Website knowledge now says what it does: a website source has a Re-import button instead of an "Auto-sync" badge that suggested it updated by itself. You can add the pages under any address of your site, exclude whole sections such as a glossary, and imports read up to 300 pages, share them fairly between the sections of your site and find pages through its sitemap.
How it behaves
- Re-import: an imported website or help center shows Re-import (and "Importing…" while it runs). Its tooltip says when the source was last imported in full. Pages are updated in place; nothing is duplicated. Only products (synced from your store) and macros (synced from your helpdesk) keep the Auto-sync badge, because they really do update on their own.
- Add pages from a URL: on a website source, the page menu's Add pages from a URL imports
the page at an address and the pages under it, for example
https://acme.com/customers, into the same source. Pages Aide already has are updated, new ones are added. The address must be on the same site (acme.comandwww.acme.comcount as one). This partial import doesn't change the source's "last imported" time, which stays the time of the last full import. - Excluded paths: when you import your website, you can list paths to skip, such as
/glossaryor/blog/tag; you can change them later from the website source's menu (Excluded paths). A path covers the page itself and every page under it (/glossaryskips/glossary/seobut not/glossary-of-care). Adding a path removes the pages already imported under it right away, and Aide stops answering from them; no re-import needed. - Removing an excluded path: later imports read that section again, but the pages removed when you excluded it stay removed, like pages you delete by hand.
- Fairer imports: an import takes pages from each section of your site in turn (
/customers,/blog,/glossary…), so a section with hundreds of links, such as a glossary, no longer uses up the page budget before the rest of your site is read. - Larger imports: a website import, including the one from sign-up, now reads up to 300 pages (it was 100). The most is still 500.
- Sitemaps: imports also read your site's sitemap (the one your robots.txt names, or
/sitemap.xml), so pages that nothing links to are found too. A missing or unreadable sitemap is simply skipped. - Links kept: the text Aide reads from a page keeps its links with their full address, so an index page (a list of case studies, for example) points the AI agent to the right subpage. Links in your site's menus and footer are still left out.
- Answers from short pages: when a short page (a list of case studies, a pricing page) matches a question, the AI agent reads the whole page rather than only the part that matched. A page's short opening line, such as an eyebrow above its title, is no longer stored on its own, so it can't be matched without the content under it. Existing sources pick this up at their next import.
Also in this change
- The help article "Add knowledge from your help center or website" covers Re-import, adding pages from a URL, skipped paths and the 300-page limit.
For developers
POST /v1/knowledge-sources/:id/importsaccepts an optionalurl: an http(s) address on the source's site (at most 2,048 characters). The crawl starts there and stays under its path, writing into the same source. Another site answers 422other_siteonurl; an address under an excluded path answers 422excluded.- Knowledge sources have
excludePaths(string[], path prefixes, stored without a trailing slash). It's accepted onPOST /v1/knowledge-sourcesandPATCH /v1/knowledge-sources/:id(up to 50 entries, each starting with/, at most 200 characters, no query or fragment) and returned on every knowledge source. Patching it deletes the source's documents under the newly added paths. - The default
maxPagesfor a new website or help center import is now 300.