starmap

Organize your GitHub starred repos with AI-assisted categorization, right from the command line.

starmap fetches everything you've starred on GitHub, asks an LLM to group it into a small number of broad, meaningful categories, and lets you review, rename, merge, and move repos between categories before anything is saved. It can also read from and write to GitHub's native Lists feature (github.com/stars), so it works whether you're starting from scratch or already curate lists there.

Nothing is ever deleted or overwritten silently: every write is backed up, and AI suggestions are staged separately until you explicitly approve them.

Features

• Fetches your starred repos via the GitHub API and stores them locally
• Uses an LLM to suggest a small number of broad categories (not a new list for every 3-5 repos) - supports Anthropic, OpenAI, Gemini, Groq, and Ollama
• Interactive terminal UI to approve, reject, rename, merge, and move repos between categories before anything is saved
• Reads and writes GitHub's native Lists feature (github.com/stars) via its GraphQL API, so you can build on lists you already curate there
• Exports your organized stars to Markdown
• Automatic backups before every write; nothing is lost by re-running a command

Installation

pip install starmap-gh

The PyPI package is named starmap-gh (the short name was already taken), but it installs a command called starmap.

To install from source:

git clone https://github.com/AndriiSonsiadlo/starmap
cd starmap
pip install -e .

Quick start

# 1. Fetch your starred repos (requires a GitHub token)
export GITHUB_TOKEN=ghp_yourtoken
starmap sync

# 2. Ask an LLM to suggest categories
starmap suggest --provider anthropic

# 3. Review the suggestions: approve, reject, rename, merge, or move repos
#    between categories. Nothing is written until you save.
starmap organize

# 4. Browse the results
starmap list
starmap list "Machine Learning"
starmap list --uncategorized

# 5. Export to Markdown
starmap export

Commands

Run starmap <command> --help for the full set of options for any command.

AI providers

starmap suggest --provider <name> supports:

The AI prompt is deliberately biased toward a small number of broad, durable categories (e.g. "Web Frameworks", "DevOps & Infra", "Data & Machine Learning") instead of a new category for every handful of repos, so your stars stay easy to browse instead of fragmenting into dozens of near-empty lists. If a very large star list causes a response to come back truncated, lower the batch size with --chunk-size/-c.

How your data is kept safe

• starmap sync never wipes out local edits: if a repo you already have is fetched again, its notes/category fields are carried over.
• starmap suggest never overwrites your approved categories. It writes to a separate staging file, and tells the AI about your existing categories so it reuses them instead of creating near-duplicates. Nothing is final until you review it in starmap organize.
• Every write to your local data files makes a rolling backup first, so you can always recover the previous version.
• starmap push-lists makes no changes to your GitHub account unless you pass --experimental; by default it writes a checklist for you to apply by hand.

starmap organize keyboard reference

Working with GitHub Lists (github.com/stars)

GitHub's Lists feature has no REST API, but it is part of GitHub's official GraphQL schema (UserList, createUserList, updateUserListsForItem, and so on, added in 2023 - see the GraphQL reference). Two commands build on that:

• starmap fetch-lists treats your real GitHub Lists as the source of truth: it fetches every list and its repos and saves them as your approved categories. Run this before suggest/organize if you already curate lists on github.com/stars and want to build on that instead of starting over.
• starmap push-lists pushes your approved categories back to GitHub. By default it writes a checklist you apply by hand and makes no API calls. Pass --experimental to create/update lists automatically via GraphQL. A repo is only added to its target list and removed from other lists starmap already knows about that it no longer belongs to; memberships in lists outside what was fetched are never touched.

Both commands require a classic Personal Access Token with the user scope - broader than the read-only scope sync needs, since Lists have no REST equivalent and live behind GraphQL mutations that require it.

Configuration

starmap stores its data in ~/.starmap/:

Set GITHUB_TOKEN in your environment to avoid passing --token on every command.

Development

git clone https://github.com/AndriiSonsiadlo/starmap
cd starmap
uv sync --all-extras
uv run pytest

License

Released under the MIT License.