跳到正文
GitHub·· 8 天前AI 评分60

repomap 开源:扫描代码库生成 CONTEXT.md 供 AI 编码智能体读取

connectjackofficial-source/repomap

AI 导读

repomap 是一个开源工具,扫描代码库后生成紧凑的 CONTEXT.md,让 Claude Code、Cursor、Codex 等 AI 编码智能体快速了解项目结构,避免反复读取文件消耗大量 token。

正文

Scan a codebase and generate a compact CONTEXT.md so any AI coding agent understands your project in 30 seconds, not 30 minutes.

Python 3.9+ License: MIT

Every time you point Claude Code / Cursor / Codex at an unfamiliar repo, it burns hundreds of thousands of tokens re-reading files to figure out where the entry points are, what the classes do, and how things connect.

repomap pre-computes that once: it walks the tree, parses Python with ast and JS/TS with heuristics, and writes a tight map of files, classes, functions, and exports. Drop CONTEXT.md at the root and your agent starts oriented.

Install

git clone https://github.com/connectjackofficial-source/repomap.git
cd repomap
python -m repomap.cli /path/to/your/repo

Usage

# generate CONTEXT.md for the current project
python -m repomap.cli .

# write somewhere else
python -m repomap.cli ~/src/awesome-app -o docs/MAP.md

# skip extra directories
python -m repomap.cli . --ignore generated temp

# only scan the top two levels
python -m repomap.cli . --max-depth 2

# machine-readable output for tooling
python -m repomap.cli . --format json -o context.json

# skip files with very few symbols (focus on real modules)
python -m repomap.cli . --min-symbols 3

# per-language file/symbol counts
python -m repomap.cli . --stats
# {"\".py\"": {"files": 12, "symbols": 87}, "\".ts\"": {"files": 4, "symbols": 19}}

Example output

# my-app - codebase map

**`src/server.py`**
- class Server (methods: listen, route, use)
- def create_app(config)

**`src/db.py`**
- class Pool (methods: query, close)
- const DEFAULT_LIMIT

How it works

Language Extraction
Python ast — classes, methods, functions, constants
JS / TS / JSX / TSX regex — exports, classes, top-level functions

Skips .git, node_modules, __pycache__, virtualenvs. No network, no third-party deps.

Use cases

  • Speed up onboarding of new AI coding agents
  • Generate a reference doc for contributors
  • Spot entry points and public exports before a refactor

License

MIT

来源:GitHub · github.com