derrick-ships/distill-it
GitHub: derrick-ships/distill-it
一个利用 Claude 自动分析 GitHub 仓库并生成功能学习文档与构建规范,同时构建交互式知识图谱的代码知识蒸馏工具。
Stars: 1 | Forks: 0
```
██████╗ ██╗███████╗████████╗██╗██╗ ██╗ ██╗████████╗
██╔══██╗██║██╔════╝╚══██╔══╝██║██║ ██║ ██║╚══██╔══╝
██║ ██║██║███████╗ ██║ ██║██║ ██║█████╗██║ ██║
██║ ██║██║╚════██║ ██║ ██║██║ ██║╚════╝██║ ██║
██████╔╝██║███████║ ██║ ██║███████╗ ███████╗ ██║ ██║
╚═════╝ ╚═╝╚══════╝ ╚═╝ ╚═╝╚══════╝ ╚══════╝ ╚═╝ ╚═╝
```




## 这是什么
一个知识库,在这里 **Claude 负责阅读,我只负责挑选 repos**。
给 Claude 一个 GitHub URL,它就会通过网络阅读代码库,端到端地追踪每个功能,并为每个功能编写两份文档:
| 层级 | 目标受众 | 目的 |
|-------|----------|---------|
| **学习文档** | 你 | 通俗易懂的深度剖析,可直接上传至 NotebookLM |
| **构建规范** | Claude(未来的会话) | 独立的技术指南,用于在任何新项目中重新实现该功能 |
所有这些都会汇入一个你可以浏览和搜索的交互式知识图谱。
## 浏览图谱
在浏览器中打开 [`graph/graph.html`](graph/graph.html)。
点击任意节点 → 查看摘要,以及指向学习文档、构建规范和原始 repo 的链接。
## 包含哪些内容
| 领域 | 功能 | 来源 |
|--------|---------|------|
| adaptive-parsing | 自适应元素重定位 | [Scrapling](https://github.com/D4Vinci/Scrapling) |
| document-conversion | Converter Pipeline | [markitdown](https://github.com/microsoft/markitdown) |
| document-conversion | PDF 转换 | [markitdown](https://github.com/microsoft/markitdown) |
| document-conversion | Office 文档转换 | [markitdown](https://github.com/microsoft/markitdown) |
| document-conversion | ZIP 归档遍历 | [markitdown](https://github.com/microsoft/markitdown) |
| plugin-architecture | 插件系统 | [markitdown](https://github.com/microsoft/markitdown) |
| file-detection | Magika 文件检测 | [markitdown](https://github.com/microsoft/markitdown) |
| media-processing | 图像 + LLM 字幕生成 | [markitdown](https://github.com/microsoft/markitdown) |
| media-processing | 音频转录 | [markitdown](https://github.com/microsoft/markitdown) |
| web-extraction | HTML 网页转换 | [markitdown](https://github.com/microsoft/markitdown) |
| web-extraction | YouTube 提取 | [markitdown](https://github.com/microsoft/markitdown) |
| ai-integration | Azure 文档智能 | [markitdown](https://github.com/microsoft/markitdown) |
| agent-architecture | Agent 输出契约 | [last30days-skill](https://github.com/mvanhorn/last30days-skill) |
| research-automation | 多源研究引擎 | [last30days-skill](https://github.com/mvanhorn/last30days-skill) |
| research-automation | 实体消歧 | [last30days-skill](https://github.com/mvanhorn/last30days-skill) |
| research-automation | 参与度信号排名 | [last30days-skill](https://github.com/mvanhorn/last30days-skill) |
| content-synthesis | 跨源聚类 | [last30days-skill](https://github.com/mvanhorn/last30days-skill) |
| credential-management | 多层级凭证 | [last30days-skill](https://github.com/mvanhorn/last30days-skill) |
| ai-automation | AI 规则引擎 | [inbox-zero](https://github.com/elie222/inbox-zero) |
| ai-automation | AI 回复起草 | [inbox-zero](https://github.com/elie222/inbox-zero) |
| inbox-cleanup | 批量退订器 | [inbox-zero](https://github.com/elie222/inbox-zero) |
| inbox-cleanup | 批量归档器 | [inbox-zero](https://github.com/elie222/inbox-zero) |
| email-platform | 邮件提供商抽象层 | [inbox-zero](https://github.com/elie222/inbox-zero) |
| agent-architecture | 有序后端路由 | [Agent-Reach](https://github.com/Panniantong/Agent-Reach) |
| diagnostics | 频道健康诊断 | [Agent-Reach](https://github.com/Panniantong/Agent-Reach) |
