[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-danielmiessler--Personal_AI_Infrastructure":3,"tool-danielmiessler--Personal_AI_Infrastructure":61},[4,18,26,36,44,53],{"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",142651,2,"2026-04-06T23:34:12",[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":32,"last_commit_at":50,"category_tags":51,"status":17},4721,"markitdown","microsoft\u002Fmarkitdown","MarkItDown 是一款由微软 AutoGen 团队打造的轻量级 Python 工具，专为将各类文件高效转换为 Markdown 格式而设计。它支持 PDF、Word、Excel、PPT、图片（含 OCR）、音频（含语音转录）、HTML 乃至 YouTube 链接等多种格式的解析，能够精准提取文档中的标题、列表、表格和链接等关键结构信息。\n\n在人工智能应用日益普及的今天，大语言模型（LLM）虽擅长处理文本，却难以直接读取复杂的二进制办公文档。MarkItDown 恰好解决了这一痛点，它将非结构化或半结构化的文件转化为模型“原生理解”且 Token 效率极高的 Markdown 格式，成为连接本地文件与 AI 分析 pipeline 的理想桥梁。此外，它还提供了 MCP（模型上下文协议）服务器，可无缝集成到 Claude Desktop 等 LLM 应用中。\n\n这款工具特别适合开发者、数据科学家及 AI 研究人员使用，尤其是那些需要构建文档检索增强生成（RAG）系统、进行批量文本分析或希望让 AI 助手直接“阅读”本地文件的用户。虽然生成的内容也具备一定可读性，但其核心优势在于为机器",93400,"2026-04-06T19:52:38",[52,14],"插件",{"id":54,"name":55,"github_repo":56,"description_zh":57,"stars":58,"difficulty_score":10,"last_commit_at":59,"category_tags":60,"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":62,"github_repo":63,"name":64,"description_en":65,"description_zh":66,"ai_summary_zh":67,"readme_en":68,"readme_zh":69,"quickstart_zh":70,"use_case_zh":71,"hero_image_url":72,"owner_login":73,"owner_name":74,"owner_avatar_url":75,"owner_bio":76,"owner_company":77,"owner_location":78,"owner_email":79,"owner_twitter":73,"owner_website":80,"owner_url":81,"languages":82,"stars":118,"forks":119,"last_commit_at":120,"license":121,"difficulty_score":32,"env_os":122,"env_gpu":123,"env_ram":123,"env_deps":124,"category_tags":129,"github_topics":130,"view_count":32,"oss_zip_url":135,"oss_zip_packed_at":135,"status":17,"created_at":136,"updated_at":137,"faqs":138,"releases":171},4804,"danielmiessler\u002FPersonal_AI_Infrastructure","Personal_AI_Infrastructure","Agentic AI Infrastructure for magnifying HUMAN capabilities.","Personal_AI_Infrastructure（简称 PAI）是一套旨在放大人类能力的个人智能体基础设施。它的核心理念是让每个人都能轻松拥有并掌控属于自己的顶级 AI 技术栈，从而将人工智能从云端服务转化为真正增强个人效率的本地化伙伴。\n\n该项目主要解决了普通用户在部署和管理复杂 AI 系统时面临的技术门槛高、配置繁琐以及数据隐私难以保障等痛点。通过提供一套标准化的安装流程和模块化组件，PAI 让用户无需深厚的技术背景即可搭建起功能强大的个人 AI 环境，实现从“使用 AI 工具”到“拥有 AI 能力”的转变。\n\n虽然它由开发者 Daniel Miessler 创建并深受技术人员喜爱，但其设计初衷是服务于所有希望提升工作效率的普通用户、创作者及知识工作者。只要对利用 AI 辅助工作或学习感兴趣，无论是否具备编程经验，都能从中受益。\n\n在技术亮点方面，PAI 采用现代技术栈构建，基于 TypeScript 和 Bun 运行时，确保了系统的高效与轻量。项目强调社区驱动与开放性，支持跨平台（包括 Linux）运行，并持续集成社区贡献的修复与功能扩展。它不仅是一个软件包，更是一个倡导\"A","Personal_AI_Infrastructure（简称 PAI）是一套旨在放大人类能力的个人智能体基础设施。它的核心理念是让每个人都能轻松拥有并掌控属于自己的顶级 AI 技术栈，从而将人工智能从云端服务转化为真正增强个人效率的本地化伙伴。\n\n该项目主要解决了普通用户在部署和管理复杂 AI 系统时面临的技术门槛高、配置繁琐以及数据隐私难以保障等痛点。通过提供一套标准化的安装流程和模块化组件，PAI 让用户无需深厚的技术背景即可搭建起功能强大的个人 AI 环境，实现从“使用 AI 工具”到“拥有 AI 能力”的转变。\n\n虽然它由开发者 Daniel Miessler 创建并深受技术人员喜爱，但其设计初衷是服务于所有希望提升工作效率的普通用户、创作者及知识工作者。只要对利用 AI 辅助工作或学习感兴趣，无论是否具备编程经验，都能从中受益。\n\n在技术亮点方面，PAI 采用现代技术栈构建，基于 TypeScript 和 Bun 运行时，确保了系统的高效与轻量。项目强调社区驱动与开放性，支持跨平台（包括 Linux）运行，并持续集成社区贡献的修复与功能扩展。它不仅是一个软件包，更是一个倡导\"AI 应赋能每一个人”的开放生态，帮助用户构建安全、可控且高度个性化的智能助手系统。","\u003Cdiv align=\"center\">\n\n\u003Cpicture>\n  \u003Csource media=\"(prefers-color-scheme: dark)\" srcset=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_9ff44e3a6ece.png\">\n  \u003Csource media=\"(prefers-color-scheme: light)\" srcset=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_9ff44e3a6ece.png\">\n  \u003Cimg alt=\"PAI Logo\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_9ff44e3a6ece.png\" width=\"300\">\n\u003C\u002Fpicture>\n\n\u003Cbr\u002F>\n\u003Cbr\u002F>\n\n# Personal AI Infrastructure\n\n[![Typing SVG](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_5f3e340731b4.png)](https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure)\n\n\u003Cbr\u002F>\n\n\u003C!-- Social Proof -->\n![Stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=social)\n![Forks](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fforks\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=social)\n![Watchers](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fwatchers\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=social)\n\n\u003C!-- Project Health -->\n![Release](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fv\u002Frelease\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=flat&logo=github&color=8B5CF6)\n![Last Commit](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flast-commit\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=flat&logo=git&color=22C55E)\n![Open Issues](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fissues\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=flat&logo=github&color=F97316)\n![Open PRs](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fissues-pr\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=flat&logo=github&color=EC4899)\n![License](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flicense\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=flat&color=60A5FA)\n\n\u003C!-- Metrics -->\n![Discussions](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdiscussions\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=flat&logo=github&label=Discussions&color=EAB308)\n![Commit Activity](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fcommit-activity\u002Fm\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=flat&logo=git&label=Commits%2Fmo&color=F59E0B)\n![Repo Size](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Frepo-size\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=flat&logo=database&label=Repo%20Size&color=D97706)\n\n\u003C!-- Content -->\n[![Get Started](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F🚀_Get_Started-Install-22C55E?style=flat)](#-installation)\n[![Release v4.0.3](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F📦_Release-v4.0.3-8B5CF6?style=flat)](Releases\u002Fv4.0.3\u002F)\n[![Contributors](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fcontributors\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=flat&logo=githubsponsors&logoColor=white&label=Contributors&color=EC4899)](https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure\u002Fgraphs\u002Fcontributors)\n\n\u003C!-- Tech Stack -->\n[![Built with Claude](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FBuilt_with-Claude-D4A574?style=flat&logo=anthropic&logoColor=white)](https:\u002F\u002Fclaude.ai)\n[![TypeScript](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FTypeScript-3178C6?style=flat&logo=typescript&logoColor=white)](https:\u002F\u002Fwww.typescriptlang.org\u002F)\n[![Bun](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FBun-000000?style=flat&logo=bun&logoColor=white)](https:\u002F\u002Fbun.sh)\n[![Community](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FCommunity-5865F2?style=flat&logo=discord&logoColor=white)](https:\u002F\u002Fdanielmiessler.com\u002Fupgrade)\n\n\u003Cbr\u002F>\n\n**Overview:** [Purpose](#the-purpose-of-this-project) · [What is PAI?](#what-is-pai) · [New to AI?](#new-to-this-start-here) · [Principles](#the-pai-principles) · [Primitives](#pai-primitives)\n\n**Get Started:** [Installation](#-installation) · [Releases](Releases\u002F) · [Packs](Packs\u002F)\n\n**Resources:** [FAQ](#-faq) · [Roadmap](#-roadmap) · [Community](#-community) · [Contributing](#-contributing)\n\n\u003Cbr\u002F>\n\n[![PAI Overview Video](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_977a3ba203b9.jpg)](https:\u002F\u002Fyoutu.be\u002FLe0DLrn7ta0)\n\n**[Watch the full PAI walkthrough](https:\u002F\u002Fyoutu.be\u002FLe0DLrn7ta0)** | **[Read: The Real Internet of Things](https:\u002F\u002Fdanielmiessler.com\u002Fblog\u002Freal-internet-of-things)**\n\n---\n\n\u003C\u002Fdiv>\n\n> [!IMPORTANT]\n> **PAI v4.0.3 Released** — 3 patch updates since v4.0.0 with 30+ community-contributed fixes: Linux compatibility, JSON parsing, installer improvements, portability, and upgrade migration.\n>\n> **[Release notes →](Releases\u002Fv4.0.3\u002FREADME.md)** | **[All releases →](Releases\u002F)**\n\n\u003Cdiv align=\"center\">\n\n# AI should magnify everyone—not just the top 1%.\n\n\u003C\u002Fdiv>\n\n## The Purpose of This Project\n\n**PAI exists to solve what I believe is the [P0 problem](https:\u002F\u002Fdanielmiessler.com\u002Ftelos) in the world:**\n\n### Only a tiny fraction of humanity's creative potential is activated on Earth.\n\nMost people don't believe they have valuable contributions to make. They think there are \"special\" people—and they aren't one of them. They've never asked who they are, what they're about, and have never articulated or written it down. This makes them catastrophically vulnerable to AI displacement. Without activation, there is no high-agency.\n\nSo our goal with PAI is to activate people.\n\n**PAI's mission is twofold:**\n\n1. **Activate as many people as possible** — Help people identify, articulate, and pursue their own purpose in life through AI-augmented self-discovery\n2. **Make the best AI available in the world accessible to everyone** — Ensure this quality of AI infrastructure isn't reserved for just the rich or technical elite.\n\nThat's why this is an open-source project instead of private.\n\n---\n\n## New to This? Start Here\n\nYou've probably used ChatGPT or Claude. Type a question, get an answer. Simple.\n\nYou can think of AI systems as **three levels**:\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_fbef7d388f7a.png\" alt=\"The AI Evolution - From chatbots to your personal AI system\" width=\"800\">\n\u003C\u002Fp>\n\n### Chatbots\n\nChatGPT, Claude, Gemini—you ask something, it answers, and then it forgets everything. Next conversation starts fresh. No memory of you, your preferences, or what you talked about yesterday.\n\n**The pattern:** Ask → Answer → Forget\n\n### Agentic Platforms\n\nTools like Claude Code. The AI can actually *do* things—write code, browse the web, edit files, run commands.\n\n**The pattern:** Ask → Use tools → Get result\n\nMore capable, but it still doesn't know *you*—your goals, your preferences, your history.\n\n### PAI (Personal AI Infrastructure)\n\nNow your DA **learns and improves**:\n- **Captures every signal** — Ratings, sentiment, verification outcomes\n- **Learns from mistakes** — Failures get analyzed and fixed\n- **Gets better over time** — Success patterns get reinforced\n- **Upgrades itself** — Skills, workflows, even the core behavior evolves\n\nPlus it knows:\n- **Your goals** — What you're working toward\n- **Your preferences** — How you like things done\n- **Your history** — Past decisions and learnings\n\n**The pattern:** Observe → Think → Plan → Execute → Verify → **Learn** → Improve\n\nThe key difference: **PAI learns from feedback**. Every interaction makes it better at helping *you* specifically.\n\n---\n\n## What is PAI?\n\nPAI is a Personalized AI Platform designed to magnify your capabilities.\n\nIt's designed for humans most of all, but can be used by teams, companies, or Federations of Planets desiring to be better versions of themselves.\n\nThe scale of the entity doesn't matter: It's a system for understanding, articulating, and realizing its principal's goals using a full-featured Agentic AI Platform.\n\n### Who is PAI for?\n\n**Everyone, full stop.** It's the anti-gatekeeping AI project.\n\n- **Small business owners** who aren't technical but want AI to handle invoicing, scheduling, customer follow-ups, and marketing\n- **Companies** who want to understand their data, optimize operations, and make better decisions\n- **Managers** who want to run their teams more effectively—tracking projects, preparing for reviews, and communicating clearly\n- **Artists and creatives** who want to find local events, galleries, and opportunities to showcase their work\n- **Everyday people** who want to improve their lives—better fitness routines, stronger social connections, personal finance, or just getting organized\n- **Developers** using AI coding assistants who want persistent memory and custom workflows\n- **Power users** who want their AI to know their goals, preferences, and context\n- **Teams** building shared AI infrastructure with consistent capabilities\n- **Experimenters** interested in AI system design and personal AI patterns\n\n### What makes PAI different?