Content and answers

Import your docs

Paste a URL when you create the assistant, or later under Content → Import site. Almanzor reads up to 60 pages per import:

• It starts from the URL you give and follows links on the same domain, plus its sitemap.xml if there is one.
• Pages under the same path come first: paste docs.acme.dev/guides to import the guides before anything else on that domain.
• Each page keeps its own URL, so answers can link the exact page.

Pages are read as HTML. If your docs only render in the browser with JavaScript, upload them as Markdown instead.

Keep it up to date

An import is a snapshot. When your docs change, import the URL again from Content and remove the old source. You can also do it from your CI with the API after each docs deploy.

Upload files

Markdown, plain text or PDF, up to 10 MB each, under Content. Useful for content that isn't on a public page: an internal FAQ, a changelog, an OpenAPI guide. Answers from files cite the file name. How many sources and how much space you have depends on the plan.

How the assistant answers

• Only from your content. It searches your sources for each question and answers from what it finds, not from what the model knows about the internet.
• With the page. Answers link the pages they come from.
• It says when it doesn't know. If the answer isn't in your content, it says so instead of making something up.
• In the language of the question. Your docs can be in English and a question in Spanish gets a Spanish answer.

The instructions under Behavior set its tone and scope. New assistants start with a template for technical docs; change it if you want, for example, shorter answers or a different audience.

Unanswered questions

Questions the assistant couldn't answer from your content are collected under Content → Unanswered questions (Starter and up). They are the pages your docs are missing: write them, import again, and the next person gets an answer.