Now liveThe Skillselion MCP - thousands of ranked skills, loaded into your agent mid-task. No install.Get it →
bitqs avatar

Maia

  • 1 repo stars
  • Updated May 29, 2026
  • bitqs/maia

Turn subjects or document corpora into bilingual, guided knowledge maps exported as self-contained HTML.

About

Maia builds teaching-oriented knowledge graphs from a topic or your markdown, PDF, and text files, then exports a fullscreen HTML map with guided tours. Use via Claude Code /maia:build when you need structured learning paths instead of flat summaries.

  • Six-stage bilingual graph pipeline
  • Self-contained HTML export with guided tours

Maia by the numbers

  • Data as of Jul 9, 2026 (Skillselion catalog sync)
/plugin marketplace add bitqs/maia
/plugin install maia@maia

Add your badge

Show developers this skill is listed on Skillselion. Paste this into your README.

Listed on Skillselion
repo stars1
Last updatedMay 29, 2026
Repositorybitqs/maia

What it does

Turn subjects or document corpora into bilingual, guided knowledge maps exported as self-contained HTML.

README.md

Maia

English · 中文


Turn any knowledge into a map that teaches you.

Why · What makes it different · Gallery · Quickstart · How it works

Maia is named for the midwife-goddess behind Socrates' maieutics — the art of drawing understanding out of someone rather than pouring it in. Maia doesn't hand you a pile of nodes; it builds a map and leads you through until the understanding is yours.

Why

You want to understand something — Stoicism, causal inference, the Daodejing, a stack of papers you've been meaning to read. The usual options fall short:

  • Search gives you ten tabs and no shape.
  • A summary flattens a system into a paragraph and loses every connection.
  • A raw knowledge graph gives you a pretty hairball you can't actually learn from.

Maia gives you the thing in between: a map that teaches. Concepts arranged by importance, readable in depth, walkable in the right order — from bedrock ideas up to the ones that depend on them.

What makes it different

Most "knowledge graph" tools stop at impressive. Maia is built to teach:

🧭 Graphs that teach, not impress Every node carries a four-part archive — what it is, why it matters, its place in the system, a key quotation. A built-in guided tour walks concepts from foundations upward, like a patient teacher.
🌏 Two languages, two worlds Following Wittgenstein — "the limits of my language are the limits of my world" — English and Chinese graphs are extracted independently, then aligned. Switching language switches worldview, not just labels.
🎨 Taste, applied per subject A Daodejing map and a quantum-mechanics map should not look the same. Maia reads the subject's character and designs a visual identity to fit it.
📄 Feeds on your own documents Point Maia at your own .md .txt .pdf .docx files. Every concept traces back to the exact passage it came from; ideas that never share a page get connected across your library.
🪶 One file, double-click to open Export as a single self-contained .html — no server, no build, no network. Fullscreen-adaptive for presentation. Share like a PDF.

Gallery

Each graph is fully bilingual — the EN / 中 toggle switches labels, inspector text, tour narration, and captions. Two languages, two world-views, one graph.

English

Laozi — Dao De Jing

Dao De Jing knowledge graph (EN)

Diamond Sutra

Diamond Sutra knowledge graph (EN)
中文

老子《道德经》

道德经知识图谱(中文)

金刚经

金刚经知识图谱(中文)

Live demos — open in browser, no download: Laozi — Dao De Jing · Diamond Sutra · Shot Editing

F to present · ← → to walk the tour · EN / 中 to switch worlds

Quickstart

As a Claude Code plugin (recommended — uses your Claude session, no extra API key):

/plugins install github:bitqs/maia

Then in any Claude Code session:

/maia:build 道德经
/maia:build Stoicism
/maia:build --corpus ./my-notes/ --title "My Reading"

Standalone (Python, bring your own graph JSON):

git clone https://github.com/bitqs/maia && cd maia/maia-plugin/backend
python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
python export_html.py ../output/graph.json out.html && open out.html

How it works

  subject  ·  or  ·  your documents
              │
              ▼
  ┌────────────────────────────────────┐
  │  1  seed      find pillar concepts  │
  │  2  expand    per-language worlds;  │
  │               peripheral ideas in  │
  │               isolated sub-agents  │
  │  3  resolve   merge synonyms EN↔ZH │
  │  4  enrich    difficulty · tour    │
  │  5  profile   four-part archive    │
  │  6  theme     visual identity      │
  └────────────────────────────────────┘
              │
              ▼
       graph.json ──► self-contained .html
  • Relevance-routed sub-agents. Peripheral concepts are explored in isolated contexts so they can't pollute the main graph — probabilistic routing, not a hard cutoff.
  • Structured outputs. Pydantic schemas throughout — no hand-parsed JSON, no regex.
  • Two models, by job. Fast model for extraction; stronger model for resolution and synthesis.

