Fainiro.io)\n\n![Embeddable AI chatbot](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fpolterguy_magic_readme_ad6ccabddb80.png)\n\n## 20x times faster than Python\n\nWhen we measure Hyperlambda and Magic Cloud, it's roughly around 20 times faster than similar solutions built in Python, such as Fast API or Flask. Compared to LangChain, it's probably around 50 times faster, in addition to making it much easier to create workflows, due to being able to create backend code using English, instead of _\"drag and drop WYWIWYG hell\"_. Hyperlambda solutions are in general on pair with C# combined with Entity Framework, both on scalabaility and performance. Below is Hyperlambda versus Fast API and Flask.\n\n![Python versus Hyperlambda](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fpolterguy_magic_readme_575371dfbd26.png)\n\nMagic Cloud is built in C# and .Net Core 10.\n\nOn average, Magic is probably around 100 to 1,000 times faster than _\"graphical workflow solutions\"_, such as N8N, LangChain, Zapier, Make, etc - Due to relying upon an actual programming language instead of JSON, XML, or Markdown based _\"workflow files\"_. Executing a piece of Hyperlambda is around 1,000 times faster than parsing _\"dynamic logic\"_ from YAMNL or JSON files.\n\nHyperlambda is almost on pair with pure C# code!\n\n## Getting started\n\nThe easiest way to get started is to use Docker and create a _\"docker-compose.yaml\"_ file with the following content;\n\n```yaml\nversion: \"3.8\"\n\nservices:\n  backend:\n    image: servergardens\u002Fmagic-backend:latest\n    platform: linux\u002Famd64\n    container_name: magic_backend\n    restart: unless-stopped\n\n    ports:\n      - \"4444:4444\"\n\n    volumes:\n      - magic_files_etc:\u002Fmagic\u002Ffiles\u002Fetc\n      - magic_files_data:\u002Fmagic\u002Ffiles\u002Fdata\n      - magic_files_config:\u002Fmagic\u002Ffiles\u002Fconfig\n      - magic_files_modules:\u002Fmagic\u002Ffiles\u002Fmodules\n\n  frontend:\n    image: servergardens\u002Fmagic-frontend:latest\n    container_name: magic_frontend\n    restart: unless-stopped\n\n    depends_on:\n      - backend\n\n    ports:\n      - \"5555:80\"\n\nvolumes:\n  magic_files_etc:\n  magic_files_data:\n  magic_files_config:\n  magic_files_modules:\n```\n\nSave it somewhere, and execute `docker compose up` or something, visit `localhost:5555`, login with _\"root\"_ \u002F _\"root\"_, and configure the system. You can [read more here](https:\u002F\u002Fdocs.ainiro.io\u002Fgetting-started\u002F) for alternatives, such as running the codebase directly on your own machine.\n\nYou can also watch me [guide you through the setup process here](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=k6eSKxc6oM8).\n\n### Bring your own OpenAI API key\n\nTo use the system you'll need an OpenAI API key. You can create one [here](https:\u002F\u002Fplatform.openai.com\u002Fapi-keys).\n\n**NOTICE** - To gain access to `gpt-5.4`, you might have to deposit $51 into your OpenAI API account. Magic depends upon OpenAI, and without depositing money into OpenAI, you won't get access to gpt-5.4, which is the default model in Magic for _\"vibe coding\"_. You might get GPT-4.1 to work during vibe coding, but 5.4 is __much better__!\n\nIf you are absolutely allergic to OpenAI, there are Ollama and HuggingFace plugins for the system, allowing you to _\"override\"_ the inference functions with Ollama or HuggingFace models and endpoints - But embeddings can only be created with OpenAI's APIs.