An Antora extension and supplemental UI that lets readers ask AI for help from the docs and triage gaps (e.g. Cloudflare Proxy vs generic DNS). It adds section-aware in-page prompts, vector search over the Antora repo, and a cached Q&A store that can be promoted into official content.
-
In-page AI prompt: Click between text sections on a page; an AI prompt is inserted so you can ask for help where the instructions lack detail.
-
Vector search: Before or instead of calling AI, a quick vector search runs over existing articles and the AI response store so "frequent flyer" questions can be answered from cache or existing docs.
-
Cached answers: Questions are indexed and answers cached. After review, maintainers can move cached Q&A into an official section; AI suggests where to store them.
Works alongside existing Antora search (e.g. @antora/lunr-extension). That extension does full-text search; this one adds semantic (vector) search and the AI help + cache flow. You can use both in the same playbook.
-
Add this repo (or the published package) to your playbook.
-
Configure
antora.extensionsandui.supplemental_filesas documented in the spec. -
Configure AI response store path and (optional) embedding/search backend.
-
Node 18+
-
pnpm for install and scripts
pnpm install
pnpm run build # when implementedFormat: pnpm exec dprint fmt