\n\nThe first thing people ask is:\n\n> How is this different from Claude Code, or any of the other agentic systems?\n\nMost agentic systems are built around tools with the user being an afterthought. They are also mostly task-based instead of being goal-based using all the context available to them. PAI is the opposite.\n\n**Three core differentiators:**\n\n1. **Goal Orientation** — PAI's primary focus is on the human running it and what they're trying to do in the world, not the tech. This is built into how the system executes all tasks.\n\n2. **Pursuit of Optimal Output** — The system's outer loop and everything it does is trying to produce the exact right output given the current situation and all the contexts around it.\n\n3. **Continuous Learning** — The system constantly captures signals about what was done, what changes were made, what outputs were produced for each request, and then how you liked or disliked the results.\n\n---\n\n## The PAI Principles\n\nThese principles guide how PAI systems are designed and built. **[Full breakdown →](https:\u002F\u002Fdanielmiessler.com\u002Fblog\u002Fpersonal-ai-infrastructure)**\n\n| # | Principle | Summary |\n|---|-----------|---------|\n| 1 | **User Centricity** | PAI is built around you, not tooling. Your goals, preferences, and context come first—the infrastructure exists to serve them. |\n| 2 | **The Foundational Algorithm** | The scientific method as a universal problem-solving loop: Observe → Think → Plan → Build → Execute → Verify → Learn. Define the ideal state, iterate until you reach it. |\n| 3 | **Clear Thinking First** | Good prompts come from clear thinking. Clarify the problem before writing the prompt. |\n| 4 | **Scaffolding > Model** | System architecture matters more than which model you use. |\n| 5 | **Deterministic Infrastructure** | AI is probabilistic; your infrastructure shouldn't be. Use templates and patterns. |\n| 6 | **Code Before Prompts** | If you can solve it with a bash script, don't use AI. |\n| 7 | **Spec \u002F Test \u002F Evals First** | Write specifications and tests before building. Measure if the system works. |\n| 8 | **UNIX Philosophy** | Do one thing well. Make tools composable. Use text interfaces. |\n| 9 | **ENG \u002F SRE Principles** | Treat AI infrastructure like production software: version control, automation, monitoring. |\n| 10 | **CLI as Interface** | Command-line interfaces are faster, more scriptable, and more reliable than GUIs. |\n| 11 | **Goal → Code → CLI → Prompts → Agents** | The decision hierarchy: clarify goal, then code, then CLI, then prompts, then agents. |\n| 12 | **Skill Management** | Modular capabilities that route intelligently based on context. |\n| 13 | **Memory System** | Everything worth knowing gets captured. History feeds future context. |\n| 14 | **Agent Personalities** | Different work needs different approaches. Specialized agents with unique voices. |\n| 15 | **Science as Meta-Loop** | Hypothesis → Experiment → Measure → Iterate. |\n| 16 | **Permission to Fail** | Explicit permission to say \"I don't know\" prevents hallucinations. |\n\n---\n\n## PAI Primitives\n\nWhile the Principles describe the *philosophy* of PAI, the Primitives are the *architecture*—the core systems that make everything work.\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_aa7d84f40c6d.png\" alt=\"PAI Primitives - A system that knows you, not a tool harness\" width=\"800\">\n\u003C\u002Fp>\n\nThese primitives work together to create the experience of working with a system that understands and knows you—as opposed to a tool harness that just executes commands.\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_a7d904467cc0.png\" alt=\"Assistant vs Agent-Based Interaction\" width=\"700\">\n\u003C\u002Fp>\n\n### Assistant vs. Agent-Based AI Interaction\n\nPAI treats AI as a [persistent assistant, friend, coach, and mentor](https:\u002F\u002Fdanielmiessler.com\u002Fblog\u002Fpersonal-ai-maturity-model) rather than a stateless agent that runs tasks. An assistant knows your goals, remembers your preferences, and improves over time. An agent executes commands and forgets.\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_995c0de09826.png\" alt=\"TELOS - Deep Goal Understanding\" width=\"700\">\n\u003C\u002Fp>\n\n### TELOS (Deep Goal Understanding)\n\n10 files that capture who you are: MISSION.md, GOALS.md, PROJECTS.md, BELIEFS.md, MODELS.md, STRATEGIES.md, NARRATIVES.md, LEARNED.md, CHALLENGES.md, IDEAS.md. Your DA knows what you're working toward because it's all documented.\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_5dec59e61dfe.png\" alt=\"User\u002FSystem Separation\" width=\"700\">\n\u003C\u002Fp>\n\n### User\u002FSystem Separation\n\nYour customizations live in USER\u002F. PAI infrastructure lives in SYSTEM\u002F. When PAI upgrades, your files are untouched. Portable identity, upgrade-safe.\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_27a0dfbc5fcf.png\" alt=\"Granular Customization\" width=\"700\">\n\u003C\u002Fp>\n\n### Granular Customization\n\nSix layers of customization: Identity (name, voice, personality), Preferences (tech stack, tools), Workflows (how skills execute), Skills (what capabilities exist), Hooks (how events are handled), and Memory (what gets captured). Start with defaults, customize when needed.\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_3ca3a7edb0ad.png\" alt=\"Skill System\" width=\"700\">\n\u003C\u002Fp>\n\n### Skill System\n\nHighly focused on consistent results. It has a structure that puts *deterministic outcomes first* by going from CODE -> CLI-BASED-TOOL -> PROMPT -> SKILL instead of a haphazard structure.\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_2368afac3b09.png\" alt=\"Memory System\" width=\"700\">\n\u003C\u002Fp>\n\n### Memory System\n\nFocused on continuous learning. Every interaction generates signals—ratings, sentiment, successes, failures—that feed back into improving the system. Three-tier architecture (hot\u002Fwarm\u002Fcold) with phase-based learning directories.\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_b0b1f63c109b.png\" alt=\"Hook System\" width=\"700\">\n\u003C\u002Fp>\n\n### Hook System\n\nResponds to lifecycle events—session start, tool use, task completion, and more. 8 event types enable voice notifications, automatic context loading, session capture, security validation, and observability.\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_d04b87fad00e.png\" alt=\"Security System\" width=\"700\">\n\u003C\u002Fp>\n\n### Security System\n\nDefines system and user-level security policies by default. You don't have to run with `--dangerously-skip-permissions` to have an uninterrupted experience. PAI's security hooks validate commands before execution, blocking dangerous operations while allowing normal workflows to proceed smoothly.\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_b1c5dd64011a.png\" alt=\"AI-Based Installation\" width=\"700\">\n\u003C\u002Fp>\n\n### AI-Based Installation\n\nThe GUI installer handles everything—prerequisites, configuration, and setup. No manual configuration, no guessing.\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_09439a021951.png\" alt=\"Notification System\" width=\"700\">\n\u003C\u002Fp>\n\n### Notification System\n\nKeeps you informed without being intrusive. Push notifications via ntfy for mobile alerts, Discord integration for team updates, and duration-aware routing that escalates for long-running tasks. Fire-and-forget design means notifications never block your workflow.\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_3b61fdfc0b2f.png\" alt=\"Voice System\" width=\"700\">\n\u003C\u002Fp>\n\n### Voice System\n\nPowered by ElevenLabs TTS. Hear task completions, session summaries, and important updates spoken aloud. Prosody enhancement makes speech sound natural. Your AI has a voice.\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_2d9a23177c4b.png\" alt=\"Terminal-Based UI\" width=\"700\">\n\u003C\u002Fp>\n\n### Terminal-Based UI\n\nRich tab titles and pane management. Dynamic status lines show learning signals, context usage, and current task state. Your terminal is a command center.\n\n---\n\n## 🚀 Installation\n\n> [!CAUTION]\n> **Project in Active Development** — PAI is evolving rapidly. Expect breaking changes, restructuring, and frequent updates. We are working on stable and development branches, but currently it's all combined.\n\n### Fresh Install\n\n```bash\n# Clone the repo\ngit clone https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure.git\ncd Personal_AI_Infrastructure\u002FReleases\u002Fv4.0.3\n\n# Copy the release and run the installer\ncp -r .claude ~\u002F && cd ~\u002F.claude && bash install.sh\n```\n\n**The installer will:**\n- Detect your system and install prerequisites (Bun, Git, Claude Code)\n- Ask for your name, AI assistant name, timezone, and temperature unit preference\n- Clone\u002Fconfigure the PAI repository into `~\u002F.claude\u002F`\n- Set up voice features with ElevenLabs (optional)\n- Configure your shell alias and verify the installation\n\n**After installation:** Run `source ~\u002F.zshrc && pai` to launch PAI.\n\n### Upgrading from a Previous Version\n\n```bash\n# 1. Back up your current installation\ncp -r ~\u002F.claude ~\u002F.claude-backup-$(date +%Y%m%d)\n\n# 2. Clone and copy the new release over your installation\ngit clone https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure.git\ncd Personal_AI_Infrastructure\u002FReleases\u002Fv4.0.3\ncp -r .claude ~\u002F\n\n# 3. Run the installer (detects existing installation, preserves your data)\ncd ~\u002F.claude && bash install.sh\n\n# 4. Rebuild your CLAUDE.md\nbun ~\u002F.claude\u002FPAI\u002FTools\u002FBuildCLAUDE.ts\n```\n\n> [!TIP]\n> The installer **auto-detects** existing installations. It preserves your `USER\u002F` files, merges `settings.json` (only updating installer-managed fields like identity and version), and never overwrites your hooks, statusline, or custom configuration.\n\n**Post-upgrade checklist:**\n- [ ] Verify your identity in `settings.json` (name, AI name, timezone)\n- [ ] Confirm the statusline displays correctly\n- [ ] Test voice notifications (if enabled)\n- [ ] Run a simple prompt to confirm PAI responds correctly\n\n---\n\n## 📦 PAI Packs\n\nDon't want to install all of PAI? **Packs** are standalone, AI-installable capabilities you can add one at a time. Each pack is self-contained — your AI reads the install guide and sets everything up for you. No PAI installation required.\n\nPoint your AI at any pack and say \"install this\":\n\n```\n\"Install the Research pack from PAI\u002FPacks\u002FResearch\u002F\"\n```\n\nYour AI walks through a 5-phase wizard: system analysis, user questions, backup, installation, verification.\n\n### Available Packs\n\n| Pack | What It Does |\n|------|-------------|\n| [ContextSearch](Packs\u002FContextSearch\u002F) | `\u002Fcontext-search` and `\u002Fcs` — instant recall of prior work sessions |\n| [Agents](Packs\u002FAgents\u002F) | Custom agent composition from traits, voices, and personalities |\n| [ContentAnalysis](Packs\u002FContentAnalysis\u002F) | Wisdom extraction from videos, podcasts, articles, and YouTube |\n| [Investigation](Packs\u002FInvestigation\u002F) | OSINT and investigation — company intel, people search, domain lookup |\n| [Media](Packs\u002FMedia\u002F) | AI image generation, diagrams, infographics, and Remotion video |\n| [Research](Packs\u002FResearch\u002F) | Multi-agent research — quick, standard, extensive, and deep modes |\n| [Scraping](Packs\u002FScraping\u002F) | Web scraping via Bright Data proxy and Apify social media actors |\n| [Security](Packs\u002FSecurity\u002F) | Recon, web app testing, prompt injection testing, security news |\n| [Telos](Packs\u002FTelos\u002F) | Life OS — goals, beliefs, wisdom, project dashboards, McKinsey reports |\n| [Thinking](Packs\u002FThinking\u002F) | First principles, council debates, red team, brainstorming, science |\n| [USMetrics](Packs\u002FUSMetrics\u002F) | 68 US economic indicators from FRED, EIA, Treasury, BLS, Census |\n| [Utilities](Packs\u002FUtilities\u002F) | CLI generation, skill scaffolding, Fabric patterns, Cloudflare, browser automation |\n\nEach pack works standalone — install one, install five, or install all of them. They're designed to give you PAI-level capabilities whether or not you run the full PAI system.