\n\n### DIY Home Cloud\n\nMagic easily installs on for instance a Mac Mini, using the Docker images. By combining it with CloudFlare tunnel, you can setup a web server in a couple of minutes, serving applications and data out of your home. The link below is running out of my house in Larnaca \u002F Cyprus, using a CloudFlare tunnel. We've tested it from US, Norway, and a whole range of countries, and it's relatively responsive, considering the connection it's being served over.\n\n* [CRM dashboard served from DIY web server](https:\u002F\u002Fhome.ainiro.io\u002Fanalytics-crm)\n\nBelow is a screenshot of the system.\n\n![Analytucs CRM Dashboard](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fpolterguy_magic_readme_e1d690ce517e.png)\n\nUse such a setup to completely bypass any censorship, and or spying, allowing you to completely avoid both spying and censoring of your content or applications.\n\n## LLM\n\nThe system internally is using OpenAI's gpt-5.4, with minimum reasoning turned on - But everything is tunable, and you can with a little bit of effort exchange the integrated defaults with Ollama or Hugging Face models. However, the Hyperlambda Generator's training dataset is _not_ made public, and we have no plans to do so either. This means that worst case scenario, you're still running your already generated systems perfectly fine, without the ability to generate new systems - Even if you were to loose the Hyperlambda Generator for some reasons.\n\nThe Hyperlambda Generator is however a fairly unique thing, due to Hyperlambda's integrated security model, something that allows for dynamically generating tools on the fly, and securely executing the generated code on the backend. Something demonstrated in our [natural language API](https:\u002F\u002Fainiro.io\u002Fnatural-language-api).\n\nThe following is a screenshot from a publicly available page (natural language API), where we accept input from any random visitor. The input is then transformed into Hyperlambda using our LLM, for then to be executed __in-process__ behind our DMZ in our militarized zone. We've offered hackers $100 if they can somehow exploit the endpoint to access PII or extract information using it. So far none have claimed money from us. We've had this offer out now for 3 months without any succeeding so far.\n\n![Natural Language API](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fpolterguy_magic_readme_42e53b1a3d41.png)\n\nIf you can hack the above API endpoint, I will give you $100!\n\n## Unique Security Model\n\nThe point about Hyperlambda, is that it's first of all running in a sandbox environment, so it doesn't have access to the file system outside of its own sandbox. In addition, it's got the ability to whitelist individual functions, according to its built-in RBAC system, allowing for your server to accept code as input, and still securely execute it - Without even knowing its origin.\n\nThe above is only possible by restricting function invocations at the execution level, which as far as I know, Hyperlambda is the _only_ programming language in the world that currently does.\n\nThis makes Hyperlambda uniquely fit for AI agents that needs to _\"generate tools on demand\"_, since it allows to owner of the solution to specify a subset of the server's vocabulary as _\"legal functions\"_, while no other functions are allowed.