Point a source at an OpenAPI document and duxt builds a reference out of it — an
overview, a page per tag, a page per operation — with a client on each operation
page that sends the request from the reader's own browser.

The pages are an ordinary collection, which is the whole point: search finds
them, `llms.txt` lists them, the sitemap carries them, the version switcher moves
between them. Nothing here is a second rendering path.

::page-cards
::