\n\n**[Browse all packs →](Packs\u002F)**\n\n---\n\n## ❓ FAQ\n\n### How is PAI different from just using Claude Code?\n\nPAI is built natively on Claude Code and designed to stay that way. We chose Claude Code because its hook system, context management, and agentic architecture are the best foundation available for personal AI infrastructure.\n\nPAI isn't a replacement for Claude Code — it's the layer on top that makes Claude Code *yours*:\n\n- **Persistent memory** — Your DA remembers past sessions, decisions, and learnings\n- **Custom skills** — Specialized capabilities for the things you do most\n- **Your context** — Goals, contacts, preferences—all available without re-explaining\n- **Intelligent routing** — Say \"research this\" and the right workflow triggers automatically\n- **Self-improvement** — The system modifies itself based on what it learns\n\nThink of it this way: Claude Code is the engine. PAI is everything else that makes it *your* car.\n\n### What's the difference between PAI and Claude Code's built-in features?\n\nClaude Code provides powerful primitives — hooks, slash commands, MCP servers, context files. These are individual building blocks.\n\nPAI is the complete system built on those primitives. It connects everything together: your goals inform your skills, your skills generate memory, your memory improves future responses. PAI turns Claude Code's building blocks into a coherent personal AI platform.\n\n### Is PAI only for Claude Code?\n\nPAI is Claude Code native. We believe Claude Code's hook system, context management, and agentic capabilities make it the best platform for personal AI infrastructure, and PAI is designed to take full advantage of those features.\n\nThat said, PAI's concepts (skills, memory, algorithms) are universal, and the code is TypeScript and Bash — so community members are welcome to adapt it for other platforms.\n\n### How is this different from fabric?\n\n[Fabric](https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002Ffabric) is a collection of AI prompts (patterns) for specific tasks. It's focused on *what to ask AI*.\n\nPAI is infrastructure for *how your DA operates*—memory, skills, routing, context, self-improvement. They're complementary. Many PAI users integrate Fabric patterns into their skills.\n\n### What if I break something?\n\nRecovery is straightforward:\n\n- **Back up first** — Before any upgrade: `cp -r ~\u002F.claude ~\u002F.claude-backup-$(date +%Y%m%d)`\n- **USER\u002F is safe** — Your customizations in `USER\u002F` are never touched by the installer or upgrades\n- **Settings merge, not overwrite** — The installer only updates identity and version fields; your hooks, statusline, and custom config are preserved\n- **Git-backed** — Version control everything, roll back when needed\n- **History is preserved** — Your DA's memory survives mistakes\n- **DA can fix it** — Your DA helped build it, it can help repair it\n- **Re-install** — Run the installer again; it detects existing installations and merges intelligently\n\n---\n\n## 🎯 Roadmap\n\n| Feature | Description |\n|---------|-------------|\n| **Local Model Support** | Run PAI with local models (Ollama, llama.cpp) for privacy and cost control |\n| **Granular Model Routing** | Route different tasks to different models based on complexity |\n| **Remote Access** | Access your PAI from anywhere—mobile, web, other devices |\n| **Outbound Phone Calling** | Voice capabilities for outbound calls |\n| **External Notifications** | Robust notification system for Email, Discord, Telegram, Slack |\n\n---\n\n## 🌐 Community\n\n**GitHub Discussions:** [Join the conversation](https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure\u002Fdiscussions)\n\n**Community Discord:** PAI is discussed in the [community Discord](https:\u002F\u002Fdanielmiessler.com\u002Fupgrade) along with other AI projects\n\n**Twitter\u002FX:** [@danielmiessler](https:\u002F\u002Ftwitter.com\u002Fdanielmiessler)\n\n**Blog:** [danielmiessler.com](https:\u002F\u002Fdanielmiessler.com)\n\n### Star History\n\n\u003Ca href=\"https:\u002F\u002Fstar-history.com\u002F#danielmiessler\u002FPersonal_AI_Infrastructure&Date\">\n \u003Cpicture>\n   \u003Csource media=\"(prefers-color-scheme: dark)\" srcset=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_c94f139001b0.png&theme=dark\" \u002F>\n   \u003Csource media=\"(prefers-color-scheme: light)\" srcset=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_c94f139001b0.png\" \u002F>\n   \u003Cimg alt=\"Star History Chart\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_c94f139001b0.png\" \u002F>\n \u003C\u002Fpicture>\n\u003C\u002Fa>\n\n---\n\n## 🤝 Contributing\n\nWe welcome contributions! See our [GitHub Issues](https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure\u002Fissues) for open tasks.\n\n1. **Fork the repository**\n2. **Make your changes** — Bug fixes, new skills, documentation improvements\n3. **Test thoroughly** — Install in a fresh system to verify\n4. **Submit a PR** with examples and testing evidence\n\n---\n\n## 📜 License\n\nMIT License - see [LICENSE](LICENSE) for details.\n\n---\n\n## 🙏 Credits\n\n**Anthropic and the Claude Code team** — First and foremost. You are moving AI further and faster than anyone right now. Claude Code is the foundation that makes all of this possible.\n\n**[IndyDevDan](https:\u002F\u002Fwww.youtube.com\u002F@indydevdan)** — For great videos on meta-prompting and custom agents that have inspired parts of PAI.\n\n### Contributors\n\n**[fayerman-source](https:\u002F\u002Fgithub.com\u002Ffayerman-source)** — Google Cloud TTS provider integration and Linux audio support for the voice system.\n\n**Matt Espinoza** — Extensive testing, ideas, and feedback for the PAI 2.3 release, plus roadmap contributions.\n\n---\n\n## 💜 Support This Project\n\n\u003Cdiv align=\"center\">\n\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fsponsors\u002Fdanielmiessler\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FSponsor-❤️-EA4AAA?style=for-the-badge&logo=github-sponsors&logoColor=white\" alt=\"Sponsor\">\u003C\u002Fa>\n\n**PAI is free and open-source forever. If you find it valuable, you can [sponsor the project](https:\u002F\u002Fgithub.com\u002Fsponsors\u002Fdanielmiessler).**\n\n\u003C\u002Fdiv>\n\n---\n\n## 📚 Related Reading\n\n- [The Real Internet of Things](https:\u002F\u002Fdanielmiessler.com\u002Fblog\u002Freal-internet-of-things) — The vision behind PAI\n- [AI's Predictable Path: 7 Components](https:\u002F\u002Fdanielmiessler.com\u002Fblog\u002Fai-predictable-path-7-components-2024) — Visual walkthrough of where AI is heading\n- [Building a Personal AI Infrastructure](https:\u002F\u002Fdanielmiessler.com\u002Fblog\u002Fpersonal-ai-infrastructure) — Full PAI walkthrough with examples\n\n---\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>📜 Update History\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n\u003Cbr\u002F>\n\n**v4.0.3 (2026-03-01) — Community PR Patch**\n- JSON array parsing fix in Inference.ts\n- 29 dead references removed from CONTEXT_ROUTING.md\n- WorldThreatModelHarness PAI_DIR portability\n- User context migration for v2.5\u002Fv3.0 upgraders\n- [Release Notes](Releases\u002Fv4.0.3\u002FREADME.md)\n\n**v4.0.2 (2026-03-01) — Bug Fix Patch**\n- 13 surgical fixes: Linux compatibility, installer, statusline, hooks\n- Cross-platform OAuth token extraction, GNU coreutils tr fix\n- Inference guard (~15s savings), lineage tracking, dead code removal\n- [Release Notes](Releases\u002Fv4.0.2\u002FREADME.md)\n\n**v4.0.1 (2026-02-28) — Upgrade Path & Preferences**\n- Upgrade documentation with backup, merge, and post-upgrade checklist\n- Configurable temperature unit (Fahrenheit\u002FCelsius) in statusline and installer\n- FAQ fixes: removed stale Python reference, improved recovery guidance\n- [Release Notes](Releases\u002Fv4.0.1\u002FREADME.md)\n\n**v4.0.0 (2026-02-27) — Lean and Mean**\n- 38 flat skill directories → 12 hierarchical categories (-68% top-level dirs)\n- Dead systems removed: Components\u002F, DocRebuild, RebuildSkill\n- CLAUDE.md template system with BuildCLAUDE.ts + SessionStart hook\n- Algorithm v3.5.0 (up from v1.4.0)\n- Comprehensive security sanitization (33+ files cleaned)\n- All version refs updated, Electron crash fix\n- 63 skills, 21 hooks, 180 workflows, 14 agents\n- [Release Notes](Releases\u002Fv4.0.0\u002FREADME.md)\n\n**v3.0.0 (2026-02-15) — The Algorithm Matures**\n- Algorithm v1.4.0 with constraint extraction and build drift prevention\n- Persistent PRDs and parallel loop execution\n- Full installer with GUI wizard\n- 10 new skills, agent teams\u002Fswarm, voice personality system\n- 38 skills, 20 hooks, 162 workflows\n- [Release Notes](Releases\u002Fv3.0\u002FREADME.md)\n\n**v2.5.0 (2026-01-30) — Think Deeper, Execute Faster**\n- Two-Pass Capability Selection: Hook hints validated against ISC in THINK phase\n- Thinking Tools with Justify-Exclusion: Opt-OUT, not opt-IN for Council, RedTeam, FirstPrinciples, etc.\n- Parallel-by-Default Execution: Independent tasks run concurrently via parallel agent spawning\n- 28 skills, 17 hooks, 356 workflows\n- [Release Notes](Releases\u002Fv2.5\u002FREADME.md)\n\n**v2.4.0 (2026-01-23) — The Algorithm**\n- Universal problem-solving system with ISC (Ideal State Criteria) tracking\n- 29 skills, 15 hooks, 331 workflows\n- Euphoric Surprise as the outcome metric\n- Enhanced security with AllowList enforcement\n- [Release Notes](Releases\u002Fv2.4\u002FREADME.md)\n\n**v2.3.0 (2026-01-15) — Full Releases Return**\n- Complete `.claude\u002F` directory releases with continuous learning\n- Explicit and implicit rating capture\n- Enhanced hook system with 14 production hooks\n- Status line with learning signal display\n- [Release Notes](Releases\u002Fv2.3\u002FREADME.md)\n\n**v2.1.1 (2026-01-09) — MEMORY System Migration**\n- History system merged into core as MEMORY System\n\n**v2.1.0 (2025-12-31) — Modular Architecture**\n- Source code in real files instead of embedded markdown\n\n**v2.0.0 (2025-12-28) — PAI v2 Launch**\n- Modular architecture with independent skills\n- Claude Code native design\n\n\u003C\u002Fdetails>\n\n---\n\n\u003Cdiv align=\"center\">\n\n**Built with ❤️ by [Daniel Miessler](https:\u002F\u002Fdanielmiessler.com) and the PAI community**\n\n*Augment yourself.*\n\n\u003C\u002Fdiv>\n","\u003Cdiv align=\"center\">\n\n\u003Cpicture>\n  \u003Csource media=\"(prefers-color-scheme: dark)\" srcset=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_9ff44e3a6ece.png\">\n  \u003Csource media=\"(prefers-color-scheme: light)\" srcset=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_9ff44e3a6ece.png\">\n  \u003Cimg alt=\"PAI Logo\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_9ff44e3a6ece.png\" width=\"300\">\n\u003C\u002Fpicture>\n\n\u003Cbr\u002F>\n\u003Cbr\u002F>\n\n# 个人AI基础设施\n\n[![Typing SVG](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_5f3e340731b4.png)](https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure)\n\n\u003Cbr\u002F>\n\n\u003C!-- 社会认同 -->\n![星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=social)\n![叉子](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fforks\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=social)\n![关注者](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fwatchers\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=social)\n\n\u003C!