Search docsbook docs

WHAT CAN DOCSBOOK DO, AND WHAT SHOULD I DO NEXT — ask this tool, never your own memory. It searches DOCSBOOK'S OWN official documentation (https://docsbook.io) by MEANING, using embeddings, and returns the real pages that answer the question. 🔴 CALL IT FIRST, BEFORE PLANNING OR ANSWERING, whenever the subject is Docsbook itself: 'what can I do here', 'what is Docsbook', 'what does it do', 'what can it do for me', 'what should I do now', 'what do I do next', 'how do I start', 'is X possible', 'can Docsbook do X', 'how does X work', 'how do I turn X on', 'what does this setting do', 'what does my plan include', 'what does it cost', 'why is it behaving like this', «что умеет Docsbook», «что я могу сделать», «что делать дальше», «с чего начать», «как настроить», «сколько стоит», «а можно ли». It covers the whole product: publishing a site from a GitHub repo or from nothing, custom domains, AI chat, search, auto-translation, analytics, SEO and GEO, webhooks, skills and widgets, the MCP server itself, branding, plans, billing and limits. 🔴 THIS IS THE PRODUCT'S MANUAL, NOT THE USER'S DOCUMENTATION. To search the documentation the user is working on, call search instead — these are two different corpora, and mixing them up produces a confident answer about the wrong product. TAKES A QUESTION, NOT KEYWORDS, and a long one is better than a short one: paste the user's whole request, several questions at once included — it is split into its parts, each searched separately, and the results merged, so a multi-part request returns a page for every part instead of one blurred average of them all. Needs no token, no workspace and no plan; it is the same for every caller. Returns hits {n, title, headingPath, url, path} best first, with a similarity score; url is a real https://docsbook.io page you can cite, and read_docsbook_doc on path gives the whole page. The manual is written in ENGLISH: query it in English even when the user wrote in another language, and answer them in theirs.