| credential-management | Cookie 凭证提取 | [Agent-Reach](https://github.com/Panniantong/Agent-Reach) |
| agent-distribution | Agent 驱动安装 | [Agent-Reach](https://github.com/Panniantong/Agent-Reach) |
| tts | ONNX TTS Pipeline | [supertonic](https://github.com/supertone-inc/supertonic) |
| tts | Flow-Matching 采样器 | [supertonic](https://github.com/supertone-inc/supertonic) |
| tts | 表情标签 | [supertonic](https://github.com/supertone-inc/supertonic) |
| ai-workflow | Agentic Loop | [open-design](https://github.com/nexu-io/open-design) |
| ai-workflow | Agent CLI 集成 | [open-design](https://github.com/nexu-io/open-design) |
| codegen | 设计工件生成 | [open-design](https://github.com/nexu-io/open-design) |
| design-systems | 设计系统库 | [open-design](https://github.com/nexu-io/open-design) |
| infrastructure | BYOK Proxy | [open-design](https://github.com/nexu-io/open-design) |
| infrastructure | Local-First 架构 | [open-design](https://github.com/nexu-io/open-design) |
| plugin-architecture | Skills 系统 | [open-design](https://github.com/nexu-io/open-design) |
| plugin-architecture | 插件生态 | [open-design](https://github.com/nexu-io/open-design) |
| reactivity | Signals 响应式引擎 | [tldraw](https://github.com/tldraw/tldraw) |
| state-management | 响应式记录存储 | [tldraw](https://github.com/tldraw/tldraw) |
| schema-migrations | Schema & 迁移 | [tldraw](https://github.com/tldraw/tldraw) |
| realtime | 多人同步 | [tldraw](https://github.com/tldraw/tldraw) |
| design-systems | Token Pipeline 编排 | [style-dictionary](https://github.com/style-dictionary/style-dictionary) |
| design-systems | 引用解析引擎 | [style-dictionary](https://github.com/style-dictionary/style-dictionary) |
| design-systems | 转换与转换组 | [style-dictionary](https://github.com/style-dictionary/style-dictionary) |
| plugin-architecture | 注册 / 可扩展性 API | [style-dictionary](https://github.com/style-dictionary/style-dictionary) |
| canvas-interaction | 平移与缩放画布 | [xyflow](https://github.com/xyflow/xyflow) |
| canvas-interaction | 节点拖拽 | [xyflow](https://github.com/xyflow/xyflow) |
| canvas-interaction | 缩略图导航 | [xyflow](https://github.com/xyflow/xyflow) |
| graph-editing | 连接手柄 | [xyflow](https://github.com/xyflow/xyflow) |
| graph-editing | 节点调整器 | [xyflow](https://github.com/xyflow/xyflow) |
| graph-rendering | 边路径算法 | [xyflow](https://github.com/xyflow/xyflow) |
| state-management | 响应式 Store 架构 | [xyflow](https://github.com/xyflow/xyflow) |
| rendering | 手绘渲染 | [excalidraw](https://github.com/excalidraw/excalidraw) |
| realtime-collab | 端到端加密协作 | [excalidraw](https://github.com/excalidraw/excalidraw) |
| realtime-collab | 场景协调 | [excalidraw](https://github.com/excalidraw/excalidraw) |
| data-structures | 分数索引 (z-order) | [excalidraw](https://github.com/excalidraw/excalidraw) |
| code-generation | 访谈驱动的应用脚手架 | [whatsapp-agentkit](https://github.com/Hainrixz/whatsapp-agentkit) |
| messaging | WhatsApp 提供商适配层 | [whatsapp-agentkit](https://github.com/Hainrixz/whatsapp-agentkit) |
| agent-architecture | 按联系人的对话记忆 | [whatsapp-agentkit](https://github.com/Hainrixz/whatsapp-agentkit) |
| lead-scoring | 基于规则的潜客评分 | [auto-crm](https://github.com/Hainrixz/auto-crm) |