-- 项目健康 -->\n![发布版本](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fv\u002Frelease\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=flat&logo=github&color=8B5CF6)\n![最近一次提交](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flast-commit\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=flat&logo=git&color=22C55E)\n![未解决的问题](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fissues\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=flat&logo=github&color=F97316)\n![未合并的拉取请求](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fissues-pr\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=flat&logo=github&color=EC4899)\n![许可证](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flicense\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=flat&color=60A5FA)\n\n\u003C!-- 指标 -->\n![讨论](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdiscussions\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=flat&logo=github&label=Discussions&color=EAB308)\n![提交活跃度](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fcommit-activity\u002Fm\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=flat&logo=git&label=Commits%2Fmo&color=F59E0B)\n![仓库大小](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Frepo-size\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=flat&logo=database&label=Repo%20Size&color=D97706)\n\n\u003C!-- 内容 -->\n[![开始使用](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F🚀_Get_Started-Install-22C55E?style=flat)](#-installation)\n[![发布 v4.0.3](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F📦_Release-v4.0.3-8B5CF6?style=flat)](Releases\u002Fv4.0.3\u002F)\n[![贡献者](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fcontributors\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure?style=flat&logo=githubsponsors&logoColor=white&label=Contributors&color=EC4899)](https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure\u002Fgraphs\u002Fcontributors)\n\n\u003C!-- 技术栈 -->\n[![基于Claude构建](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FBuilt_with-Claude-D4A574?style=flat&logo=anthropic&logoColor=white)](https:\u002F\u002Fclaude.ai)\n[![TypeScript](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FTypeScript-3178C6?style=flat&logo=typescript&logoColor=white)](https:\u002F\u002Fwww.typescriptlang.org\u002F)\n[![Bun](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FBun-000000?style=flat&logo=bun&logoColor=white)](https:\u002F\u002Fbun.sh)\n[![社区](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FCommunity-5865F2?style=flat&logo=discord&logoColor=white)](https:\u002F\u002Fdanielmiessler.com\u002Fupgrade)\n\n\u003Cbr\u002F>\n\n**概览:** [项目目的](#the-purpose-of-this-project) · [什么是PAI？](#what-is-pai) · [新手入门](#new-to-this-start-here) · [原则](#the-pai-principles) · [基础组件](#pai-primitives)\n\n**开始使用:** [安装](#-installation) · [发布版本](Releases\u002F) · [套装](Packs\u002F)\n\n**资源:** [常见问题](#-faq) · [路线图](#-roadmap) · [社区](#-community) · [贡献](#-contributing)\n\n\u003Cbr\u002F>\n\n[![PAI概述视频](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_977a3ba203b9.jpg)](https:\u002F\u002Fyoutu.be\u002FLe0DLrn7ta0)\n\n**[观看完整的PAI演示](https:\u002F\u002Fyoutu.be\u002FLe0DLrn7ta0)** | **[阅读：真正的物联网](https:\u002F\u002Fdanielmiessler.com\u002Fblog\u002Freal-internet-of-things)**\n\n---\n\n\u003C\u002Fdiv>\n\n> [!IMPORTANT]\n> **PAI v4.0.3发布** — 自v4.0.0以来已发布3个补丁更新，包含30多项社区贡献的修复：Linux兼容性、JSON解析、安装程序改进、可移植性以及升级迁移。\n>\n> **[发布说明 →](Releases\u002Fv4.0.3\u002FREADME.md)** | **[所有版本 →](Releases\u002F)**\n\n\u003Cdiv align=\"center\">\n\n# AI 应该赋能每一个人——而不仅仅是前1%的人。\n\n\u003C\u002Fdiv>\n\n## 项目的宗旨\n\n**PAI的存在旨在解决我认为世界上存在的[P0问题](https:\u002F\u002Fdanielmiessler.com\u002Ftelos)：**\n\n### 地球上只有极小一部分人类的创造潜力被激发出来。\n\n大多数人并不相信自己能够做出有价值的贡献。他们认为世界上有“特殊”的人——而自己并不是其中之一。他们从未真正问过自己是谁、想要什么，也从未将这些思考清晰地表达或记录下来。这使得他们在面对AI带来的冲击时显得格外脆弱。如果没有自我激活，就无法拥有高度的自主能动性。\n\n因此，我们通过PAI的目标就是帮助人们实现自我激活。\n\n**PAI的使命是双重的：**\n\n1. **尽可能多地激活人们** — 借助AI增强的自我发现能力，帮助人们识别、表达并追求自己的人生目标。\n2. **让世界上最好的AI技术惠及所有人** — 确保这种高质量的AI基础设施不会仅仅局限于富人或技术精英群体。\n\n这就是为什么这是一个开源项目，而非私有项目。\n\n---\n\n## 新手入门？从这里开始\n\n你可能已经用过ChatGPT或Claude。输入一个问题，就能得到答案。很简单。\n\n你可以把AI系统看作是**三个层次**：\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_fbef7d388f7a.png\" alt=\"AI演进 - 从聊天机器人到你的个人AI系统\" width=\"800\">\n\u003C\u002Fp>\n\n### 聊天机器人\n\nChatGPT、Claude、Gemini——你提出一个问题，它会回答，然后就把之前的一切都忘记了。下一次对话又会从头开始。它对你、你的偏好，或者昨天聊过的内容一无所知。\n\n**模式：** 提问 → 回答 → 忘记\n\n### 主体型平台\n\n像Claude Code这样的工具。AI可以真正“做”事情——编写代码、浏览网页、编辑文件、执行命令。\n\n**模式：** 提问 → 使用工具 → 得到结果\n\n虽然功能更强大了，但它仍然不了解“你”——你的目标、偏好和过往经历。\n\n### PAI（个人AI基础设施）\n\n现在，你的DA会不断**学习和改进**：\n- **捕捉每一个信号** — 评分、情感反馈、验证结果\n- **从错误中学习** — 失败会被分析并修正\n- **随着时间推移变得更好** — 成功的模式会被强化\n- **自我升级** — 技能、工作流程，甚至核心行为都会进化\n\n此外，它还了解：\n- **你的目标** — 你在努力实现什么\n- **你的偏好** — 你喜欢怎样的方式\n- **你的历史** — 过去的决策和经验\n\n**模式：** 观察 → 思考 → 计划 → 执行 → 验证 → **学习** → 改进\n\n关键区别在于：**PAI会根据反馈不断学习**。每一次互动都会让它更好地为你提供帮助。\n\n---\n\n## 什么是 PAI？\n\nPAI 是一个个性化 AI 平台，旨在放大您的能力。\n\n它首先也是最重要的是为个人设计的，但同样适用于希望成为更好版本的自己的人团队、公司，甚至是星际联邦。\n\n无论规模大小，PAI 都是一个系统：它利用功能齐全的代理式 AI 平台，帮助用户理解、明确并实现自己的目标。\n\n### PAI 适合哪些人？\n\n**所有人，仅此而已。** 它是一个反对准入门槛的 AI 项目。\n\n- **非技术背景的小企业主**，希望借助 AI 处理开票、日程安排、客户跟进和市场营销等事务。\n- **企业**，希望更好地理解自身数据、优化运营并做出更明智的决策。\n- **管理者**，希望更高效地管理团队——跟踪项目进展、准备绩效评估并清晰沟通。\n- **艺术家和创意工作者**，希望找到本地活动、画廊以及展示作品的机会。\n- **普通大众**，希望改善生活——制定更好的健身计划、建立更牢固的人际关系、管理个人财务，或只是让自己更有条理。\n- **开发者**，使用 AI 编码助手时，需要持久记忆和自定义工作流。\n- **高级用户**，希望自己的 AI 能够了解他们的目标、偏好和上下文。\n- **团队**，构建具备一致能力的共享 AI 基础设施。\n- **对 AI 系统设计和个人化 AI 模式感兴趣的实验者**。\n\n### PAI 有何不同？\n\n人们最常问的问题是：\n\n> 这与 Claude Code 或其他代理式系统有什么不同？\n\n大多数代理式系统都是围绕工具构建的，而用户往往被放在次要位置。此外，这些系统大多以任务为导向，而非基于目标，并且很少充分利用所有可用的上下文信息。而 PAI 则恰恰相反。\n\n**三大核心差异：**\n\n1. **目标导向** — PAI 的首要关注点是运行它的用户及其在现实世界中想要达成的目标，而不是技术本身。这一点贯穿于系统的每一个执行环节。\n   \n2. **追求最优输出** — 系统的外层循环以及所有操作，都致力于根据当前情境及周边环境，生成完全符合需求的输出结果。\n\n3. **持续学习** — 系统会不断记录每次请求的执行情况、所做的调整、产生的结果，以及您对这些结果的反馈（满意或不满意），从而持续优化自身表现。\n\n---\n\n## PAI 原则\n\n这些原则指导着 PAI 系统的设计与构建。**[完整解读 →](https:\u002F\u002Fdanielmiessler.com\u002Fblog\u002Fpersonal-ai-infrastructure)**\n\n| 序号 | 原则 | 概述 |\n|---|-----------|---------|\n| 1 | **以用户为中心** | PAI 是围绕您构建的，而非围绕工具。您的目标、偏好和上下文始终放在首位——基础设施的存在是为了服务于它们。 |\n| 2 | **基础算法** | 科学方法作为通用的问题解决循环：观察 → 思考 → 计划 → 构建 → 执行 → 验证 → 学习。明确理想状态，反复迭代直至达成。 |\n| 3 | **清晰思考优先** | 优质的提示词源于清晰的思考。在编写提示词之前，先明确问题。 |\n| 4 | **架构优先于模型** | 系统架构的重要性远胜于所使用的具体模型。 |\n| 5 | **确定性基础设施** | AI 具有概率性；而您的基础设施不应如此。应使用模板和模式。 |\n| 6 | **代码优先于提示词** | 如果可以用 Bash 脚本解决的问题，就不要使用 AI。 |\n| 7 | **先规范、再测试、后评估** | 在构建之前先编写规范和测试用例。通过测量来判断系统是否有效。 |\n| 8 | **UNIX 哲学** | 一件事做到最好。让工具可组合。使用文本界面。 |\n| 9 | **软件工程与 SRE 原则** | 将 AI 基础设施视为生产级软件：版本控制、自动化、监控。 |\n| 10 | **命令行界面优先** | 命令行界面比图形用户界面更快、更易脚本化且更可靠。 |\n| 11 | **目标 → 代码 → CLI → 提示词 → 代理** | 决策层次：先明确目标，再写代码，然后使用命令行，接着是提示词，最后才是代理。 |\n| 12 | **技能管理** | 模块化的功能可根据上下文智能路由。 |\n| 13 | **记忆系统** | 所有值得记住的信息都会被记录下来。历史会为未来的上下文提供支持。 |\n| 14 | **代理人格** | 不同的工作需求需要不同的处理方式。专门化的代理拥有独特的“声音”。 |\n| 15 | **科学作为元循环** | 假设 → 实验 → 测量 → 迭代。 |\n| 16 | **允许失败的权利** | 明确允许说“我不知道”，可以有效防止幻觉现象的发生。 |\n\n---\n\n## PAI 基本组件\n\n如果说原则描述的是 PAI 的*哲学*，那么基本组件就是其*架构*——支撑一切运转的核心系统。\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_aa7d84f40c6d.png\" alt=\"PAI 基本组件 - 一个了解你的系统，而非工具集合\" width=\"800\">\n\u003C\u002Fp>\n\n这些基本组件协同工作，为您带来一种与真正理解并熟悉您的系统互动的体验，而不是仅仅执行指令的工具集合。\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_a7d904467cc0.png\" alt=\"助理与代理式交互\" width=\"700\">\n\u003C\u002Fp>\n\n### 助理与代理式 AI 交互\n\nPAI 将 AI 视为一位[持久的助理、朋友、教练和导师](https:\u002F\u002Fdanielmiessler.com\u002Fblog\u002Fpersonal-ai-maturity-model)，而非只负责执行任务的无状态代理。助理会了解您的目标、记住您的偏好，并随着时间推移不断改进；而代理则只是执行命令后便遗忘一切。\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_995c0de09826.png\" alt=\"TELOS - 深度目标理解\" width=\"700\">\n\u003C\u002Fp>\n\n### TELOS（深度目标理解）\n\n10 个文件记录了您的身份与目标：MISSION.md、GOALS.md、PROJECTS.md、BELIEFS.md、MODELS.md、STRATEGIES.md、NARRATIVES.md、LEARNED.md、CHALLENGES.md 和 IDEAS.md。您的 DA 清楚您正在努力的方向，因为这一切都被详细记录了下来。\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_5dec59e61dfe.png\" alt=\"用户\u002F系统分离\" width=\"700\">\n\u003C\u002Fp>\n\n### 用户\u002F系统分离\n\n您的个性化设置存储在 USER\u002F 目录下。而 PAI 的基础设施则位于 SYSTEM\u002F 目录中。当 PAI 升级时，您的文件不会受到影响。实现身份可移植、升级安全。\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_27a0dfbc5fcf.png\" alt=\"精细化定制\" width=\"700\">\n\u003C\u002Fp>\n\n### 精细化定制\n\n六层定制选项：身份（姓名、语气、个性）、偏好（技术栈、工具）、工作流程（技能如何执行）、技能（具备哪些能力）、钩子（事件如何处理）以及记忆（记录哪些内容）。您可以从默认设置开始，按需进行个性化调整。\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_3ca3a7edb0ad.png\" alt=\"技能系统\" width=\"700\">\n\u003C\u002Fp>\n\n### 技能系统\n\n高度注重持续稳定的结果。其结构以*确定性结果优先*，按照 CODE -> 命令行工具 -> 提示词 -> 技能 的顺序进行，而非随意的架构。\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_2368afac3b09.png\" alt=\"记忆系统\" width=\"700\">\n\u003C\u002Fp>\n\n### 记忆系统\n\n专注于持续学习。每次交互都会产生信号——评分、情感、成功或失败——这些信号会反馈回系统，用于不断优化。采用三层架构（热\u002F温\u002F冷），并配备基于阶段的学习目录。\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_b0b1f63c109b.png\" alt=\"钩子系统\" width=\"700\">\n\u003C\u002Fp>\n\n### 钩子系统\n\n响应生命周期事件——会话开始、工具使用、任务完成等。8种事件类型支持语音通知、自动加载上下文、会话捕获、安全验证和可观性等功能。\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_d04b87fad00e.png\" alt=\"安全系统\" width=\"700\">\n\u003C\u002Fp>\n\n### 安全系统\n\n默认定义了系统级和用户级的安全策略。您无需使用 `--dangerously-skip-permissions` 参数即可获得顺畅的体验。PAI 的安全钩子会在命令执行前进行验证，阻止危险操作，同时确保正常工作流程顺利进行。\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_b1c5dd64011a.png\" alt=\"基于AI的安装\" width=\"700\">\n\u003C\u002Fp>\n\n### 基于AI的安装\n\nGUI 安装程序负责处理所有内容——包括先决条件、配置和设置。无需手动配置，也无需猜测。\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_09439a021951.png\" alt=\"通知系统\" width=\"700\">\n\u003C\u002Fp>\n\n### 通知系统\n\n在不打扰您的情况下保持信息畅通。通过 ntfy 发送移动设备提醒，集成 Discord 进行团队更新，并根据任务持续时间智能路由，对于长时间运行的任务会自动升级通知级别。一次性触发的设计意味着通知绝不会阻塞您的工作流。\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_3b61fdfc0b2f.png\" alt=\"语音系统\" width=\"700\">\n\u003C\u002Fp>\n\n### 语音系统\n\n由 ElevenLabs TTS 提供支持。您可以听到任务完成、会话摘要以及重要更新的语音播报。韵律增强功能使语音听起来更加自然。您的 AI 现在有了声音。\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_2d9a23177c4b.png\" alt=\"终端界面\" width=\"700\">\n\u003C\u002Fp>\n\n### 终端界面\n\n丰富的标签标题和窗格管理功能。动态状态栏显示学习信号、上下文使用情况以及当前任务状态。您的终端现在就是一个指挥中心。\n\n---\n\n## 🚀 安装\n\n> [!CAUTION]\n> **项目处于积极开发中** — PAI 正在快速演进。请预期会出现破坏性变更、结构调整以及频繁的更新。我们正在维护稳定分支和开发分支，但目前两者尚未分离。\n\n### 全新安装\n\n```bash\n# 克隆仓库\ngit clone https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure.git\ncd Personal_AI_Infrastructure\u002FReleases\u002Fv4.0.3\n\n# 复制发布版本并运行安装程序\ncp -r .claude ~\u002F && cd ~\u002F.claude && bash install.sh\n```\n\n**安装程序将：**\n- 检测您的系统并安装先决条件（Bun、Git、Claude Code）\n- 询问您的姓名、AI 助手名称、时区及温度单位偏好\n- 将 PAI 仓库克隆并配置到 `~\u002F.claude\u002F`\n- 设置 ElevenLabs 语音功能（可选）\n- 配置您的 shell 别名并验证安装\n\n**安装完成后：** 运行 `source ~\u002F.zshrc && pai` 启动 PAI。\n\n### 从旧版本升级\n\n```bash\n# 1. 备份当前安装\ncp -r ~\u002F.claude ~\u002F.claude-backup-$(date +%Y%m%d)\n\n# 2. 克隆并覆盖现有安装\ngit clone https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure.git\ncd Personal_AI_Infrastructure\u002FReleases\u002Fv4.0.3\ncp -r .claude ~\u002F\n\n# 3. 运行安装程序（检测现有安装，保留您的数据）\ncd ~\u002F.claude && bash install.sh\n\n# 4. 重新构建 CLAUDE.md\nbun ~\u002F.claude\u002FPAI\u002FTools\u002FBuildCLAUDE.ts\n```\n\n> [!TIP]\n> 安装程序会**自动检测**现有安装。它会保留您的 `USER\u002F` 文件，合并 `settings.json`（仅更新安装程序管理的字段，如身份和版本），并且绝不会覆盖您的钩子、状态栏或自定义配置。\n\n**升级后检查清单：**\n- [ ] 在 `settings.json` 中确认您的身份信息（姓名、AI 名称、时区）\n- [ ] 确认状态栏显示正确\n- [ ] 测试语音通知（如果已启用）\n- [ ] 运行一个简单提示，确认 PAI 响应正常\n\n---\n\n## 📦 PAI 套件\n\n不想安装整个 PAI？**套件**是可独立安装的 AI 能力模块，您可以一次添加一个。每个套件都是自包含的——您的 AI 会读取安装指南并为您完成所有设置。无需安装完整的 PAI 系统。\n\n只需让您的 AI 指向任意套件并说“安装这个”：\n\n```\n“从 PAI\u002FPacks\u002FResearch\u002F 安装 Research 套件”\n```\n\n您的 AI 会引导您完成一个五阶段向导：系统分析、用户提问、备份、安装和验证。