\n\nThe above allows you to deliver AI agents that creates their own tools __on demand__ - Without widening attack surface or creating security holes.\n\n## Technology\n\nMagic Cloud is built in .Net Core 10, and its dashboard is Angular. Hyperlambda again was entirely invented and created by yours truly, and you can find some articles about its unique technology below.\n\n* [Make C# more Dynamic with Hyperlambda](https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Farchive\u002Fmsdn-magazine\u002F2017\u002Fjune\u002Fcsharp-make-csharp-more-dynamic-with-hyperlambda)\n* [Active events, one design pattern instead of a dozen](https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Farchive\u002Fmsdn-magazine\u002F2017\u002Fmarch\u002Fpatterns-active-events-one-design-pattern-instead-of-a-dozen)\n\nHowever, Hyperlambda, and hence Magic Cloud by association, was built on a unique design pattern called _\"Active Events\"_, or _\"Slots and Signals\"_, which is an in-process model for executing _\"dynamic functions\"_, that's 100% unique for Magic Cloud. Active Events is at the core of Hyperlambda, and completely eliminates 100% of all cross projects dependencies, resulting in 100% _\"perfect\"_ encapsulation and cohesion.\n\nThe above design pattern, and Hyperlambda combined, is what facilitates for such extreme levels of security in Magic Cloud, where we can confidently trust that no AI-generated code does anything harmful - Simply because it doesn't have _permissions to do something malicious_ - Unless somebody explicitly gave it such permissions.\n\n> I'm so confident in its codebase quality, I'll give you $100 if you can find a (severe security) related bug in its backend code!\n\n## Maintenance\n\nMagic Cloud and Hyperlambda is developed and maintained by [AINIRO.IO](https:\u002F\u002Fainiro.io). We offer hosting, support, and software development services on top of Magic Cloud, in addition to delivering AI agents, chatbots, and AI solutions.\n\n## License\n\nThis project, and all of its satellite project, is licensed under the terms of the MIT license, as published by the Open Source Initiative. See LICENSE file for details. For licensing inquiries you can contact Thomas Hansen thomas@ainiro.io\n\n## Copyright and maintenance\n\nThe projects is copyright of Thomas Hansen 2019 - 2025, and professionally maintained by [AINIRO.IO](https:\u002F\u002Fainiro.io).\n","# 魔法云——全自主AI驱动的软件开发助手\n\n> Magic 是一款开源、可自托管的 AI 软件开发平台，能够通过自然语言生成并运行全栈业务应用和 AI 代理。\n\nMagic 基于 [OpenAI](https:\u002F\u002Fopenai.com) 和 [Hyperlambda](https:\u002F\u002Fainiro.io\u002Fhyperlambda\u002F) 构建而成。Hyperlambda 是一种专门用于解决后端软件开发相关问题的领域特定语言，被誉为“AI 代理编程语言”。在开源环境中，你可以轻松创建全栈应用，就像 Lovable、Bolt 或 Replit 一样。只需使用自然语言作为输入，如果你愿意，还可以将其托管在自己的硬件上。\n\n**无需额外的“后端连接器”或“数据库连接器”！**\n\n**因此，完全无锁定！！**\n\n一切均与 SQLite 完全集成，同时支持可选的 MySQL、PostgreSQL 和 Microsoft SQL Server。基本上，如果你想，完全可以将整个系统运行在你自己的硬件上……\n\n## 开源“氛围编码”平台\n\n以下是一个应用，它是根据如下提示创建的：\n\n> 为一家汽车经销商创建一个管理 VIP 客户的全栈应用\n\n整个过程总共花了约 30 分钟，错误不到一只手就能数过来，过程中仅对大语言模型进行了 5 到 10 次修正或反馈。所有 bug 都由经验丰富的软件开发者轻松定位并消除。\n\n![汽车经销商 CRM 系统](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fpolterguy_magic_readme_9add79b1428b.png)\n\n在自动生成数据库之前，Magic 提出了一些控制性问题；随后基于集成的 Hyperlambda 生成器创建后端代码，最后再根据 API 组装前端——包括身份验证和授权，100% 安全（当然啦！）。所有内容都本地部署在内置的 Web 服务器上。\n\n**一旦保存代码，你就可以直接测试！无需任何“部署”或“发布”操作即可测试代码！**\n\n## 强大的 AI 代理！\n\n以下是 Magic 中的 AI 代理，它完全自主地浏览网页并填写了一个“联系我们”表单。这个具体示例使用了内置的无头浏览器，让 AI 代理能够“看到”网页，自主浏览并完成任务。