pip install legacy-doc-ai
One command. Works offline. Configure your LLM provider with a single env var.
Batch-generate docstrings, detect stale documentation, and generate whole-repo READMEs — without sending your full codebase to an LLM.
Free for 1 run. $49/repo/month. Bring your own LLM key.
60-70% of codebase documentation becomes partially or fully inaccurate within 6 months of a significant refactor*
*Published developer productivity research, 2024
Generate docstrings for all undocumented functions in a repo with one command. Supports Python and JavaScript/TypeScript.
Finds functions where docstrings no longer match the actual signature — missing params, wrong types, outdated descriptions.
Preview every proposed change with unified diffs before writing anything. The same review workflow you already use for code changes.
Generates a complete README from actual code analysis — not directory listings. Architecture overviews, API tables, docstring coverage.
Bring your own API key — OpenAI, Anthropic, or any OpenAI-compatible endpoint. Your code never touches a server we control.
AST parsing runs locally. Only function signatures and metadata are sent to the LLM — not your code. Works on air-gapped repos.
One command. Works offline. Configure your LLM provider with a single env var.
legacy-doc-ai check ./src/ — see coverage stats, stale docs, and missing docs in seconds.
legacy-doc-ai docstrings ./src/ --dry-run — see every proposed change as a unified diff.
Run without --dry-run and commit. The same workflow you use for PRs, applied to documentation.
Per-repo. No seat licenses. No usage tiers.