\n\n### 可用套件\n\n| 套件 | 功能描述 |\n|------|-------------|\n| [ContextSearch](Packs\u002FContextSearch\u002F) | `\u002Fcontext-search` 和 `\u002Fcs` — 快速调用之前的工作会话 |\n| [Agents](Packs\u002FAgents\u002F) | 根据特质、声音和个性定制代理 |\n| [ContentAnalysis](Packs\u002FContentAnalysis\u002F) | 从视频、播客、文章和 YouTube 中提取智慧 |\n| [Investigation](Packs\u002FInvestigation\u002F) | OSINT 和调查——公司情报、人员搜索、域名查询 |\n| [Media](Packs\u002FMedia\u002F) | AI 图像生成、图表、信息图以及 Remotion 视频 |\n| [Research](Packs\u002FResearch\u002F) | 多代理研究——快速、标准、广泛和深度模式 |\n| [Scraping](Packs\u002FScraping\u002F) | 使用 Bright Data 代理和 Apify 社交媒体机器人进行网页抓取 |\n| [Security](Packs\u002FSecurity\u002F) | 情报收集、Web 应用测试、提示注入测试以及安全新闻 |\n| [Telos](Packs\u002FTelos\u002F) | 生活操作系统——目标、信念、智慧、项目仪表盘和麦肯锡报告 |\n| [Thinking](Packs\u002FThinking\u002F) | 第一原理、议会辩论、红队演练、头脑风暴和科学研究 |\n| [USMetrics](Packs\u002FUSMetrics\u002F) | 来自 FRED、EIA、美国财政部、BLS 和人口普查局的 68 个美国经济指标 |\n| [Utilities](Packs\u002FUtilities\u002F) | CLI 生成、技能搭建、Fabric 模式、Cloudflare 以及浏览器自动化 |\n\n每个套件都可以独立使用——您可以只安装一个，也可以全部安装。它们旨在为您提供 PAI 级别的能力，无论您是否运行完整的 PAI 系统。\n\n**[浏览所有套件 →](Packs\u002F)**\n\n---\n\n## ❓ 常见问题\n\n### PAI 与单纯使用 Claude Code 有何不同？\n\nPAI 原生构建于 Claude Code 之上，并且设计初衷就是保持这种关系。我们选择 Claude Code，是因为它的钩子系统、上下文管理和智能体架构，为个人 AI 基础设施提供了最佳的底层支持。\n\nPAI 并不是 Claude Code 的替代品——它是位于 Claude Code 之上的那一层，让 Claude Code 真正“属于你”：\n\n- **持久化记忆**——你的 DA 能够记住过去的会话、决策和学习成果。\n- **自定义技能**——针对你最常做的事情，提供专业化的功能。\n- **你的上下文**——目标、联系人、偏好等信息，无需重复说明即可随时调用。\n- **智能路由**——只需说一句“研究一下这个”，相应的流程就会自动触发。\n- **自我改进**——系统会根据自身学到的内容不断优化。\n\n可以这样理解：Claude Code 是引擎，而 PAI 则是让这台“车”真正成为“你的车”的其他所有部件。\n\n### PAI 与 Claude Code 自带功能有何区别？\n\nClaude Code 提供了强大的基础组件——钩子、斜杠命令、MCP 服务器、上下文文件等。这些都只是独立的构建模块。\n\n而 PAI 是基于这些基础组件构建的完整系统。它将所有部分有机地连接起来：你的目标指导技能的选择，技能生成记忆，记忆又反过来提升未来的响应质量。PAI 把 Claude Code 的各个构建模块整合成一个连贯的个人 AI 平台。\n\n### PAI 是否仅适用于 Claude Code？\n\nPAI 是 Claude Code 的原生应用。我们认为，Claude Code 的钩子系统、上下文管理和智能体能力使其成为构建个人 AI 基础设施的最佳平台，而 PAI 正是充分利用这些特性的产物。\n\n不过，PAI 的核心概念（技能、记忆、算法）具有普适性，其代码采用 TypeScript 和 Bash 编写，因此欢迎社区成员将其移植到其他平台上。\n\n### 这与 Fabric 有何不同？\n\n[Fabric](https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002Ffabric) 是一系列用于特定任务的 AI 提示词（模式），专注于“应该向 AI 提出什么问题”。\n\n而 PAI 则是关于“你的 DA 如何运作”的基础设施——包括记忆、技能、路由、上下文和自我改进机制。两者相辅相成。许多 PAI 用户会将 Fabric 的模式集成到自己的技能中。\n\n### 如果我弄坏了怎么办？\n\n恢复过程非常简单：\n\n- **先备份**——在任何升级之前，请执行：`cp -r ~\u002F.claude ~\u002F.claude-backup-$(date +%Y%m%d)`\n- **USER\u002F 目录安全**——你在 `USER\u002F` 中的自定义配置不会被安装程序或升级所触及。\n- **设置合并而非覆盖**——安装程序只会更新身份和版本字段，你的钩子、状态栏以及自定义配置都会保留。\n- **Git 版本控制**——对所有内容进行版本管理，必要时可回滚。\n- **历史记录保留**——即使出现错误，你的 DA 的记忆也不会丢失。\n- **DA 可以修复**——既然你的 DA 参与了构建过程，它同样可以帮助修复问题。\n- **重新安装**——再次运行安装程序；它会检测现有安装并智能合并。\n\n---\n\n## 🎯 路线图\n\n| 功能 | 描述 |\n|------|------|\n| **本地模型支持** | 使用本地模型（Ollama、llama.cpp）运行 PAI，以保障隐私并控制成本 |\n| **精细化模型路由** | 根据任务复杂度，将不同任务分配给不同的模型处理 |\n| **远程访问** | 从任何地方访问你的 PAI——手机、网页或其他设备 |\n| **外呼电话功能** | 支持语音外呼的能力 |\n| **外部通知** | 强大的通知系统，支持邮件、Discord、Telegram 和 Slack |\n\n---\n\n## 🌐 社区\n\n**GitHub 讨论区：** [加入讨论](https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure\u002Fdiscussions)\n\n**社区 Discord：** PAI 与其他 AI 项目一同在 [社区 Discord](https:\u002F\u002Fdanielmiessler.com\u002Fupgrade) 中讨论。\n\n**Twitter\u002FX：** [@danielmiessler](https:\u002F\u002Ftwitter.com\u002Fdanielmiessler)\n\n**博客：** [danielmiessler.com](https:\u002F\u002Fdanielmiessler.com)\n\n### 星标历史\n\n\u003Ca href=\"https:\u002F\u002Fstar-history.com\u002F#danielmiessler\u002FPersonal_AI_Infrastructure&Date\">\n \u003Cpicture>\n   \u003Csource media=\"(prefers-color-scheme: dark)\" srcset=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_c94f139001b0.png&theme=dark\" \u002F>\n   \u003Csource media=\"(prefers-color-scheme: light)\" srcset=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_c94f139001b0.png\" \u002F>\n   \u003Cimg alt=\"星标历史图表\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_readme_c94f139001b0.png\" \u002F>\n \u003C\u002Fpicture>\n\u003C\u002Fa>\n\n---\n\n## 🤝 贡献\n\n我们欢迎任何形式的贡献！请查看我们的 [GitHub Issues](https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure\u002Fissues) 获取当前开放的任务。\n\n1. **Fork 仓库**\n2. **进行修改**——修复 bug、新增技能、改进文档\n3. **充分测试**——在全新系统上安装并验证\n4. **提交 PR**，附上示例和测试证据\n\n---\n\n## 📜 许可证\n\nMIT 许可证——详情请参阅 [LICENSE](LICENSE) 文件。\n\n---\n\n## 🙏 致谢\n\n**Anthropic 和 Claude Code 团队**——首先也是最重要的感谢对象。你们正在以无人能及的速度推动 AI 的发展。Claude Code 是这一切得以实现的基础。\n\n**[IndyDevDan](https:\u002F\u002Fwww.youtube.com\u002F@indydevdan)**——他关于元提示和自定义智能体的精彩视频，启发了 PAI 的部分设计思路。\n\n### 贡献者\n\n**[fayerman-source](https:\u002F\u002Fgithub.com\u002Ffayerman-source)**——为语音系统集成了 Google Cloud TTS 提供商，并添加了 Linux 音频支持。\n\n**Matt Espinoza**——对 PAI 2.3 版本进行了大量测试、提出了许多创意和反馈，并参与了路线图的制定。\n\n---\n\n## 💜 支持本项目\n\n\u003Cdiv align=\"center\">\n\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fsponsors\u002Fdanielmiessler\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FSponsor-❤️-EA4AAA?style=for-the-badge&logo=github-sponsors&logoColor=white\" alt=\"赞助\">\u003C\u002Fa>\n\n**PAI 永远免费且开源。如果你觉得它有价值，可以通过 [赞助该项目](https:\u002F\u002Fgithub.com\u002Fsponsors\u002Fdanielmiessler) 来支持我们。**\n\n\u003C\u002Fdiv>\n\n---\n\n## 📚 相关阅读\n\n- [真正的物联网](https:\u002F\u002Fdanielmiessler.com\u002Fblog\u002Freal-internet-of-things) — PAI 背后的愿景\n- [AI 的可预测路径：7 大组件](https:\u002F\u002Fdanielmiessler.com\u002Fblog\u002Fai-predictable-path-7-components-2024) — AI 发展方向的可视化解读\n- [构建个人 AI 基础设施](https:\u002F\u002Fdanielmiessler.com\u002Fblog\u002Fpersonal-ai-infrastructure) — 包含示例的完整 PAI 操作指南\n\n---\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>📜 更新历史\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n\u003Cbr\u002F>\n\n**v4.0.3 (2026-03-01) — 社区 PR 修复**\n- 修复了 Inference.ts 中的 JSON 数组解析问题\n- 从 CONTEXT_ROUTING.md 中移除了 29 处无效引用\n- 提升了 WorldThreatModelHarness 的 PAI_DIR 可移植性\n- 为 v2.5 和 v3.0 用户提供了用户上下文迁移功能\n- [发布说明](Releases\u002Fv4.0.3\u002FREADME.md)\n\n**v4.0.2 (2026-03-01) — 错误修复补丁**\n- 进行了 13 处针对性修复：包括 Linux 兼容性、安装程序、状态栏及钩子等方面的改进\n- 实现了跨平台 OAuth 令牌提取，并修复了 GNU coreutils 中的 tr 命令问题\n- 优化了推理保护机制（节省约 15 秒），增强了 lineage 追踪功能，同时移除了冗余代码\n- [发布说明](Releases\u002Fv4.0.2\u002FREADME.md)\n\n**v4.0.1 (2026-02-28) — 升级路径与偏好设置**\n- 更新了升级文档，增加了备份、合并及升级后检查清单\n- 在状态栏和安装程序中新增了可配置的温度单位切换功能（华氏\u002F摄氏）\n- 修复了常见问题解答中的过时 Python 引用，并优化了故障恢复指导\n- [发布说明](Releases\u002Fv4.0.1\u002FREADME.md)\n\n**v4.0.0 (2026-02-27) — 精简高效**\n- 将 38 个扁平化的技能目录整合为 12 个层级化类别（顶级目录减少 68%）\n- 移除了已废弃的系统模块：Components\u002F、DocRebuild 和 RebuildSkill\n- 引入了基于 CLAUDE.md 的模板系统，配合 BuildCLAUDE.ts 和 SessionStart 钩子\n- 算法版本升级至 v3.5.0（原为 v1.4.0）\n- 进行了全面的安全清理工作（共清理 33 余份文件）\n- 更新了所有版本引用，并修复了 Electron 应用崩溃问题\n- 当前包含 63 个技能、21 个钩子、180 条工作流和 14 个智能体\n- [发布说明](Releases\u002Fv4.0.0\u002FREADME.md)\n\n**v3.0.0 (2026-02-15) — 算法逐步成熟**\n- 算法版本升级至 v1.4.0，新增约束提取与构建漂移预防功能\n- 支持持久化 PRD 和并行循环执行\n- 完整的图形化安装向导\n- 新增 10 个技能、代理团队\u002F蜂群模式以及语音个性系统\n- 总计 38 个技能、20 个钩子和 162 条工作流\n- [发布说明](Releases\u002Fv3.0\u002FREADME.md)\n\n**v2.5.0 (2026-01-30) — 思考更深入，执行更快**\n- 两步能力选择流程：在 THINK 阶段根据 ISC 验证钩子提示\n- 引入思考工具 Justify-Exclusion：对于 Council、RedTeam、FirstPrinciples 等采用 opt-OUT 而非 opt-IN 方式\n- 默认启用并行执行：通过并行启动多个智能体来同时处理独立任务\n- 包含 28 个技能、17 个钩子和 356 条工作流\n- [发布说明](Releases\u002Fv2.5\u002FREADME.md)\n\n**v2.4.0 (2026-01-23) — 算法系统**\n- 构建了通用问题解决系统，支持 ISC（理想状态标准）追踪\n- 包含 29 个技能、15 个钩子和 331 条工作流\n- 以“欣快感”作为结果评估指标\n- 加强了安全防护，引入 AllowList 强制执行机制\n- [发布说明](Releases\u002Fv2.4\u002FREADME.md)\n\n**v2.3.0 (2026-01-15) — 完整发布回归**\n- 恢复了完整的 `.claude\u002F` 目录发布方式，并支持持续学习\n- 明确记录显性和隐性评分\n- 优化了钩子系统，新增 14 个生产级钩子\n- 在状态栏中增加了学习信号显示\n- [发布说明](Releases\u002Fv2.3\u002FREADME.md)\n\n**v2.1.1 (2026-01-09) — MEMORY 系统迁移**\n- 将历史系统合并至核心模块，更名为 MEMORY 系统\n\n**v2.1.0 (2025-12-31) — 模块化架构**\n- 源代码从嵌入式 Markdown 中分离出来，存放在实际文件中\n\n**v2.0.0 (2025-12-28) — PAI v2 正式发布**\n- 采用模块化架构，各技能相互独立\n- Claude Code 原生设计\n\u003C\u002Fdetails>\n\n---\n\n\u003Cdiv align=\"center\">\n\n**由 [Daniel Miessler](https:\u002F\u002Fdanielmiessler.com) 与 PAI 社区倾情打造**\n\n*提升自我。*\n\n\u003C\u002Fdiv>","# Personal AI Infrastructure (PAI) 快速上手指南\n\nPersonal AI Infrastructure (PAI) 是一个开源的个人 AI 平台，旨在通过以目标为导向、具备持续学习能力的代理系统，放大每个人的创造力。与传统的聊天机器人不同，PAI 会记忆你的偏好、目标和历史，并通过反馈循环不断进化。\n\n## 环境准备\n\n在开始之前，请确保你的系统满足以下要求：\n\n*   **操作系统**：\n    *   推荐：**Linux** (Ubuntu\u002FDebian, Fedora, Arch 等) 或 **macOS**。\n    *   Windows 用户建议使用 **WSL2** (Windows Subsystem for Linux) 以获得最佳兼容性。\n*   **核心运行时**：\n    *   **Bun**: PAI 基于 Bun 构建（比 Node.js 更快）。\n    *   **Git**: 用于克隆代码仓库。\n*   **AI 模型访问**：\n    *   你需要一个可用的 LLM API Key（如 Anthropic Claude, OpenAI 等），并在配置阶段填入。\n\n### 前置依赖安装\n\n**macOS \u002F Linux:**\n\n```bash\n# 安装 Git (如果未安装)\n# macOS\nbrew install git\n# Ubuntu\u002FDebian\nsudo apt update && sudo apt install -y git\n\n# 安装 Bun (官方一键脚本)\ncurl -fsSL https:\u002F\u002Fbun.sh\u002Finstall | bash\n\n# 加载环境变量 (根据提示选择对应的 shell，通常是 bash 或 zsh)\nsource ~\u002F.bashrc \n# 或\nsource ~\u002F.zshrc\n```\n\n**Windows (WSL2):**\n\n请在 WSL2 终端中执行上述 Linux 命令。\n\n> **国内加速提示**：如果 `bun.sh` 下载缓慢，可尝试使用国内镜像源安装 Bun：\n> ```bash\n> export BUN_INSTALL=\"$HOME\u002F.bun\"\n> curl -fsSL https:\u002F\u002Fmirrors.cloud.tencent.com\u002Fbun\u002Finstall.sh | bash\n> # 或者手动下载二进制文件并配置 PATH\n> ```\n\n## 安装步骤\n\n### 1. 克隆项目仓库\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure.git\ncd Personal_AI_Infrastructure\n```\n\n### 2. 安装依赖\n\n使用 Bun 安装项目所需的 TypeScript 依赖：\n\n```bash\nbun install\n```\n\n### 3. 初始化配置\n\n运行安装\u002F设置脚本。该脚本将引导你创建配置文件并设置 API Key。\n\n```bash\nbun run setup\n```\n\n*   按照提示输入你的 **LLM Provider** (例如 `anthropic`)。\n*   输入你的 **API Key**。\n*   设置你的个人基本信息（用于 PAI 建立初始用户画像）。\n\n> **注意**：配置文件通常保存在 `.env` 或 `config\u002F` 目录下，请勿将包含 API Key 的文件上传至公共仓库。\n\n### 4. 验证安装\n\n运行版本检查或帮助命令以确保一切正常：\n\n```bash\nbun run pai --version\n# 或\nbun run pai --help\n```\n\n## 基本使用\n\nPAI 的核心在于通过命令行界面 (CLI) 与具有“记忆”和“目标导向”的 AI 代理交互。\n\n### 启动交互式会话\n\n最简单的使用方式是启动主代理进行对话。PAI 会自动加载你的上下文、目标和历史记忆。\n\n```bash\nbun run pai chat\n```\n\n**示例场景**：\n你可以直接询问涉及你个人目标的问题，例如：\n> \"根据我上周设定的健身目标，帮我规划今天的训练计划，并回顾我昨天的饮食记录。\"\n\nPAI 不仅会生成计划，还会引用你之前的输入（如果已记录），并根据反馈机制调整建议。\n\n### 执行特定任务 (Agentic Mode)\n\nPAI 可以调用工具执行实际操作（如文件操作、网络搜索等，取决于具体安装的 Packs）。\n\n```bash\nbun run pai execute \"分析当前目录下的日志文件，总结错误模式并生成修复建议文档\"\n```\n\n### 查看与管理系统状态\n\n查看当前的技能包 (Packs) 或系统状态：\n\n```bash\n# 列出已激活的功能包\nbun run pai packs list\n\n# 查看系统记忆摘要\nbun run pai memory summary\n```\n\n---\n\n**下一步建议**：\n*   阅读 `Packs\u002F` 目录了解可扩展的功能模块。\n*   查阅 `Releases\u002F` 获取最新版本的更新日志。\n*   加入社区讨论以分享你的自定义工作流。","资深安全研究员李明需要每天从数十个分散的安全博客、GitHub 仓库和威胁情报源中筛选关键漏洞信息，并整理成可执行的防御策略报告。\n\n### 没有 Personal_AI_Infrastructure 时\n- **信息过载与遗漏**：手动浏览多个来源效率极低，容易错过隐蔽但高危的最新漏洞情报。\n- **上下文割裂**：不同来源的数据格式杂乱，难以将碎片化信息与李明现有的知识库自动关联。\n- **响应滞后**：从发现威胁到形成结构化报告耗时数小时，无法在攻击窗口期内及时部署防御。\n- **重复劳动**：每次都需要重新编写提示词来格式化数据，无法沉淀个性化的分析逻辑。\n- **能力受限**：单靠人力难以同时监控全球多语言源，导致防御视野存在盲区。\n\n### 使用 Personal_AI_Infrastructure 后\n- **智能聚合过滤**：PAI 自动抓取并预筛选全网情报，仅推送与李明关注领域相关的高价值告警。\n- **知识自动增强**：PAI 将新情报与李明的本地笔记和过往案例自动关联，生成带有上下文的深度分析。\n- **实时策略生成**：分钟级输出包含具体修复命令和检测规则的防御报告，大幅缩短响应时间。\n- **个性化代理进化**：PAI 记住李明的分析偏好和报告模板，越用越懂他的思维模式，无需重复指令。\n- **能力无限延伸**：7x24 小时不间断监控多语言源，相当于为李明配备了一支全天候的全球情报团队。\n\nPersonal_AI_Infrastructure 将李明从繁琐的信息搬运工转变为战略决策者，真正实现了用 AI 架构放大人类专业能力的愿景。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdanielmiessler_Personal_AI_Infrastructure_fbef7d38.png","danielmiessler","Daniel Miessler 🛡️","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fdanielmiessler_33f5146d.png","⚒️ Building AI that ᴜᴘɢʀᴀᴅᴇs humans.","Unsupervised Learning","San Francisco Bay Area","daniel@danielmiessler.com","https:\u002F\u002Fdanielmiessler.com\u002F","https:\u002F\u002Fgithub.com\u002Fdanielmiessler",[83,87,91,95,99,103,107,111,115],{"name":84,"color":85,"percentage":86},"TypeScript","#3178c6",76.3,{"name":88,"color":89,"percentage":90},"Python","#3572A5",9.8,{"name":92,"color":93,"percentage":94},"Shell","#89e051",4.6,{"name":96,"color":97,"percentage":98},"Vue","#41b883",3.1,{"name":100,"color":101,"percentage":102},"JavaScript","#f1e05a",2.6,{"name":104,"color":105,"percentage":106},"CSS","#663399",1.8,{"name":108,"color":109,"percentage":110},"Handlebars","#f7931e",1.4,{"name":112,"color":113,"percentage":114},"HTML","#e34c26",0.2,{"name":116,"color":117,"percentage":114},"Swift","#F05138",11122,1546,"2026-04-06T22:39:09","MIT","Linux, macOS","未说明",{"notes":125,"python":126,"dependencies":127},"该项目是一个基于 Bun 和 TypeScript 的个人 AI 基础设施框架，而非单纯的 Python 深度学习模型。README 明确提到了 v4.0.3 版本增加了 Linux 兼容性、安装程序改进和可移植性。由于核心运行时为 Bun (JavaScript\u002FTypeScript 运行时)，因此没有传统的 Python 版本或 PyTorch\u002FCUDA 依赖要求。