Sphinx

Yes. Kelu crawls Sphinx documentation hosted on readthedocs.io and on Read the Docs custom domains. Sitemap discovery works for all Read the Docs deployments, including projects using the Read the Docs Sphinx theme or custom themes.

Yes. Sphinx autodoc output is rendered as standard HTML pages. Kelu crawls and indexes these pages, so users can ask questions about specific Python classes, methods, and module-level functions and receive cited answers from the API reference.

Add the Kelu widget script to html_js_files in conf.py (Sphinx 1.8 or later). Sphinx adds it to every page, so the widget will appear across all pages of your Sphinx documentation site.

Yes. Create a separate source for each Sphinx doc version and give each a Version label (v1.x, v2.x, latest). Kelu passes the label to the model with every excerpt, so it can say which version an answer applies to. To answer only from one version, scope that version's widget to its own source group, on the Enterprise plan.

For large API references with hundreds of pages, the first index typically completes within a few minutes. Subsequent incremental syncs are faster because Kelu only re-processes pages whose content has changed since the previous crawl.