Project layout

maia-plugin/
  backend/
    build.py          web mode — research a subject
    build_corpus.py   corpus mode — read your documents
    pipeline.py       six-stage pipeline
    export_html.py    single-file HTML export
    schemas.py        Pydantic structured-output models
    graph.py          bilingual graph contract
    evaluation/       precision/recall harness + gold set
  skills/
    build/            /maia:build Claude Code skill
examples_*.json       worked graphs (Daodejing, Stoicism)

License

MIT — do anything, keep the notice.


把任何知识,变成一张会教你的地图。

把任何知识体系变成一张会自我讲解的活地图。

为什么 · 有何不同 · 成品展示 · 快速开始 · 原理

Maia 得名于苏格拉底"助产术"(maieutics)背后的助产女神——把理解从人内部引出,而不是灌进去。Maia 不是把一堆节点扔给你,而是建一张地图,带你走完,直到理解变成你自己的

为什么

你想弄懂某件事——斯多葛主义、因果推断、道德经、一叠一直没时间读的论文。常见选择都有缺陷:

  • 搜索 给你十个标签页,没有结构。
  • 摘要 把一个体系压成一段话,所有连接都消失了。
  • 原始知识图谱 看起来很酷,但根本学不进去。

Maia 给你中间那个东西:一张能教会你的地图。概念按重要度排列,可以深读,可以按正确顺序走——从基础往上,一直到依赖它们的那些。

有何不同

大多数"知识图谱"工具止步于好看。Maia 的目标是教会

🧭 教会,不只是展示 每个节点有四部分档案:是什么、为何重要、在体系中的位置、一句关键引言。内置引导游览,从基础概念向上展开,像一位有耐心的老师。
🌏 两种语言,两个世界 沿维特根斯坦——"语言的界限就是世界的界限"——中英两张图独立提取再对齐。切换语言切换的是世界观,不只是标签。
🎨 因材施形 道德经的图和量子力学的图不该长一个样。Maia 读取主题的气质,设计匹配的视觉风格
📄 读你自己的文档 把自己的 .md .txt .pdf .docx 喂给 Maia。每个概念溯源到原文段落;从不同文档里提取的想法跨文件连接。
🪶 一个文件,双击打开 导出为单一自包含 .html——无需服务器、无需构建、不需网络。全屏自适应,能演示。像 PDF 一样分享。

成品展示

每张图完全双语——顶部 EN / 中 切换标签、详情面板文字、引导词、图注全部切换。两种语言,两种世界观,同一张图。

English

Laozi — Dao De Jing

Dao De Jing knowledge graph (EN)

Diamond Sutra

Diamond Sutra knowledge graph (EN)
中文

老子《道德经》

道德经知识图谱(中文)

金刚经

金刚经知识图谱(中文)

在线 Demo — 浏览器直接打开,无需下载: 老子《道德经》 · 金刚经 · 镜头剪辑

F 演示 · ← → 步进引导 · EN / 中 切换世界

快速开始

作为 Claude Code 插件(推荐——使用你的 Claude 对话,无需额外 API Key):

/plugins install github:bitqs/maia

在任意 Claude Code 会话中:

/maia:build 道德经
/maia:build Stoicism
/maia:build --corpus ./my-notes/ --title "我的笔记"

独立使用(Python,自带 graph JSON):

git clone https://github.com/bitqs/maia && cd maia/maia-plugin/backend
python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
python export_html.py ../output/graph.json out.html && open out.html

原理

  主题  ·  或  ·  你的文档
              │
              ▼
  ┌────────────────────────────────────┐
  │  1  播种      找核心概念              │
  │  2  展开      中英独立扩展;           │
  │               低相关概念隔离子智能体  │
  │  3  对齐      合并中英同义概念        │
  │  4  丰富      难度·引导顺序          │
  │  5  建档      四部分档案             │
  │  6  定形      匹配主题的视觉风格      │
  └────────────────────────────────────┘
              │
              ▼
       graph.json ──► 单文件 .html
  • 相关度路由子智能体。 边缘概念在隔离上下文中探索,不污染主图——概率路由,非硬截断。
  • 结构化输出。 全程 Pydantic schema,无手工解析 JSON,无正则。
  • 两种模型,各司其职。 快速模型做高并发提取;强模型做对齐与综合。

项目结构

maia-plugin/
  backend/
    build.py          网络模式——研究一个主题
    build_corpus.py   语料模式——读你的文档
    pipeline.py       六阶段流水线
    export_html.py    单文件 HTML 导出
    schemas.py        Pydantic 结构化输出模型
    graph.py          双语图谱契约
    evaluation/       精确率/召回率评测框架
  skills/
    build/            /maia:build Claude Code 技能
examples_*.json       示例图谱(道德经、斯多葛主义)

许可证

MIT — 随意使用,保留声明即可。

Built to be read, walked, and understood — not just looked at.

Related skills

AI & Agent Buildingresearchautomation

This week in AI coding

Five minutes, every Monday - the tools, releases and tactics for developers.

unsubscribe anytime.