具体的 GPU 和内存需求取决于用户在此架构中集成的后端 AI 模型（如本地运行的 LLM 或调用的 API），框架本身对此未作硬性规定。","未说明 (项目主要基于 Bun\u002FTypeScript)",[128,84],"Bun",[13,15,14],[131,132,133,134],"ai","augmentation","productivity","humans",null,"2026-03-27T02:49:30.150509","2026-04-07T11:35:31.152362",[139,144,149,153,157,162,167],{"id":140,"question_zh":141,"answer_zh":142,"source_url":143},21819,"PAI_DIR 环境变量应该指向哪里？它是指克隆的仓库目录还是安装目录？","PAI_DIR 是您安装位置的单一事实来源。默认情况下，它指向 `~\u002F.claude`（Claude Code 的标准位置），但也可以自定义为其他任意目录（如 `~\u002F.config\u002Fpai`）。克隆的仓库目录与实际安装目录是分开的。解析顺序为：优先使用 `PAI_DIR` 环境变量，其次是 `PAI_HOME`，最后默认为 `~\u002F.claude`。如果您担心覆盖现有的 Claude 配置，可以将其设置为独立目录，并通过符号链接或配置钩子来集成。","https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure\u002Fissues\u002F265",{"id":145,"question_zh":146,"answer_zh":147,"source_url":148},21820,"如何在非 macOS 系统（如 Linux 或 WSL）上安装和使用 Kai 语音系统？","Kai 语音系统原生仅支持 macOS。在 Linux 或 WSL 上使用时需要进行以下修改：\n1. 将音频播放命令 `afplay` 替换为 `mpg123`、`aplay` 或 `paplay`。\n2. 将桌面通知命令 `osascript` 替换为 `notify-send`。\n3. 确保安装了必要的依赖包（如 `mpg123`）。\n安装脚本现在会在检测到非 macOS 系统时发出警告并询问确认，但在 Windows 上会直接停止安装。","https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure\u002Fissues\u002F274",{"id":150,"question_zh":151,"answer_zh":152,"source_url":148},21821,"安装 Kai 语音系统时遇到端口冲突或需要自定义端口怎么办？","默认端口为 8888。如果该端口被占用，安装程序会提示并显示占用该端口的进程。您可以通过设置环境变量 `VOICE_SERVER_PORT` 或 `PAI_VOICE_PORT` 来指定替代端口。注意：部分旧版脚本可能硬编码了端口，需手动修改 `manage.sh` 以读取环境变量，或在 `.env` 文件中配置 `VOICE_SERVER_PORT`。",{"id":154,"question_zh":155,"answer_zh":156,"source_url":148},21822,"在 .env 文件中配置 ElevenLabs Voice ID 时出现 'invalid_uid' 错误如何解决？","这是因为安装脚本错误地在值周围添加了双引号。错误的格式是：`ELEVENLABS_VOICE_ID=\"cgSgspJ2msm6clMCkdW9\"`。正确的 .env 格式不应包含引号，应修改为：`ELEVENLABS_VOICE_ID=cgSgspJ2msm6clMCkdW9`。请检查您的 `.env` 文件，移除 ID 值周围的所有引号。",{"id":158,"question_zh":159,"answer_zh":160,"source_url":161},21823,"为什么我在 Linux 上安装后遇到 'SessionStart hook error' 或找不到脚本的错误？","这通常是因为 PAI v1.x 架构存在平台兼容性问题（如大小写敏感的目录名 `Skills` vs `skills`，或路径变量 `__HOME__` 未正确解析）。该项目已迁移至 **PAI v2.0**，采用了模块化的 'Packs' 系统来解决这些依赖问题。建议不要继续使用 v1.x 的安装方式，而是阅读最新的 README 文档，按照 v2.0 的指南重新安装独立的模块化组件。","https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure\u002Fissues\u002F224",{"id":163,"question_zh":164,"answer_zh":165,"source_url":166},21824,"Claude Code 更新后状态行（Status Line）被截断只显示一行怎么办？","这是 Claude Code 2.1.87+ 版本的限制，不再支持多行状态行。解决方案是创建一个“密集”的单行状态行脚本，将上下文、使用情况、当前路径（PWD）和 Git 分支等信息压缩在一行内显示。您可以参考社区提供的 `statusline-dense.sh` 脚本来实现这一功能，或者等待未来版本恢复多行支持。","https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure\u002Fissues\u002F1013",{"id":168,"question_zh":169,"answer_zh":170,"source_url":148},21825,"我想在服务器上运行 PAI 语音服务，但让它在桌面端通知，如何配置？","您可以通过环境变量指定语音服务器的完整 URL。虽然具体变量名可能随版本更新（如 `VOICE_SERVER_URL` 或在客户端配置中指定），但核心思路是在桌面端的配置中指向服务器运行的地址，而不是默认的 localhost。这使得分离式部署（服务端运行 PAI，桌面端接收通知）成为可能。",[172,177,182,187,192,197,202,207,212,217,222,227,232,237,242,247,252,257,262],{"id":173,"version":174,"summary_zh":175,"released_at":176},127825,"v4.0.3","4 个社区贡献的修复 — 无新功能，无破坏性变更。\n\n## 变更内容\n\n### 推理与解析\n- **#800** `Inference.ts` 中的 JSON 解析仅匹配对象 `{}` — 现在也支持数组 `[]`，并通过 `JSON.parse` 进行验证。\n\n### 文档与可移植性\n- **#836** 移除了 `CONTEXT_ROUTING.md` 中 29 处指向 v4.0 版本中已移除文件的无效链接 — 统一为 4 个 README 指向。\n- **#817** `WorldThreatModelHarness` 中硬编码了 `~\u002F.claude\u002F` 路径 — 现改为使用 `$PAI_DIR` 以提高可移植性。\n\n### 安装程序\n- **#846** 从 v2.5 或 v3.0 升级时，用户上下文会滞留在 `skills\u002FPAI\u002FUSER\u002F` 目录下 — 现在安装程序会将文件迁移到 `PAI\u002FUSER\u002F` 并创建符号链接以保持向后兼容。\n\n## 安装步骤\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure.git\ncd Personal_AI_Infrastructure\u002FReleases\u002Fv4.0.3\ncp -r .claude ~\u002F && cd ~\u002F.claude && bash install.sh\n```\n\n请参阅 [Releases 目录](https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure\u002Ftree\u002Fmain\u002FReleases)，获取所有版本及升级说明。\n\n## 贡献者\n\n感谢 @rikitikitavi2012-debug、@Spirotot 和 @catchingknives 的贡献！","2026-03-02T02:30:06",{"id":178,"version":179,"summary_zh":180,"released_at":181},127826,"v4.0.1","## PAI v4.0.1 — 升级路径与偏好设置\n\n**补丁版本，修复了社区反馈的升级困难、偏好设置缺失以及状态栏错误等问题。**\n\n### 变更内容\n\n#### 1. 升级路径文档\n主 `README` 现在分别提供了 **全新安装** 和 **从旧版本升级** 两部分，其中详细列出了备份、合并及升级后的检查清单步骤。\n\n#### 2. 温度单位偏好设置\n`settings.json` 中新增了 `preferences.temperatureUnit` 字段，可选择华氏或摄氏。安装程序会在设置过程中提示用户选择，状态栏也会根据该设置显示相应的温度单位。\n\n#### 3. 状态栏错误修复\n根据社区报告的 PR #762、#780 和 #806 进行了以下修复：\n- **移除硬编码时区** — 现在从 `settings.json` 中读取 `principal.timezone`，而非硬编码的 `America\u002FLos_Angeles`。\n- **修复上下文回退问题** — `\u002Fclear` 命令不再显示过时的上下文百分比。\n- **移除启动估算值** — 去除了会夸大新会话上下文百分比的自校准估算值。\n- **修复 f-string 语法问题** — Python f-string 内部嵌套的转义双引号在某些 Shell 中会导致解析错误。\n\n#### 4. 常见问题解答更新\n- 移除了过时的 Python 引用 — PAI v4.0 已经使用 TypeScript 和 Bash，而非 Python。\n- 改进了“如果我弄坏了怎么办？”的回答，增加了备份、合并和保留数据的指导说明。\n\n### 文件变更（相对于 v4.0.0）\n\n| 文件 | 变更 |\n|------|--------|\n| `settings.json` | 添加了 `preferences.temperatureUnit` 字段 |\n| `statusline-command.sh` | 温度单位偏好、动态时区、上下文\u002F启动估算及 f-string 相关修复 |\n| `PAI-Install\u002Fengine\u002Ftypes.ts` | 将 `temperatureUnit` 添加到收集的数据和 PAIConfig 中 |\n| `PAI-Install\u002Fengine\u002Fconfig-gen.ts` | 输出 `preferences` 部分 |\n| `PAI-Install\u002Fengine\u002Factions.ts` | 在身份验证步骤中添加温度单位提示，并在升级时保留用户偏好 |\n\n### 快速手动升级（从 v4.0.0）\n\n如果您只想应用这些修复而无需重新运行安装程序：\n\n1. 在您的 `~\u002F.claude\u002Fsettings.json` 中添加以下内容：\n```json\n\"preferences\": {\n  \"temperatureUnit\": \"celsius\"\n}\n```\n2. 将本版本中的更新版 `statusline-command.sh` 复制覆盖您现有的文件。\n3. 删除天气缓存：`rm -f ~\u002F.claude\u002FMEMORY\u002FSTATE\u002Fweather-cache.json`\n\n**[完整发布说明 →](https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure\u002Fblob\u002Fmain\u002FReleases\u002Fv4.0.1\u002FREADME.md)**","2026-02-28T18:38:37",{"id":183,"version":184,"summary_zh":185,"released_at":186},127827,"v4.0.0","## PAI v4.0.0 — 精简高效\n\n**38 个扁平化的技能目录被压缩为 12 个层级化分类。已移除废弃系统，上下文占用空间减半。**\n\n### 亮点\n\n1. **技能重组** — 将 38 个扁平目录整合为 12 个具有清晰归属边界的层级化分类。\n2. **算法 v3.5.0** — 重大推理升级，优化了阶段执行与 ISC 跟踪。\n3. **CLAUDE.md 模板系统** — `BuildCLAUDE.ts` + `SessionStart` 钩子可根据 `settings.json` 中的变量，基于模板生成 CLAUDE.md。\n4. **安全净化** — 清理了 33+ 个文件中的硬编码路径、API 密钥及个人数据。\n5. **上下文精简** — 启动时上下文占比约 19%（相比 v3.0 的约 38% 显著降低）。\n\n### 数据统计\n\n- 技能数量：63 个（原 38 个）\n- 钩子数量：21 个（原 20 个）\n- 工作流数量：180 个（原 162 个）\n- 算法版本：v3.5.0（原 v1.4.0）\n- 技能分类：13 类\n\n### 破坏性变更\n\n- 技能目录结构调整 — 扁平的 `skills\u002FSkillName\u002F` 改为层级化的 `skills\u002FCategory\u002FSkillName\u002F`\n- `CLAUDE.md` 现由模板生成 — 请编辑 `CLAUDE.md.template` 替代\n- 移除多个已弃用的系统（旧语音服务器、遗留迁移工具）\n\n**[完整发布说明 →](https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure\u002Fblob\u002Fmain\u002FReleases\u002Fv4.0.0\u002FREADME.md)**","2026-02-28T18:38:18",{"id":188,"version":189,"summary_zh":190,"released_at":191},127828,"v3.0.0","## PAI v3.0.0 — 算法走向成熟\n\n**约束提取、构建漂移预防、持久化需求文档以及并行循环执行**\n\n### 七大核心特性\n\n1. **算法 v1.4.0** — 从原型迈向生产级。新增约束提取、自我质询、构建漂移预防、验证预演、支持并行工作者的循环模式，以及与计划模式的集成。\n2. **完整安装系统** — 基于 Electron 的 GUI 向导 + CLI 备用方案，具备 API 密钥自动发现与基于模板的配置生成功能。\n3. **10 项新技能** — IterativeDepth、Science、Remotion、WorldThreatModelHarness、WriteStory、Evals、USMetrics、ExtractWisdom、Cloudflare、Sales。\n4. **代理团队\u002F蜂群** — 多代理协同执行，共享任务列表、并行工作者及消息传递机制。\n5. **持久化需求文档** — 能够跨会话保存的需求文档，并实现全生命周期跟踪。\n6. **语音人格系统** — 可配置的人格特质，用于塑造语音表达风格。\n7. **内联验证方法** — 每个 ISC 准则都标注了明确的验证方式（CLI、测试、静态分析、浏览器、Grep、读取、自定义）。\n\n### 数据统计\n\n- 技能数量：38 项（原 28 项）\n- 钩子数量：20 个（原 17 个）\n- 工作流数量：162 条\n- 文件总数：1,229 个\n- 算法版本：v1.4.0（原 v0.2.25）\n\n### 重大变更\n\n- `INSTALL.ts` 被 `PAI-Install\u002F` 目录取代（全新安装系统）\n- 算法版本由 v0.2.x 升级至 v1.4.0（推理逻辑全面重写）\n- 语音服务器新增 Qwen3 本地 TTS 功能，与 ElevenLabs 共存\n- `settings.json` 结构更新（由安装程序自动生成）\n\n**[完整发布说明 →](https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPAI\u002Fblob\u002Fmain\u002FReleases\u002Fv3.0\u002FREADME.md)**\n\n---\n\n*这一版本标志着算法从原型阶段迈入系统化阶段。*","2026-02-15T09:12:13",{"id":193,"version":194,"summary_zh":195,"released_at":196},127829,"v2.5.0","\u003Cdiv align=\"center\">\n\n\u003Cimg src=\"https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FTheAlgorithm\u002Fraw\u002Fmain\u002Fassets\u002Falgorithm-blocks.png\" alt=\"PAI v2.5\" width=\"200\">\n\n# PAI v2.5.0 — 思考更深入，执行更快\n\n**该算法学会了反思自己的思考方式，并能并行处理更多工作**\n\n[![GitHub Release](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FRelease-v2.5.0-8B5CF6?style=flat&logo=github)](https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPAI\u002Freleases\u002Ftag\u002Fv2.5.0)\n[![技能](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FSkills-28-22C55E?style=flat)](skills\u002F)\n[![钩子](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FHooks-17-3B82F6?style=flat)](hooks\u002F)\n[![工作流](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FWorkflows-356-F97316?style=flat)](skills\u002F)\n\n![讨论](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdiscussions\u002Fdanielmiessler\u002FPAI?style=flat&logo=github&label=Discussions&color=EAB308)\n![每月提交](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fcommit-activity\u002Fm\u002Fdanielmiessler\u002FPAI?style=flat&logo=git&label=Commits%2Fmo&color=F59E0B)\n![仓库大小](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Frepo-size\u002Fdanielmiessler\u002FPAI?style=flat&logo=database&label=Repo%20Size&color=D97706)\n\n\u003C\u002Fdiv>\n\n---\n\n## 变化内容\n\nv2.4 引入了“算法”——一种包含可验证标准的七阶段问题解决方法。它确实有效，但仍有盲点：它依据习惯而非分析来选择能力，未加论证就跳过思维工具，并且在任务可以同时进行时仍逐一执行。\n\nv2.5 解决了这三个问题。\n\n### v2.5 所解决的问题\n\nv2.4 的算法：\n```\n挑选熟悉的工具 → 依次执行 → 希望选对了所需的能力\n```\n\nv2.5 的算法：\n```\n钩子建议能力 → 逆向推导真实意图 → 根据 ISC 标准验证建议 → 为每项工具的取舍提供理由 → 并行执行独立任务 → 核实结果\n```\n\n关键区别在于元认知。现在，该算法会在行动前先反思自身的思考方式。\n\n---\n\n## 三大新特性\n\n### 1. 双重能力筛选\n\n过去，能力筛选是隐式的——系统会默认采用熟悉模式（通常是工程师 + 研究），而不管任务实际需要什么。\n\n如今，筛选分为两个明确的步骤：\n\n**第一步 — 钩子提示（在算法开始前）**\n\nFormatReminder 钩子会对原始提示进行 AI 推理，提出能力、技能和思维工具的建议。这些只是初步建议，相当于一个起点，而非最终决定。\n\n**第二步 — THINK 验证（OBSERVE 完成后）**\n\n在完成逆向推导和 ISC 标准分析之后，THINK 阶段会对所有建议进行验证。技能会被添加、移除或确认。第二步具有权威性。\n\n```\n🔍 技能检查（根据 ISC 标准验证钩子建议）：\n│ 钩子建议：创建技能：更新技能\n│ ISC 要求：架构设计（而不仅仅是技能更新）\n│ 最终技能：创建技能：更新技能 + 架构咨询\n```\n\n**为什么需要两步？** 钩子可以从原始提示中提供一个起点。然而，OBSERVE 阶段会改变整体情况。一个看似简单的请求，实际上可能涉及……","2026-01-31T08:03:28",{"id":198,"version":199,"summary_zh":200,"released_at":201},127830,"v2.4.0","# PAI v2.4.0 — 算法\n\n**我们首次尝试在 PAI 中内置一种通用问题解决工具，以追求欣快式的惊喜。**\n\n## 算法\n\nv2.4 引入了一种结构化的七阶段问题解决方法。这是一次让 AI 工作更具可验证性的实验。\n\n| 阶段 | 名称 | 目的 |\n|-------|------|---------|\n| 1 | **观察** | 收集信息，理解当前状态 |\n| 2 | **思考** | 分析意图、期望结果及失败模式 |\n| 3 | **计划** | 构建理想状态标准（ISC）——可测试的成功条件 |\n| 4 | **构建** | 组装解决方案的各个组件 |\n| 5 | **执行** | 采取行动，并根据标准跟踪进展 |\n| 6 | **验证** | 用证据确认所有标准均已满足 |\n| 7 | **学习** | 汲取经验教训，明确下一步行动 |\n\n## 理想状态标准（ISC）\n\n其核心机制是仅由 8 个词组成的、可二元测试的成功条件。只要能在前期定义出可测试的标准，就能真正验证某件事是否成功。\n\n## 新特性\n\n| 特性 | 描述 |\n|---------|-------------|\n| **算法** | 带有 ISC 跟踪的七阶段问题解决流程 |\n| **29 种技能** | 从 20 种增至 29 种，覆盖更多领域 |\n| **15 个钩子** | 优化后的钩子架构 |\n| **331 个工作流** | 全面的自动化库 |\n| **改进的向导** | 更好的交互式设置，支持语音 |\n| **语音通知** | 集成 ElevenLabs，提供语音提醒 |\n| **可观测性仪表板** | 实时系统监控 |\n| **白名单强制执行** | 用于发布的安全控制 |\n\n## 安装\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPAI.git\ncd PAI\u002FReleases\u002Fv2.4\n\n[ -d ~\u002F.claude ] && mv ~\u002F.claude ~\u002F.claude-backup-$(date +%Y%m%d)\ncp -r .claude ~\u002F\n\ncd ~\u002F.claude && bun run PAIInstallWizard.ts\n```\n\n## 命名代理\n\n| 代理 | 专长 |\n|-------|-----------|\n| **算法** | ISC 跟踪与验证工作 |\n| **架构师** | 系统设计、分布式系统 |\n| **工程师** | 测试驱动开发、实现模式 |\n| **艺术家** | 视觉内容、提示工程 |\n| **设计师** | 用户体验\u002F用户界面、可访问性 |\n| **质量保证测试员** | 浏览器自动化、验证 |\n| **渗透测试员** | 安全测试 |\n| **实习生** | 高自主性的通才 |\n\n## 完整文档\n\n完整文档请参阅 [Releases\u002Fv2.4\u002FREADME.md](https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPAI\u002Fblob\u002Fmain\u002FReleases\u002Fv2.4\u002FREADME.md)。","2026-01-24T05:45:33",{"id":203,"version":204,"summary_zh":205,"released_at":206},127831,"v2.3.0","# PAI v2.3.0 - 完整发行版回归\n\n**自 v2.0 以来最重要的版本。** PAI v2.3 在保留模块化 Packs 系统的同时，重新推出了完整且功能齐全的发行版。现在你可以通过两种方式安装 PAI：要么单独获取特定功能的 Packs，要么直接下载一个完整的 `.claude\u002F` 目录并立即开始使用。\n\n---\n\n## 🎯 亮点\n\n### 完整发行版回归\n`Releases\u002Fv2.3\u002F.claude\u002F` 目录包含一个 **完整、可运行的 PAI 安装包**。将其复制到 `~\u002F.claude\u002F` 即可直接运行，无需任何额外组装。\n\n这回应了自 v2.0 发布以来最常见的反馈：“我非常喜欢模块化的设计，但有时候就是想要一个完整的解决方案。”现在你既可以选择模块化方式，也可以直接使用完整发行版。\n\n### 臻于愉悦的惊喜\nPAI v2.3 引入了“臻于愉悦的惊喜”这一核心理念——它将成为驱动每一次交互的唯一目标。我们追求的不仅是任务的完成，更是结果的彻底与周到，让你真正感到惊喜与满足。这一理念现已贯穿于整个文档和技能设计之中。\n\n### 自定义为第一原则\nPAI 的创始原则已更新。**以你的目标为导向的自定义**现已成为第一原则——PAI 的存在就是为了帮助你实现人生目标。它致力于 democratize 访问个性化代理基础设施的机会，这些基础设施能够理解你的上下文、偏好和历史记录。