Skip to main content
The knowledge base is the bot’s single source of truth. It never answers from general internet knowledge — only from the sources you’ve added to the workspace. Better knowledge means better answers.

​
Source types

Ingestion is charged per source, not per page — a 200-page documentation site still costs 15 credits. See Credits.

​
Processing statuses

Every source moves through these statuses after you add it:
  • Pending — queued for processing.
  • Processing — being read and indexed.
  • Ready — the bot can answer from it.
  • Error — something went wrong; the error message is shown on the source. Fix the cause (e.g. an unreachable URL) and re-add or reindex it.

​
Limits

How many knowledge sources a workspace can hold depends on the plan: Only sources you add yourself count toward the limit — files, web pages, and documentation sites. Learned Q&A pairs from auto-learning are unlimited on every plan.

​
Organizing knowledge

  • Tags — label sources to keep a large knowledge base navigable.
  • Delete — removing a source immediately removes it from the bot’s answers.
  • For crawled documentation sites, the dashboard shows the tree of indexed pages so you can see exactly what the bot has read.
Write knowledge the way you’d answer a member: short, factual, one topic per document. Marketing copy and long unstructured pages produce weaker answers than a clean FAQ.