Training your chatbot

1. 1

   Enter a URL

   Open Chatbots → Knowledge → Add from website and paste the URL you want indexed. A section root such as https://example.com/help works as well as a homepage.

2. 2

   Review the discovered pages

   OyeChats reads your robots.txt for Sitemap: directives and then the sitemaps themselves. If that turns up nothing usable, it scans the page you gave it for same-domain links and follows them. Either way you get the page list with an estimated credit cost before anything is charged, and your typed URL is always included.

3. 3

   Deselect what you do not want

   Pruning the list here is the cheapest moment to do it. Login pages, cart pages, tag archives and paginated blog indexes usually add cost without adding answers.

4. 4

   Start the crawl

   Progress updates live. Pages are indexed in waves as they arrive, so early pages become answerable before the whole crawl finishes. You can cancel mid-run; pages already indexed are kept.

Use Add files for anything not on your website: spec sheets, policy documents, price lists, internal FAQs.

Before charging, the uploader shows a cost preview. Unsupported extensions and oversized files are rejected at that point, so they never consume credits.

Plans also cap total knowledge size, in pages, files, and characters. Hitting a cap blocks further ingestion for that chatbot until you remove content or move up a tier; it never silently truncates a document. See Plans and limits.

Manual re-crawl. Full

Re-fetches and re-indexes every discovered page, and charges for all of them at 5 credits each even if nothing changed. Available on every plan, and it is the only manual option below Standard.

Manual re-crawl. Updated pages only

Re-fetches everything but only re-indexes and charges for pages whose content actually changed. Included on Standard and above. Before it runs you get a diff showing how many pages are unchanged, new and removed.

Automatic re-crawl

A weekly refresh of every previously-crawled URL. Pages whose content genuinely changed are re-indexed; unchanged pages are skipped. Cosmetic churn such as a copyright year rolling over does not count as a change. Included on Standard and above; the toggle is visible but locked on lower tiers.

Re-index a file

Re-processes a single uploaded document, useful after you upload a corrected version under the same name.

Deleting a page or file from the Knowledge tab removes it and its indexed passages, and the chatbot immediately stops being able to answer from it. Deletion does not refund the credits already spent indexing it.

• Index the pages that answer questions. Docs, help centre, FAQs, pricing, policies. Before marketing pages.
• Skip near-duplicate pages. Ten variants of the same landing page make retrieval less certain, not more thorough.
• Prefer one thorough page over several thin ones on the same topic.
• Watch Analytics → Unanswered questions. It is a direct list of what your knowledge base is missing.