\n\n---\n\n## 📦 v2.3 包含的内容\n\n### 完整发行版 (`Releases\u002Fv2.3\u002F.claude\u002F`)\n\n| 组件 | 数量 | 描述 |\n|-----------|-------|-------------|\n| **技能** | 22 | 包含工作流的完整技能模块 |\n| **命名代理** | 11 | 具有不同人格的代理（艺术家、建筑师、工程师等）|\n| **钩子** | 15 | 生命周期事件处理器 |\n| **织物模式** | 100+ | 可复用的提示模板 |\n\n**包含的技能：**\n- **核心**：CORE、System、THEALGORITHM\n- **研究**：Research、OSINT、PrivateInvestigator、AnnualReports\n- **创意**：Art、Prompting、Council、RedTeam\n- **开发**：Browser、CreateCLI、CreateSkill、Agents\n- **基础设施**：PAIUpgrade、BrightData、Telos、FirstPrinciples、Recon\n\n**命名代理：**\nArtist、QATester、Architect、Designer、GeminiResearcher、GrokResearcher、CodexResearcher、Pentester、Intern、Engineer、ClaudeResearcher\n\n### 更新后的 Packs（共 23 个）\n\n所有 Packs 均已升级至 v2.3 标准：\n- 统一采用 AI 向导格式的 INSTALL.md 文件\n- 所有 README 中添加图标引用\n- 统一的版本管理\n\n---\n\n## ⚡ 安装指南\n\n### 选项 1：完整安装（推荐新用户）\n\n```bash\n# 下载 v2.3 发行版\n# 将 .claude\u002F 复制到主目录\ncp -r Releases\u002Fv2.3\u002F.claude ~\u002F.claude\n\n# 运行安装向导\ncd ~\u002F.claude && bun run install.ts\n\n# 开始使用 PAI\nclaude\n```\n\n### 选项 2：单独安装 Packs（现有用户）\n\n浏览 [`Packs\u002F`](Packs\u002F) 并根据需求安装特定功能的 Packs。每个 Pack 都是独立的，自带完整的使用说明。\n\n---\n\n## 🔄 自 v2.1.0 以来的变化\n\n### 重大更新\n- **恢复完整发行版** - 在 `Releases\u002Fv2.3\u002F` 中提供完整的 `.claude\u002F` 目录\n- **23 个 Packs 更新** 至 v2.3 版本\n- **新增“臻于愉悦的惊喜”理念** as ","2026-01-16T06:24:13",{"id":208,"version":209,"summary_zh":210,"released_at":211},127832,"v2.1.0","# PAI v2.1.0 - 核心重大升级与全新升级技能\n\n本次发布对核心基础设施进行了重大更新，并引入了一项用于跟踪系统升级的新技能。\n\n## 重大更新\n\n### pai-core-install v1.2.0（重大升级）\n\n核心包迎来了迄今为止最大的一次更新：\n\n- **新增：MEMORY\u002F 骨架** - 包含会话历史、学习内容、决策记录及状态管理的11目录结构\n- **新增：settings.json.template** - 完整的钩子配置模板，预配置了所有事件\n- **USER\u002F 目录**（15个模板） - 用于身份、联系人、偏好等个人配置\n- **SYSTEM\u002F 目录**（17个模板） - 提供技能、钩子、记忆、委托等方面的系统架构文档\n- **向导式安装** - 安装程序现在会以清晰的阶段引导您逐步完成设置\n\n此次更新将 pai-core-install 从一个基础框架转变为一个完整的基础设施平台。\n\n### pai-browser-skill v1.2.0\n\n以调试为先的理念增强了浏览器自动化功能：\n- 始终开启的诊断与控制台捕获\n- 会话自动启动能力\n- 通过基于文件的 MCP 模式节省99%的令牌\n\n### 新增：pai-upgrades-skill v1.0.0\n\n一项用于跟踪和管理 PAI 系统升级的新技能：\n- 监控新功能与能力\n- 跟踪升级机会\n- 维护升级历史\n- 帮助规划与优先安排系统改进\n\n## 文档更新\n\n- 更新了软件包计数徽章（8 → 10个包）\n- 修复了捆绑包引用（Kai\u002F → Official\u002F）\n- 根据新技能更新了安装顺序\n- 在所有文档中刷新了版本号\n\n## 安装\n\n```bash\n# 全新安装\ncd PAI\u002FBundles\u002FOfficial && bun run install.ts\n\n# 升级现有安装\ncd PAI\u002FBundles\u002FOfficial && bun run install.ts --update\n```\n\n## 下一步计划\n\n- 更多技能即将加入软件包生态\n- 所有软件包都将增强向导式安装体验\n- 欢迎社区贡献！\n\n---\n\n**完整变更日志：** https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPAI\u002Fcompare\u002Fv2.0.0...v2.1.0","2026-01-08T12:11:49",{"id":213,"version":214,"summary_zh":215,"released_at":216},127833,"v2.0.0","## PAI v2.0.0 - PAI Packs 系统发布\n\n### 重大架构升级\n- 从“镜像系统”模式转变为模块化的 **PAI Packs**\n- Packs 是自包含、可安装的 AI 能力包\n- 平台无关设计：兼容 Claude Code、OpenCode、Gemini Code、GPT-Codex 或自定义系统\n\n### 可用 Packs（8 个）\n- **kai-core-install** - 基础包，提供技能路由、身份管理和架构追踪功能\n- **kai-hook-system** - 事件驱动的钩子架构\n- **kai-history-system** - 自动上下文追踪\n- **kai-voice-system** - 使用 ElevenLabs TTS 的语音通知功能\n- **kai-art-skill** - 视觉内容创作\n- **kai-prompting-skill** - 元提示系统\n- **kai-agents-skill** - 动态代理组合\n- **kai-observability-server** - 实时代理监控\n\n### 可用 Bundle（1 个）\n- **kai-core-bundle** - 完整的 Kai 基础设施（包含 4 个基础 Pack）\n\n### 为何进行此次变更？\n- v1.x 试图完整镜像整个 Kai 系统，但过于脆弱且依赖关系复杂\n- v2.0 将经过实战检验的功能拆分为独立可安装的模块\n- 每个 Pack 就像在《黑客帝国》中学习功夫一样，是一次完整的技能下载\n\n### 文档\n- 完整的 Pack 模板规范\n- 全面的 Pack 系统文档\n- 更新的 README，新增 14 条创始原则及安装指南","2025-12-30T16:48:11",{"id":218,"version":219,"summary_zh":220,"released_at":221},127834,"v0.9.0","## PAI v0.9.0 - 平台无关发布\n\n本次发布致力于使 PAI 完全可移植且易于分叉。**你的 AI，你的身份，你的系统。**\n\n### 亮点\n\n#### 可观测性仪表板\n- 在 `.claude\u002FObservability\u002F` 提供完整的实时代理监控\n- 所有代理活动的 WebSocket 流式传输\n- 实时脉搏图、事件时间线和泳道视图\n- 多种主题（东京之夜、Nord、猫屎咖啡等）\n- 敏感数据的安全混淆处理\n- 启动命令：`~\u002F.claude\u002FObservability\u002Fmanage.sh start`\n\n#### 通用化代理身份\n- 所有代理引用现在使用 `process.env.DA || 'main'`\n- 不再使用硬编码名称——您的 DA 名称贯穿整个系统\n- 可观测性仪表板显示您配置的身份\n- 颜色系统支持任何 DA 名称，并提供智能默认值\n\n#### 平台无关的配置\n- 在 `settings.json` 中新增 `_envDocs`，解释所有环境变量\n- 清晰分离：`settings.json` 用于身份和路径配置，`.env` 用于 API 密钥\n- `DA`（数字助理名称）——您的 AI 的身份\n- `PAI_DIR`——所有配置的根目录\n- `TIME_ZONE`——可配置的时区，用于时间戳\n\n#### 状态栏增强\n- 终端颜色兼容模式，适用于不同终端模拟器\n- 动态显示 Claude Code 版本\n- 更整洁的欢迎信息格式\n\n#### 技能系统改进\n- 全面采用标准的标题大小写文件命名规范\n- 标准化的技能-工作流-通知脚本，便于仪表板检测\n- 所有路径均使用 `${PAI_DIR}\u002F`，实现位置无关安装\n\n### 破坏性变更\n无——本次发布完全向后兼容。\n\n### 升级说明\n```bash\ncd ~\u002F.claude  # 或您的 PAI_DIR\ngit pull origin main\n```\n\n如果您是全新安装，请按照 README.md 中的快速入门指南进行操作。\n\n---\n\n**完整更新日志**：https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure\u002Fcompare\u002Fv0.7.0...v0.9.0","2025-12-02T04:01:42",{"id":223,"version":224,"summary_zh":225,"released_at":226},127835,"v0.7.0","## What's New in v0.7.0\n\nMajor release with 18 commits of improvements since v0.6.0.\n\n### 🔒 Security\n- Comprehensive prompt injection and input validation guidance\n- Security hardening for art skill (removed Discord\u002FMidjourney integration)\n\n### ✨ New Skills\n- **BrightData** - Four-tier progressive web scraping skill with complete MCP installation docs\n- **Art** - PAI visual aesthetic system (Tron + Anthropic + Excalidraw fusion)\n- **Story-explanation** - Narrative summary generation\n- **System-createskill** and **system-createcli** - Meta-skills for building skills and CLIs\n\n### 🏗️ Infrastructure\n- `PAI_DIR` wrapper and protection system\n- PAI vs Kai clarity documentation\n- Comprehensive event capture hook system\n- Complete MCP installation instructions\n\n### 📚 Documentation\n- Fixed all broken documentation links and references\n- CORE skill is now the canonical documentation source\n- Added 10 sanitized CORE skill files for public template\n- Hook system branding updated from Kai to PAI\n\n---\n\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure\u002Fcompare\u002Fv0.6.0...v0.7.0","2025-11-26T19:12:03",{"id":228,"version":229,"summary_zh":230,"released_at":231},127836,"v0.6.0","# 🔥 v0.6.0 - MAJOR UPGRADE: Repository Restructured with .claude\u002F Directory\n\n## ⚠️ BREAKING CHANGE\n\n**The repository structure has been completely reorganized.** All PAI infrastructure now lives in the `.claude\u002F` directory to properly mirror how PAI works in production.\n\n---\n\n## 🎯 What Changed\n\n### Repository Structure (Before → After)\n\n**Before (v0.5.0):**\n```\n\u002FPAI\u002F\n├── agents\u002F\n├── commands\u002F\n├── documentation\u002F\n├── hooks\u002F\n├── skills\u002F\n├── voice-server\u002F\n├── settings.json\n└── README.md\n```\n\n**After (v0.6.0):**\n```\n\u002FPAI\u002F\n├── .claude\u002F                 # ← All PAI infrastructure\n│   ├── agents\u002F\n│   ├── commands\u002F\n│   ├── documentation\u002F\n│   ├── hooks\u002F\n│   ├── skills\u002F\n│   ├── voice-server\u002F\n│   ├── settings.json\n│   ├── .mcp.json\n│   └── setup.sh\n├── README.md               # GitHub files stay at root\n├── LICENSE\n└── .gitignore\n```\n\n---\n\n## 🚀 Why This Matters\n\n### The Problem\nUsers reported issues with PAI not working correctly because the repository structure didn't match the actual working system. The real PAI system expects all infrastructure to live in `~\u002F.claude\u002F`, but the repo had everything at root level, causing:\n- Confusion about where files should go\n- Compatibility issues during setup\n- Difficulty using the repo as a reference\n\n### The Solution\n1. **Proper Emulation:** Repository now accurately represents how PAI works in production\n2. **Easier Setup:** Users can see exactly how their `~\u002F.claude\u002F` directory should be structured\n3. **Less Confusion:** Clear separation between GitHub files and PAI infrastructure\n4. **Better Documentation:** Structure itself serves as documentation\n5. **Reference Implementation:** Can be copied\u002Freferenced directly for setup\n\n---\n\n## 📦 Installation (New Users)\n\n```bash\n# Clone the repository\ngit clone https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure.git\ncd Personal_AI_Infrastructure\n\n# Copy .claude directory to your home\ncp -r .claude ~\u002F.claude\n\n# Configure environment\ncp ~\u002F.claude\u002F.env.example ~\u002F.claude\u002F.env\nvim ~\u002F.claude\u002F.env  # Add your API keys\n\n# Set environment variables (add to ~\u002F.zshrc or ~\u002F.bashrc)\nexport PAI_DIR=\"$HOME\u002FPAI\"  # Points to repo root (for development)\nexport PAI_HOME=\"$HOME\"\n```\n\n---\n\n## 🔄 Migration (Existing Users)\n\n**No action required for existing installations!** Your `~\u002F.claude\u002F` directory is already set up correctly. This change only affects the repository structure, not your working installation.\n\nIf you want to update your installation:\n```bash\ncd $PAI_DIR  # Your PAI repo\ngit pull\n# Your existing ~\u002F.claude\u002F continues to work as-is\n```\n\n---\n\n## 📋 What's Included\n\n### Files Moved to `.claude\u002F`\n- ✅ 8 agent definitions (`agents\u002F`)\n- ✅ 7 command files (`commands\u002F`)\n- ✅ 13 documentation files (`documentation\u002F`)\n- ✅ 11 hook scripts (`hooks\u002F`)\n- ✅ Full skills system including Fabric repo (`skills\u002F`)\n- ✅ Complete voice server (`voice-server\u002F`)\n- ✅ Configuration files (`settings.json`, `.mcp.json`, `.env.example`)\n- ✅ Setup and utility scripts\n\n### Files Kept at Root (GitHub Infrastructure)\n- ✅ README.md\n- ✅ LICENSE\n- ✅ SECURITY.md\n- ✅ .gitignore\n- ✅ .github\u002F\n\n---\n\n## 📚 Documentation Updates\n\nAll documentation has been updated to reflect the new structure:\n- ✅ README.md - Prominent MAJOR UPGRADE notice\n- ✅ architecture.md - v0.6.0 structure notice\n- ✅ how-to-start.md - Installation instructions updated\n- ✅ QUICK-REFERENCE.md - Path references updated\n- ✅ All path examples changed from `${PAI_DIR}\u002F*` to `~\u002F.claude\u002F*`\n\n---\n\n## 🔗 Related Issues\n\nThis release addresses user-reported compatibility issues and confusion about repository structure. The `.claude\u002F` organization makes PAI a proper reference implementation.\n\n---\n\n## 🙏 Thank You\n\nThank you to everyone who reported issues and provided feedback about the repository structure. This change makes PAI more accessible and easier to understand for new users.\n\n---\n\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure\u002Fcompare\u002Fv0.5.0...v0.6.0","2025-10-21T19:09:44",{"id":233,"version":234,"summary_zh":235,"released_at":236},127837,"v0.4.0","## 🔥 BREAKING CHANGE: Repository Restructure\n\nPAI has been restructured for better GitHub UX and renamed for clarity!\n\n### ⚠️ BREAKING CHANGES\n\n**1. PAI_DIR Environment Variable:**\n```bash\n# OLD (v0.3.x and earlier):\nexport PAI_DIR=\"\u002Fpath\u002Fto\u002FPAI\u002FPAI_DIRECTORY\"\n\n# NEW (v0.4.0+):\nexport PAI_DIR=\"\u002Fpath\u002Fto\u002FPAI\"\n```\n\n**2. Repository Renamed:**\n- **Old:** `danielmiessler\u002FPAI`\n- **New:** `danielmiessler\u002FPersonal_AI_Infrastructure`\n- GitHub automatically redirects old URL to new URL\n- All existing links continue to work\n\n### 🏗️ What Changed\n\n**Flattened Repository Structure:**\n```\nBefore (v0.3.x):           After (v0.4.0):\nPAI\u002F                       PAI\u002F\n├── PAI_DIRECTORY\u002F         ├── agents\u002F\n│   ├── agents\u002F            ├── commands\u002F\n│   ├── commands\u002F          ├── documentation\u002F\n│   ├── skills\u002F            ├── hooks\u002F\n│   └── ...                ├── skills\u002F\n├── README.md              ├── voice-server\u002F\n└── ...                    ├── .env.example\n                           ├── .mcp.json\n                           ├── settings.json\n                           ├── README.md\n                           └── ...