\n\n![Magic 中的无头浏览器填写表单](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fpolterguy_magic_readme_1d6f0ed9f5b5.png)\n\n不过，你也可以将 AI 代理与你的 CRM 系统、ERP 系统、旧数据库等“任意”系统集成起来。从根本上说，Magic 就是一个用于构建软件和 AI 代理的 AI 代理。至于你用它来做什么，完全取决于你自己。\n\n除了仪表板中的 AI 代理，它能仅凭自然语言输入就生成完整的全栈应用之外，系统还提供了一系列其他组件，帮助你实现软件开发的自动化，比如：\n\n* CRUD 生成器，利用数据库元信息创建 API 端点\n* SQL Studio，让你能够以可视化方式设计和管理你的 SQL 数据库\n* 内置 RBAC\n* Hyper IDE，让你能在类似 VS Code 的环境中手动编辑代码\n* 任务管理器，用于管理和调度任务\n* 机器学习组件，让你能够管理 AI 代理和聊天机器人\n* 插件仓库，支持安装各种类型的前端网站和后端代码\n* 以及更多更多……\n\n以下是 Hyper IDE 的截图。\n\n![Hyper IDE](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fpolterguy_magic_readme_599bf006cd00.png)\n\n上述展示了 Magic 如何助力“注释驱动开发”，即通过声明式注释提供“自然语言指令”，然后由系统利用内置的 AI 代码生成器自动实现代码。\n\n## 同时也是一个 Web 服务器\n\nMagic 本身也是一款 Web 服务器，让你能够即时部署一切，无需编译、构建流程或复杂的管道连接器等等。因此，整个流程如下：\n\n1. 创建你的提示\n2. 按下回车键\n3. 测试！\n\n或者，你还可以使用内置的无头浏览器自动生成 AI 工作流，完成后自动测试你的系统！\n\n这与那些不够成熟的工具形成了鲜明对比，比如 Lovable 和 Bolt44，它们要求你先部署到两个不同的第三方服务商，才能测试代码。因此，在大多数实际场景中，Magic 的开发模式速度大约快了 10 倍，而且更加优化……\n\n除了能够生成纯 JS、CSS 和 HTML 前端并立即提供服务，而无需任何部署管道之外，系统还预置了多个现成的前端系统，比如【AI 专家系统】（https:\u002F\u002Fainiro.io\u002Fai-expert-system），它允许你提供受密码保护的 AI 代理，甚至交付完整的 SaaS AI 解决方案。\n\n该系统尤其适合用于创建 AI 代理。\n\n## 无头浏览器\n\nMagic 内置了一个无头浏览器，特别是 PuppeteerSharp，让你能够像人类一样浏览网页、填写表单、点击按钮等等。\n\n* “访问 xyz 网站，识别他们的‘联系我们’表单，如果需要的话修改 URL，并填写他们的‘联系我们’表单。”\n\n你可以在下面的截图中看到这个提示的示例。\n\n![Magic 中的无头浏览器填写表单](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fpolterguy_magic_readme_1d6f0ed9f5b5.png)\n\n## Git 集成\n\n与其他氛围编码工具不同，Magic Cloud 从第一天起就是专为软件开发者打造的。这意味着它集成了 Git，成为平台不可或缺的一部分。这样，你可以根据需要设置任意数量的管道，使用 Git 管理代码，或使用 GitHub 工作流进行部署。\n\n1. 创建新项目\n2. 使用氛围编码技术配置所有工具，甚至可以配置你的 GitHub 工作流\n3. 提交并推送\n\n以下是集成的 AI 代理客观对比 Magic Cloud 与 Lovable 和 Bolt44 的结果。\n\n![Lovable、Bolt44 与 Magic Cloud 的对比](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fpolterguy_magic_readme_0cfebcf908cc.png)\n\n## Python、终端与C#的集成\n\n即时生成并执行Python脚本，并让大语言模型将这些脚本用作“工具”。此外，您还可以使用BASH和底层终端，并通过C#创建Hyperlambda扩展关键字。\n\n由于Magic默认以受保护的服务账号运行，因此实际上相当安全——不过，显然请勿开放端点，允许第三方用户生成并执行任意Python代码。\n\n您还可以持久化Python脚本，并在之后将其作为“工具”引用，从而永久扩展AI代理的功能；或者，您甚至可以将Python执行集成到您的端点和服务中。\n\n![从Magic Cloud执行Python](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fpolterguy_magic_readme_2b2eb0d7f8c4.png)\n\n**注意**——显然，您必须以root身份登录，才能生成和执行Python脚本、终端脚本以及创建C#扩展。不过，Magic拥有独特的安全模型，能够彻底消除各类与安全相关的“漏洞”。但您仍需保持清醒的头脑。Magic可不是（双关！）“灵丹妙药”。\n\n## 随处部署\n\n如果您选择创建AI代理而非全栈应用——这正是该系统特别擅长的领域——您可以将它们作为受密码保护的AI专家系统提供，或嵌入式AI聊天机器人，内嵌于某个网站。以下是我们的AI聊天机器人，您可在此试用[这里](https:\u002F\u002Fainiro.io)。\n\n![可嵌入的AI聊天机器人](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fpolterguy_magic_readme_ad6ccabddb80.png)\n\n## 比Python快20倍\n\n当我们对比Hyperlambda与Magic Cloud时，其速度大约是用Python构建的类似解决方案（如Fast API或Flask）的20倍左右。与LangChain相比，它的速度可能快50倍，而且由于能用英语编写后端代码，而不是“拖拽式的所见即所得地狱”，创建工作流也变得简单得多。总体而言，Hyperlambda解决方案在可扩展性和性能上与C#结合Entity Framework不相上下。以下是Hyperlambda与Fast API和Flask的对比。\n\n![Python与Hyperlambda的对比](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fpolterguy_magic_readme_575371dfbd26.png)\n\nMagic Cloud基于C#和.NET Core 10构建。\n\n平均而言，Magic的速度可能是“图形化工作流解决方案”（如N8N、LangChain、Zapier、Make等）的100到1000倍——这是因为Magic依赖的是真正的编程语言，而非基于JSON、XML或Markdown的“工作流文件”。执行一段Hyperlambda代码的速度，大约是解析YAMNL或JSON文件中的“动态逻辑”的1000倍。\n\nHyperlambda几乎与纯C#代码不相上下！