Skip to main content
Knowledge sources provide context to the AI during queries. Documents are indexed and scoped per workspace. The AI can only search knowledge from the workspace it is queried in.

Source types

TypeDescription
confluenceConfluence wiki pages from a specified space
fileUploaded files (PDF, text, markdown)
inlineText content entered directly
urlWeb pages fetched and indexed

Sync lifecycle

Each knowledge source goes through a sync process:
StatusMeaning
pendingSource added but not yet synced
syncingSync in progress (fetching, chunking, indexing)
syncedReady for use. Chunk count is available.
errorSync failed. Check source configuration.

Chunks

Documents are split into searchable chunks during sync. The chunk count is tracked per source. During a query, relevant chunks are retrieved and passed to the language model as context. The number of knowledge chunks used per query is recorded in the audit log.

Knowledge gaps

When the AI cannot answer a question from available knowledge, it is recorded as a knowledge gap. Gaps are tracked per workspace and surfaced in:
  • The Knowledge tab (gaps sub-tab)
  • The workspace dashboard
  • Conversation stats
Each gap includes the topic, a description, the conversation it came from, and the user who asked.