aryrabelo/repodocs

RepoDocs is a command-line tool that automatically generates documentation wikis for software repositories. You point it at any codebase and it produces a self-contained HTML wiki file where every statement includes a citation linking back to the specific file and line of source code it came from. The idea is to give maintainers documentation they can trust without having to write it by hand. The tool works by scanning your codebase deterministically, then driving an external AI coding assistant, such as Claude Code, OMP, or Codex CLI, to plan and write the documentation pages. You need Python 3.10 or higher and a tool called uv installed on your machine. You also need one of the supported AI backends logged in and ready to go. Running a single command kicks off the full pipeline, which produces an offline wiki.html file you can open in any browser. RepoDocs runs entirely on your local machine, which means your code never gets uploaded to a hosted service. This is a key distinction from hosted alternatives like DeepWiki, which require uploading your repository to their servers. The core tool has zero runtime dependencies beyond Python's standard library, though there is an optional diagram rendering feature that requires additional software like Bun and a browser for rendering images. The pipeline offers several individual commands for scanning, planning, generating pages, translating, building HTML, and publishing. Publishing has built-in safety measures: it refuses to publish directly to main branches, scans output for private keys or tokens, and requires explicit confirmation before making anything public. The tool always recommends doing a dry run first to review what will be published, since generated documentation could still expose sensitive details about your source code. For developers who want to work from within their AI coding assistant, RepoDocs is designed to be called as a command-line tool rather than running as a plugin inside the assistant itself. The full README is longer than what was shown.