\n\n## 开始使用\n\n最简单的入门方式是使用Docker，并创建一个包含以下内容的_docker-compose.yaml_文件：\n\n```yaml\nversion: \"3.8\"\n\nservices:\n  backend:\n    image: servergardens\u002Fmagic-backend:latest\n    platform: linux\u002Famd64\n    container_name: magic_backend\n    restart: unless-stopped\n\n    ports:\n      - \"4444:4444\"\n\n    volumes:\n      - magic_files_etc:\u002Fmagic\u002Ffiles\u002Fetc\n      - magic_files_data:\u002Fmagic\u002Ffiles\u002Fdata\n      - magic_files_config:\u002Fmagic\u002Ffiles\u002Fconfig\n      - magic_files_modules:\u002Fmagic\u002Ffiles\u002Fmodules\n\n  frontend:\n    image: servergardens\u002Fmagic-frontend:latest\n    container_name: magic_frontend\n    restart: unless-stopped\n\n    depends_on:\n      - backend\n\n    ports:\n      - \"5555:80\"\n\nvolumes:\n  magic_files_etc:\n  magic_files_data:\n  magic_files_config:\n  magic_files_modules:\n```\n\n将它保存到某个位置，然后执行`docker compose up`或其他命令，访问`localhost:5555`，使用“root”\u002F“root”登录并配置系统。您可[在此阅读更多内容](https:\u002F\u002Fdocs.ainiro.io\u002Fgetting-started\u002F)，了解其他替代方案，比如直接在本地机器上运行代码库。\n\n您也可以观看我[在这里指导您完成设置过程](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=k6eSKxc6oM8)。\n\n### 自行提供OpenAI API密钥\n\n要使用该系统，您需要一个OpenAI API密钥。您可[在此创建](https:\u002F\u002Fplatform.openai.com\u002Fapi-keys)。\n\n**注意**——要获得`gpt-5.4`的访问权限，您可能需要向OpenAI API账户存入51美元。Magic依赖于OpenAI，如果不向OpenAI存钱，您就无法获得gpt-5.4的访问权限，而gpt-5.4正是Magic中用于“氛围编码”的默认模型。您或许能在氛围编码期间使用GPT-4.1，但5.4版本__好太多了__！\n\n如果您对OpenAI完全过敏，系统还提供了Ollama和HuggingFace插件，让您能够用Ollama或HuggingFace模型和端点“覆盖”推理功能——不过，嵌入向量只能通过OpenAI的API生成。\n\n### DIY家庭云\n\nMagic非常容易安装在Mac Mini等设备上，只需使用Docker镜像即可。通过与CloudFlare隧道结合，您可在几分钟内搭建起一个Web服务器，从家中为应用和数据提供服务。下面这个链接就是从我在塞浦路斯拉纳卡的家中运行的，使用了CloudFlare隧道。我们已在美国、挪威及众多国家进行了测试，考虑到连接质量，响应速度还算不错。\n\n* [由DIY Web服务器提供的CRM仪表板](https:\u002F\u002Fhome.ainiro.io\u002Fanalytics-crm)\n\n以下是系统的截图。\n\n![Analytics CRM仪表板](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fpolterguy_magic_readme_e1d690ce517e.png)\n\n利用这样的部署方式，您可以完全绕过任何审查或监控，从而彻底避免内容或应用被监视或审查。\n\n## 大语言模型\n\n该系统内部使用的是OpenAI的gpt-5.4，且已开启最小化推理模式——但所有参数均可调整，只需稍加努力，您便可将内置默认设置替换为Ollama或Hugging Face模型。不过，Hyperlambda生成器的训练数据集并未公开，我们也没有计划公开。这意味着，在最坏情况下，您依然可以完美运行已生成的系统，而无需再生成新系统——即便您因某种原因丢失了Hyperlambda生成器，也无妨。\n\n然而，Hyperlambda生成器本身相当独特，这得益于Hyperlambda集成的安全模型，它能够动态地即时生成工具，并在后端安全地执行生成的代码。这一点已在我们的[自然语言API](https:\u002F\u002Fainiro.io\u002Fnatural-language-api)中得到验证。\n\n以下是来自一个公开页面（自然语言API）的截图，我们接受任何随机访问者的输入。输入随后会通过我们的大语言模型转换为Hyperlambda格式，然后在我们的DMZ军事化区域内的__进程内__执行。我们曾向黑客们提出，如果他们能以某种方式利用该接口获取个人身份信息或从中提取敏感信息，就奖励100美元。迄今为止，尚无人向我们索要这笔奖金。这项悬赏已经持续了3个月，目前仍未有人成功得手。\n\n![自然语言API](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fpolterguy_magic_readme_42e53b1a3d41.png)\n\n如果你能攻破上述API接口，我将给你100美元！\n\n## 独特的安全模型\n\nHyperlambda的核心在于，它首先运行于沙箱环境中，因此无法访问自身沙箱之外的文件系统。此外，它还具备根据内置RBAC系统对单个函数进行白名单管理的能力，允许您的服务器接受代码作为输入，却仍能安全地执行——甚至无需知晓其来源。\n\n这种功能之所以可行，是因为Hyperlambda在执行层面限制了函数调用，据我所知，目前全球只有Hyperlambda这一种编程语言能做到这一点。\n\n这使得Hyperlambda特别适合需要“按需生成工具”的AI代理，因为解决方案的所有者可以指定服务器词汇表中的子集作为“合法函数”，而其他函数则一律禁止调用。\n\n这样一来，您便能交付能够__按需__创建自身工具的AI代理——既不会扩大攻击面，也不会留下安全漏洞。\n\n## 技术\n\nMagic Cloud基于.NET Core 10构建，其仪表板采用Angular开发。而Hyperlambda完全由我本人发明并打造，您可在下方找到一些介绍其独特技术的文章：\n\n* [用Hyperlambda让C#更具动态性](https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Farchive\u002Fmsdn-magazine\u002F2017\u002Fjune\u002Fcsharp-make-csharp-more-dynamic-with-hyperlambda)\n* [主动事件：一种设计模式胜过一打](https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Farchive\u002Fmsdn-magazine\u002F2017\u002Fmarch\u002Fpatterns-active-events-one-design-pattern-instead-of-a-dozen)\n\n不过，Hyperlambda及其关联的Magic Cloud，均基于一种独特的设计模式——“主动事件”或“槽与信号”。这是一种用于执行“动态函数”的进程内模型，是Magic Cloud独有的特色。