| ai-integration | AI 潜客分类 (Claude) | [auto-crm](https://github.com/Hainrixz/auto-crm) |
| lead-ingestion | Webhook 潜客获取 | [auto-crm](https://github.com/Hainrixz/auto-crm) |
| agent-architecture | MCP CRM Server | [auto-crm](https://github.com/Hainrixz/auto-crm) |
| canvas-interaction | 看板 Pipeline (拖拽) | [auto-crm](https://github.com/Hainrixz/auto-crm) |
| analytics | CRM 仪表盘 KPI | [auto-crm](https://github.com/Hainrixz/auto-crm) |
| data-portability | CSV 导入 / 导出 | [auto-crm](https://github.com/Hainrixz/auto-crm) |
| activity-tracking | 活动追踪与跟进桶 | [auto-crm](https://github.com/Hainrixz/auto-crm) |
| agent-distribution | 自定义 CRM (Claude 命令) | [auto-crm](https://github.com/Hainrixz/auto-crm) |
| content-synthesis | AI 轮播图生成 | [carousel-generator](https://github.com/FranciscoMoretti/carousel-generator) |
| rendering | DOM 转 PDF 轮播图导出 | [carousel-generator](https://github.com/FranciscoMoretti/carousel-generator) |
| design-systems | OKLCH 主题调色板 | [carousel-generator](https://github.com/FranciscoMoretti/carousel-generator) |
| data-portability | Zod 表单持久化与 JSON 可移植性 | [carousel-generator](https://github.com/FranciscoMoretti/carousel-generator) |
| infrastructure | BYOK + 限速 AI 操作 | [carousel-generator](https://github.com/FranciscoMoretti/carousel-generator) |
| pipeline-orchestration | Graph 执行引擎 | [scrapegraph-ai](https://github.com/ScrapeGraphAI/Scrapegraph-ai) |
| web-extraction | SmartScraper Pipeline | [scrapegraph-ai](https://github.com/ScrapeGraphAI/Scrapegraph-ai) |
| web-extraction | 多源 Fetch 节点 | [scrapegraph-ai](https://github.com/ScrapeGraphAI/Scrapegraph-ai) |
| structured-extraction | Map-Reduce 答案生成 | [scrapegraph-ai](https://github.com/ScrapeGraphAI/Scrapegraph-ai) |
| ai-integration | 与提供商无关的模型层 | [scrapegraph-ai](https://github.com/ScrapeGraphAI/Scrapegraph-ai) |
| research-automation | 搜索驱动的抓取 | [scrapegraph-ai](https://github.com/ScrapeGraphAI/Scrapegraph-ai) |
| state-management | 基于变更的突变模型 | [penpot](https://github.com/penpot/penpot) |
| design-systems | 原生设计 Token | [penpot](https://github.com/penpot/penpot) |
| rendering | WASM/Skia 渲染引擎 | [penpot](https://github.com/penpot/penpot) |
| macos-ui | 刘海形状的常驻顶层窗口 | [boring.notch](https://github.com/TheBoredTeam/boring.notch) |
| media-control | 多提供商媒体控制 | [boring.notch](https://github.com/TheBoredTeam/boring.notch) |
| macos-ui | 系统 HUD 替代 | [boring.notch](https://github.com/TheBoredTeam/boring.notch) |
## 组织方式
```
distill-it/
├── graph/
│ ├── graph.json ← single source of truth (all nodes + edges)
│ └── graph.html ← open in browser for the interactive map
├── features/ ← organized by WHAT it does, not which repo
│ └── /
│ ├── _domain.md
│ ├── study/ ← human-readable, NotebookLM-ready
│ └── build/ ← agent-ready transplant spec
└── repos/
└── .md ← index of everything extracted from one repo
```
功能是根据**它们的作用**来归档的——因此所有“文件检测”的方法都被放在一起,而不管它们来自哪个 repo。
## 添加新的 repo
在 Claude Code 中运行 `/distill https://github.com/owner/repo`。
Claude 会通过网络读取 repo(无需克隆),向你展示功能列表,为你选择的任何功能编写学习 + 构建文档,更新图谱,然后提交并推送。
*使用 [Claude Code](https://claude.ai/code) 和 `repository-intelligence` skill 构建。*
标签:AI辅助编程, Homebrew安装, SOC Prime, 代码分析, 凭证管理, 后端开发, 多模态安全, 开发工具, 文档生成, 防御加固