\n```\n\n**Repository Renamed:**\n- Repository renamed from `PAI` to `Personal_AI_Infrastructure`\n- Clearer, more descriptive name\n- Better SEO and discoverability\n- GitHub automatically redirects old URL\n\n**Benefits:**\n- ✅ Functional directories immediately visible on GitHub\n- ✅ No extra click required to see actual content\n- ✅ More conventional repository structure\n- ✅ Cleaner navigation for new users\n- ✅ Better first impression for newcomers\n\n### 📦 Migration Steps\n\n**1. Update PAI_DIR in shell config:**\n```bash\n# Edit ~\u002F.zshrc or ~\u002F.bashrc\nexport PAI_DIR=\"\u002Fpath\u002Fto\u002FPAI\"  # Remove \u002FPAI_DIRECTORY\n```\n\n**2. Reload your shell:**\n```bash\nsource ~\u002F.zshrc  # or source ~\u002F.bashrc\n```\n\n**3. Pull latest changes:**\n```bash\ncd \u002Fpath\u002Fto\u002FPAI\ngit pull\n```\n\n**4. Update git remote URL (optional but recommended):**\n```bash\ngit remote set-url origin git@github.com:danielmiessler\u002FPersonal_AI_Infrastructure.git\n# Or for HTTPS:\ngit remote set-url origin https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure.git\n```\n\n### 📖 Documentation Updates\n\n- Added prominent v0.4.0 breaking change notice to README\n- Updated installation instructions\n- Updated all path references throughout documentation\n- Updated environment variables section\n- Added comprehensive migration guide\n- Made core mission statement prominent and centered\n\n### 🎯 Why This Change\n\nThe previous nested structure (`PAI_DIRECTORY\u002F`) added an unnecessary layer that made the repository less intuitive on GitHub. Users had to click into `PAI_DIRECTORY\u002F` to see the actual functional directories.\n\nThis restructure brings PAI in line with conventional repository layouts where functional code is immediately visible at the root.\n\n### 📊 Statistics\n\n- 810+ files moved to root\n- Complete Fabric repository (242 patterns) structure preserved\n- All git history maintained (100% renames)\n- Zero content changes - pure structural refactor\n\n### 🔗 Links\n\n- **Repository:** https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure\n- **Documentation:** https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure\u002Ftree\u002Fmain\u002Fdocumentation\n- **Issues:** https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure\u002Fissues\n\n---\n\n**Full Changelog:** https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPersonal_AI_Infrastructure\u002Fcompare\u002Fv0.3.2...v0.4.0","2025-10-17T23:03:47",{"id":238,"version":239,"summary_zh":240,"released_at":241},127838,"v0.2.4","## 🧹 README Cleanup & Organization\n\nThis minor release focuses on improving the README organization and reducing visual clutter.\n\n### ✨ What's New\n\n- **📂 Collapsed Updates Section**: Implemented two-level collapsing structure using nested `\u003Cdetails>` tags (#27)\n- **🎯 Better Navigation**: Significantly reduced vertical space taken by updates section\n- **📊 Space Optimization**: Updates now cleanly collapsed while maintaining full accessibility\n- **🧹 Issue Management**: Closed PR #24 after recent voice system updates\n- **✅ Improved UX**: All information remains accessible through intuitive collapse\u002Fexpand interface\n\n### 🔧 Changes\n\n- Added parent `\u003Cdetails>` wrapper around all update entries\n- Each update entry maintains its own collapsible state\n- Most recent update (October 2, 2025) open by default\n- Cleaner README with less scrolling required\n\n### 📦 Installation\n\nSee the [Quick Start Guide](https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPAI#-quick-start) for installation instructions.\n\n---\n\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPAI\u002Fcompare\u002Fv0.2.3...v0.2.4","2025-10-07T01:35:57",{"id":243,"version":244,"summary_zh":245,"released_at":246},127839,"v0.2.3","## 🎯 Release v0.2.3 - Complete Visibility & Portability Update\n\n### 🚀 Overview\n\nThis release completes the transformation of PAI into a fully portable, vendor-agnostic, and visible personal AI infrastructure system. No more hidden files, no more hardcoded paths, and full compatibility with all file management tools.\n\n### ✨ Key Improvements\n\n#### 📁 **Complete Visibility**\n- **No More Hidden Files**: All configuration files are now visible\n  - `.claude` → `PAI_DIRECTORY` \n  - `.zshrc` → `zshrc-aliases`\n  - `.env-example` → `env-example`\n- **Tool Compatibility**: Works perfectly with Obsidian, file managers, and all tools\n\n#### 🌍 **True Vendor Independence**  \n- **Not Just for Claude**: PAI works with any AI platform\n- **Supports**: Claude, GPT-4, Gemini, and any future AI systems\n- **Flexible Architecture**: Adapt PAI to your preferred AI vendor\n\n#### 🔧 **Full Portability**\n- **Dynamic Paths**: All hardcoded paths replaced with `${PAI_DIR}`\n- **Install Anywhere**: Put PAI wherever you want on your system\n- **Consistent Configuration**: `.env` lives in `${PAI_DIR}\u002F.env`\n- **No Broken Links**: Removed all problematic symbolic links\n\n### 🐛 Issues Fixed\n- ✅ **#21**: Fixed hardcoded path in session-start-hook.ts\n- ✅ **#20**: Resolved .env file location confusion\n- ✅ Removed broken symbolic links\n- ✅ Updated all documentation for consistency\n\n### 📖 Documentation Updates\n- All references updated to use `${PAI_DIR}`\n- Clear installation instructions with new paths\n- Emphasized vendor-agnostic design throughout\n- Consistent configuration file locations\n\n### 🔄 Migration Notes\n\nIf upgrading from v0.2.2 or earlier:\n1. Rename your `.claude` directory to `PAI_DIRECTORY`\n2. Update your `PAI_DIR` environment variable\n3. Move any `.env` from home directory to `${PAI_DIR}\u002F.env`\n4. Configuration files no longer have dot prefixes\n\n### 🙏 Acknowledgments\nThanks to @saisat and @frederikb96 for issue reports and the community for feedback!\n\n---\n\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPAI\u002Fcompare\u002Fv0.2.2...v0.2.3","2025-09-28T22:33:19",{"id":248,"version":249,"summary_zh":250,"released_at":251},127840,"v0.2.2","## 🎯 Release v0.2.2 - Directory Visibility & Portability Update\n\n### 🚀 What's New\n\nThis release focuses on improving file visibility and system portability by removing hidden file prefixes and ensuring all paths are dynamic.\n\n### 📋 Changes\n\n#### 📁 **Directory Structure Improvements**\n- **Renamed `.claude` to `PAI_DIRECTORY`** for better visibility\n- PAI is now truly vendor-agnostic (works with Claude, GPT, Gemini, etc.)\n- Better compatibility with tools like Obsidian that have issues with dot-prefixed directories\n\n#### 🔧 **Path Portability**\n- **Fixed hardcoded paths** in `session-start-hook.ts` (Fixes #21)\n- All paths now use `${PAI_DIR}` environment variable\n- System is fully portable across different installation locations\n\n#### 📄 **Configuration Updates** \n- **.env location standardized** to `${PAI_DIR}\u002F.env` (Fixes #20)\n- All documentation updated to use consistent path\n- Central configuration location within PAI installation\n\n#### 👁️ **File Visibility**\n- Renamed `.zshrc` to `zshrc-aliases` (no dot prefix)\n- Renamed `.env-example` to `env-example` (no dot prefix)\n- Removed broken symbolic links\n- All configuration files now visible in file managers\n\n### 🐛 Bug Fixes\n- Fixed session-start-hook.ts hardcoded path issue (#21)\n- Resolved .env file location confusion (#20)\n- Removed broken symbolic links\n\n### 📖 Documentation\n- Updated all documentation to reflect new directory structure\n- Clarified PAI's vendor-agnostic architecture\n- Consistent use of `${PAI_DIR}` throughout\n\n### 🙏 Thanks\nThanks to @saisat and @frederikb96 for reporting issues!\n\n---\n\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPAI\u002Fcompare\u002Fv0.2.1...v0.2.2","2025-09-28T22:32:13",{"id":253,"version":254,"summary_zh":255,"released_at":256},127841,"v0.2.1","## 🚀 PAI v0.2.1 - Hooks System Update\n\n### What's Changed\n\n#### 🪝 Hooks System Overhaul\n- **Sanitized for Public Use**: All hook files cleaned of sensitive data\n- **Portable Installation**: All hooks now use `${PAI_DIR}` variable\n- **Voice Integration**: Enhanced intelligent voice responses \n- **Tab Title System**: Improved terminal tab title generation\n- **Session Management**: Better initialization and state tracking\n\n#### 📁 Files Updated\n- `session-start-hook.ts` - Session initialization with voice notifications\n- `stop-hook.ts` - Terminal tab titles and completion announcements\n- `user-prompt-submit-hook.ts` - Voice greeting system\n- `subagent-stop-hook.ts` - Agent-specific voice notifications  \n- `load-dynamic-requirements.ts` - Dynamic context loading wrapper\n- `hooks\u002FREADME.md` - Hook system documentation\n\n#### 📚 Documentation\n- Fixed date formatting issues in main README\n- Removed redundant configuration warnings\n- Updated September 26, 2025 release notes\n\n### Installation Note\nRemember to set the `PAI_DIR` environment variable to your PAI installation's `.claude` directory for the hooks to work correctly.\n\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPAI\u002Fcompare\u002Fv0.2...v0.2.1","2025-09-27T01:39:11",{"id":258,"version":259,"summary_zh":260,"released_at":261},127842,"v0.2","## 🚀 What's New in v0.2\n\n### ✨ Features\n- **Human 3.0 Integration**: Added direct link to projects fit together blog post\n- **Enhanced Documentation**: Improved README structure and navigation\n- **Better Badge Styling**: Fixed purple color consistency across badges\n\n### 🐛 Bug Fixes\n- Fixed purple color issues in badges (now using proper hex `#8B5CF6`)\n- Fixed stats display badge configuration\n- Improved table formatting in architecture section\n- Minor display improvements throughout documentation\n\n### 📚 Documentation\n- Added comprehensive v0.2 changelog to Recent Updates section\n- Enhanced readability and structure\n- Better navigation and information hierarchy\n\n### 📦 Installation\n```bash\n# Clone the repository\ngit clone https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPAI.git\ncd PAI\n\n# Follow setup instructions in README\nexport PAI_HOME=\"$HOME\"\n```\n\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fdanielmiessler\u002FPAI\u002Fcompare\u002Fv0.1...v0.2","2025-09-22T14:29:36",{"id":263,"version":264,"summary_zh":265,"released_at":266},127843,"v0.1","## 🎉 PAI v0.1 - Initial Release\n\n### **Open-Source Personal AI Infrastructure for Everyone**\n\nThis is the first official release of PAI (Personal AI Infrastructure), an open-source platform designed to democratize AI access and give everyone on Earth the power to augment their personal and professional lives with AI.\n\n### ✨ **Key Features**\n\n- **🏗️ UFC Context System** - Persistent memory architecture with hierarchical organization\n- **⚡ 60+ Custom Commands** - Ready-to-use commands for work and life management\n- **🤖 5 Specialized AI Agents** - Researcher, Engineer, Designer, Pentester, Architect\n- **🔌 MCP Server Ecosystem** - Integrations with Playwright, Stripe, Analytics, and more\n- **🪝 Dynamic Hook System** - Smart context loading and routing based on intent\n- **🎙️ Voice Server** - ElevenLabs integration for natural voice interaction\n- **📁 Plain Text Configuration** - Complete portability and transparency\n- **🤖 Platform Independent** - Works with Claude, GPT, Gemini, or any AI platform\n\n### 🚀 **Getting Started**\n\n1. Clone the repository\n2. Install Bun runtime\n3. Set PAI_HOME environment variable\n4. Configure API keys in .env\n5. Start using your personal AI infrastructure!\n\n### 📚 **Documentation**\n\nFull documentation available at:\n- Quick Start Guide\n- Architecture Overview\n- UFC Context System\n- Agent Documentation\n- Voice Server Setup\n\n### 🎯 **Mission**\n\nPAI embodies the belief that AI should augment human capability, not replace it. This release represents the first step toward Human 3.0 - where every person has access to powerful AI tools that enhance their unique abilities and help them thrive in an AI-powered world.\n\n### 🙏 **Acknowledgments**\n\nThanks to the early contributors and testers who helped shape this initial release. Special thanks to the Claude Code team at Anthropic for making this development possible.\n\n---\n\n**Created by [Daniel Miessler](https:\u002F\u002Fdanielmiessler.com)**\n\n🌐 [Blog](https:\u002F\u002Fdanielmiessler.com\u002Fblog) • 📧 [Newsletter](https:\u002F\u002Fnewsletter.danielmiessler.com) • 🎬 [YouTube](https:\u002F\u002Fwww.youtube.com\u002F@unsupervised-learning)","2025-09-21T23:02:05"]