主动事件正是Hyperlambda的核心所在，它彻底消除了所有跨项目依赖，从而实现了100%的“完美”封装与内聚。\n\n正是上述设计模式与Hyperlambda的结合，成就了Magic Cloud如此极致的安全水平。我们可以放心地相信，任何由AI生成的代码都不会做出有害行为——因为它根本就没有权限执行恶意操作——除非有人明确赋予它这样的权限。\n\n> 我对其代码库的质量充满信心，如果您能在它的后端代码中发现（严重安全）漏洞，我将奖励您100美元！\n\n## 维护\n\nMagic Cloud和Hyperlambda由[AINIRO.IO](https:\u002F\u002Fainiro.io)负责开发与维护。我们不仅提供Magic Cloud的托管、支持与软件开发服务，还致力于交付AI代理、聊天机器人及各类AI解决方案。\n\n## 许可协议\n\n本项目及其所有附属项目均采用开源倡议发布的MIT许可协议授权。详情请参阅LICENSE文件。如有关于许可的疑问，欢迎联系托马斯·汉森，邮箱：thomas@ainiro.io。\n\n## 版权与维护\n\n本项目版权归托马斯·汉森所有，自2019年至2025年受[AINIRO.IO](https:\u002F\u002Fainiro.io)专业维护。","# Magic Cloud 快速上手指南\n\n## 环境准备\n\n- **系统要求**：支持 Linux、macOS、Windows（推荐使用 Docker 环境）\n- **前置依赖**：\n  - Docker 和 docker-compose\n  - OpenAI API Key（[申请地址](https:\u002F\u002Fplatform.openai.com\u002Fapi-keys)）\n  - 可选：国内用户可使用阿里云\u002F腾讯云镜像加速 Docker 拉取\n\n> 注：Magic 默认使用 `gpt-5.4` 模型，需在 OpenAI 账户充值 $51 才能启用。如需规避 OpenAI，可使用 Ollama 或 HuggingFace 替代推理模型（但嵌入仍需 OpenAI）。\n\n## 安装步骤\n\n1. 创建 `docker-compose.yaml` 文件：\n\n```yaml\nversion: \"3.8\"\n\nservices:\n  backend:\n    image: servergardens\u002Fmagic-backend:latest\n    platform: linux\u002Famd64\n    container_name: magic_backend\n    restart: unless-stopped\n\n    ports:\n      - \"4444:4444\"\n\n    volumes:\n      - magic_files_etc:\u002Fmagic\u002Ffiles\u002Fetc\n      - magic_files_data:\u002Fmagic\u002Ffiles\u002Fdata\n      - magic_files_config:\u002Fmagic\u002Ffiles\u002Fconfig\n      - magic_files_modules:\u002Fmagic\u002Ffiles\u002Fmodules\n\n  frontend:\n    image: servergardens\u002Fmagic-frontend:latest\n    container_name: magic_frontend\n    restart: unless-stopped\n\n    depends_on:\n      - backend\n\n    ports:\n      - \"5555:80\"\n\nvolumes:\n  magic_files_etc:\n  magic_files_data:\n  magic_files_config:\n  magic_files_modules:\n```\n\n2. 在终端执行：\n\n```bash\ndocker compose up -d\n```\n\n3. 等待服务启动后，访问：[http:\u002F\u002Flocalhost:5555](http:\u002F\u002Flocalhost:5555)\n\n4. 登录默认账户：`root` \u002F `root`\n\n5. 进入设置页面，填写你的 OpenAI API Key。\n\n## 基本使用\n\n1. 在首页输入自然语言提示，例如：\n\n> Create me a full stack app to manage VIP customer for a car dealership\n\n2. 按回车，Magic 将自动：\n   - 生成数据库结构（SQLite）\n   - 生成后端 API（Hyperlambda）\n   - 生成前端界面（JS\u002FHTML\u002FCSS）\n   - 集成认证与授权\n\n3. 代码生成后，**无需部署**，直接点击“Test”即可在本地运行完整应用。\n\n4. 可通过内置 Hyper IDE 手动编辑代码，或使用 Headless Browser 自动测试表单提交、网页交互等 AI 任务。\n\n> 所有功能均在本地运行，支持一键 Git 提交，可对接 GitHub Actions 实现自动化部署。","一家中小型汽车经销商的IT专员小李，负责为VIP客户管理开发一个内部CRM系统，但团队只有他一人，且无后端开发经验。\n\n### 没有 magic 时\n- 需要花两周时间学习数据库设计、API开发和前端框架，进度严重滞后。\n- 为连接客户数据，必须手动配置MySQL、编写REST接口、搭建认证系统，每个环节都容易出错。\n- 前端页面需用React或Vue手动开发，与后端联调时频繁出现字段不匹配、权限漏洞等问题。\n- 部署时要配置Nginx、Docker、端口映射，服务器环境搭建耗时且易出错。\n- 修改需求时（如新增“试驾记录”字段），需重写多个文件，测试周期长达数天。\n\n### 使用 magic 后\n- 小李仅用一句话：“创建一个管理VIP客户的全栈系统，包含客户信息、试驾记录、销售跟进和权限控制”，magic 30分钟自动生成完整应用。\n- 数据库、API、认证、RBAC权限系统全部自动生成，无需手动配置任何连接器或中间件，直接基于内置SQLite运行。\n- 前端界面自动适配后端API，支持拖拽调整布局，修改字段后刷新即生效，无需重新编译或部署。\n- 内置Web服务器一键启动，本地测试无需Docker或Nginx，修改代码后直接保存即可预览。\n- 当老板要求增加“客户画像分析”功能时，小李在Hyper IDE中添加注释“根据购车历史生成客户价值评分”，magic 自动添加ML模型和前端图表。\n\nmagic 让非专业开发者也能在几小时内交付企业级全栈应用，彻底打破“开发=团队+时间+复杂流程”的传统桎梏。