Introduction
sticky reads a public GitHub repository and writes a wiki you can actually read. No accounts, no installs, no setup — paste a URL and an AI-written tour of the codebase streams onto the page.
What is sticky for
Most repos are easy to clone and hard to understand. The README usually tells you what the project does, sometimes how to run it, and rarely how it's put together. sticky fills that gap: a short, opinionated walkthrough of a project's moving parts, written from the code itself.
It's aimed at:
- New contributors sizing up an unfamiliar codebase before opening a PR.
- Engineers evaluating a dependency who want a faster signal than the README provides.
- The curious — the kind of person who reads source code on a Sunday.
The 30-second flow
- Paste a GitHub URL on the home page.
- Choose whether to add the result to the public Library.
- A wiki streams in over the next 30–90 seconds: overview, architecture, key modules, getting started, notable details.
- The finished wiki is saved to a stable URL like
/w/repo-name. Refreshing or sharing the link renders the cached version instantly.
Terminology
- Wiki
- The AI-generated document for a single repository. Stored as Markdown in the database and rendered with GitHub-flavored extensions.
- Slug
- The URL path for a wiki. Listed wikis get a human-readable slug (
/w/commander-js); unlisted wikis get an opaque one (/w/u-abc123def). - Listed
- A boolean on each wiki. Listed wikis appear in the public Library and are deduped — two people pasting
facebook/reactshare one wiki. Unlisted ones are unindexed and each submission gets its own copy. - Snapshot
- The slice of the repo sticky actually reads — repo metadata, the full file tree (paths only, capped at 500), and the contents of up to 40 selected files.
Next: How it works — the pipeline from paste to persisted wiki.