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fpolterguy_magic_1d6f0ed9.png","polterguy","AINIRO.IO","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fpolterguy_ba17fddc.png","AI-based Low-Code and No-Code Software Development Automation","Cyprus",null,"AIniroTeam","https:\u002F\u002Fainiro.io","https:\u002F\u002Fgithub.com\u002Fpolterguy",[85,89,93,97,101,105,109],{"name":86,"color":87,"percentage":88},"C#","#178600",53,{"name":90,"color":91,"percentage":92},"TypeScript","#3178c6",18.8,{"name":94,"color":95,"percentage":96},"CSS","#663399",12.7,{"name":98,"color":99,"percentage":100},"HTML","#e34c26",9.8,{"name":102,"color":103,"percentage":104},"JavaScript","#f1e05a",3.9,{"name":106,"color":107,"percentage":108},"SCSS","#c6538c",1.6,{"name":110,"color":111,"percentage":112},"Shell","#89e051",0.1,1124,167,"2026-04-05T09:48:23","MIT","Linux, macOS, Windows","未说明",{"notes":120,"python":121,"dependencies":122},"系统基于 .NET Core 10 和 C# 构建，主要依赖 Docker 部署；需 OpenAI API 密钥以使用 gpt-5.4 模型；支持通过 Ollama 或 HuggingFace 替代 LLM，但嵌入向量仍需 OpenAI；建议使用 Docker-compose 快速部署，本地运行需配置卷挂载；执行 Python 脚本需以 root 权限运行，存在安全风险需谨慎；支持在 Mac Mini 等家用设备上部署并通过 Cloudflare 隧道对外服务。","3.8+",[],[14,15,13],[125,126,127,128,129,130],"ai","openai","automation-framework","low-code","no-code","vibe-coding","2026-03-27T02:49:30.150509","2026-04-06T06:44:23.492442",[134,139,144,149,154,159],{"id":135,"question_zh":136,"answer_zh":137,"source_url":138},9144,"前端发起 POST 请求时出现 CORS 错误，如何解决？","当请求的 credentials mode 为 'include' 时，服务器响应头 'Access-Control-Allow-Origin' 不能使用通配符 '*'。需在后端配置中明确指定前端域名，例如设置为 'http:\u002F\u002Flocalhost:3000'，而不是允许所有来源。同时确保后端在预检请求（OPTIONS）中返回正确的 'Access-Control-Allow-Credentials: true' 和 'Access-Control-Allow-Origin' 头。","https:\u002F\u002Fgithub.com\u002Fpolterguy\u002Fmagic\u002Fissues\u002F191",{"id":140,"question_zh":141,"answer_zh":142,"source_url":143},9145,"下载的前端文件权限错误，如何修复？","ZIP 文件本身不支持存储 Unix 文件权限，因此解压后文件权限会继承当前用户默认设置。目前无法通过工具自动修复，建议手动执行命令：chmod -R 644 \u003C目录> 为文件设置读权限，chmod -R 755 \u003C目录> 为目录设置执行权限。未来版本可能通过 JSZip 的 unixPermissions 参数改进，但当前无内置解决方案。","https:\u002F\u002Fgithub.com\u002Fpolterguy\u002Fmagic\u002Fissues\u002F219",{"id":145,"question_zh":146,"answer_zh":147,"source_url":148},9146,"注册用户后收不到确认邮件，如何排查？","尽管 SMTP 测试邮件能成功发送，但注册流程可能未正确使用配置的 SMTP 设置。请检查注册请求中是否包含正确的 'frontendUrl' 字段，该字段用于生成验证链接。同时确认后端配置文件中 SMTP 设置是否被正确加载，而非仅在 Evaluator 模块中生效。建议参考官方文档：https:\u002F\u002Fdocs.aista.com\u002Ftutorials\u002Fregistering\u002F。","https:\u002F\u002Fgithub.com\u002Fpolterguy\u002Fmagic\u002Fissues\u002F190",{"id":150,"question_zh":151,"answer_zh":152,"source_url":153},9147,"训练模型时传入 'epochs' 参数报错 'Additional properties are not allowed'，如何解决？","当前系统不支持直接通过 'epochs' 参数指定训练轮数。该参数不在允许的请求结构中，需使用系统默认值或通过其他配置方式（如模型配置文件）调整训练参数。建议查阅最新 API 文档或使用 Dashboard 界面进行训练配置，避免手动传递不支持的字段。","https:\u002F\u002Fgithub.com\u002Fpolterguy\u002Fmagic\u002Fissues\u002F218",{"id":155,"question_zh":156,"answer_zh":157,"source_url":158},9148,"如何让聊天机器人捕获用户邮箱和姓名并发送邮件给客服？","已在系统中实现该功能。用户可向机器人提问如 \"how do I contact you?\"，机器人将自动提示输入姓名和邮箱，并通过后端 API 将信息发送至客服邮箱。请确保已正确配置 SMTP 设置，并在机器人对话流程中启用联系表单模块。可访问 https:\u002F\u002Fainiro.io 测试该功能。","https:\u002F\u002Fgithub.com\u002Fpolterguy\u002Fmagic\u002Fissues\u002F221",{"id":160,"question_zh":161,"answer_zh":162,"source_url":163},9149,"Aista Magic 是否支持导出 API 为 OpenAPI 3.0 规范？","目前不支持自动生成 OpenAPI 3.0 规范，因为 Magic 只有一个统一的 API 端点，无法直接映射为 Swagger 格式。但未来计划支持，可通过开发独立插件（C# 或 Hyperlambda）实现。如需此功能，可自行开发并提交高质量 PR，维护团队欢迎贡献。","https:\u002F\u002Fgithub.com\u002Fpolterguy\u002Fmagic\u002Fissues\u002F189",[165,169,173,177,181,185,190,194,199,204,209,214,219,224,229,234,239,244,249,253],{"id":166,"version":167,"summary_zh":80,"released_at":168},106556,"v22.8.1","2026-04-05T06:49:20",{"id":170,"version":171,"summary_zh":80,"released_at":172},106557,"v22.8.0","2026-04-04T07:11:20",{"id":174,"version":175,"summary_zh":80,"released_at":176},106558,"v22.7.34","2026-04-03T17:04:56",{"id":178,"version":179,"summary_zh":80,"released_at":180},106559,"v22.7.33","2026-04-03T12:14:20",{"id":182,"version":183,"summary_zh":80,"released_at":184},106560,"v22.7.32","2026-04-03T11:35:43",{"id":186,"version":187,"summary_zh":188,"released_at":189},106561,"v22.7.31","Allowing the LLM to autonomously attach images and files it can analyse on a need to know basis","2026-04-03T09:54:12",{"id":191,"version":192,"summary_zh":80,"released_at":193},106562,"v22.7.30","2026-04-03T08:48:20",{"id":195,"version":196,"summary_zh":197,"released_at":198},106563,"v22.7.29","Now supporting all files OpenAI support, and also injecting local filename, giving the LLM visibility into file paths.","2026-04-03T07:46:15",{"id":200,"version":201,"summary_zh":202,"released_at":203},106564,"v22.7.28","Allowing you to log in once, use the same data directory, and have the browser start in \"logged in mode\"","2026-04-02T14:35:54",{"id":205,"version":206,"summary_zh":207,"released_at":208},106565,"v22.7.27","Avoiding opening up a new tab when we create Puppeteer session.","2026-04-02T10:02:57",{"id":210,"version":211,"summary_zh":212,"released_at":213},106566,"v22.7.26","Most importantly, we can't deal with `{{` characters in workflows, so I changed the template substitution logic to look for `[[` instead, plus other improvements to `patch-file`, and also significantly improved `get-context`, etc, etc, etc ...","2026-04-01T12:17:38",{"id":215,"version":216,"summary_zh":217,"released_at":218},106567,"v22.7.24","1. Backend exposes errors in AI functions\r\n2. get-context is no longer authorised (other models might need it)\r\n3. Support for arguments to widgets\r\n4. Updating prompting in workflows\r\n5. Fixed direct lookup bug in get-context\r\n6. Terminal slots will now resolve root folder correctly (working-directory)\r\n7. Add a bunch of design components (I'm not really sure I'll keep, so don't get used to them)\r\n8. Stronger patch-file implementation with more fault tolerance","2026-03-31T10:19:43",{"id":220,"version":221,"summary_zh":222,"released_at":223},106568,"v22.7.22","By saving files locally, and giving access to LLM through image and file URLs.\r\n\r\nNotice, requires .config to serve PDF files as `application\u002Fpdf`","2026-03-29T17:56:04",{"id":225,"version":226,"summary_zh":227,"released_at":228},106569,"v22.7.21","In addition to some design profiles, you can either use as templates for frontend designs, or use as is","2026-03-29T16:03:56",{"id":230,"version":231,"summary_zh":232,"released_at":233},106570,"v22.7.20","Optimising usage of tokens significantly!","2026-03-28T07:06:23",{"id":235,"version":236,"summary_zh":237,"released_at":238},106571,"v22.7.19","One file missing, added in this version.","2026-03-27T15:41:37",{"id":240,"version":241,"summary_zh":242,"released_at":243},106572,"v22.7.18","Allowing you to create your own ChatGPT GUI using natural language.","2026-03-27T11:01:40",{"id":245,"version":246,"summary_zh":247,"released_at":248},106573,"v22.7.17","By persisting these as a part of the source and importing during initialisation","2026-03-26T09:53:06",{"id":250,"version":251,"summary_zh":80,"released_at":252},106574,"v22.7.16","2026-03-19T15:51:27",{"id":254,"version":255,"summary_zh":256,"released_at":257},106575,"v22.7.15","Mostly to prompting","2026-03-17T00:35:40"]