Capture knowledge
Feed the agent with real sources
The knowledge console accepts files, plain text and crawled URLs. Items are scoped and prepared so they can be retrieved during a chat.
Files and pages
You upload the formats supported by your admin panel (common text and office documents, images, and video files when your deployment allows). A summary or business note can be indexed alongside the asset to steer answers.
URLs can be crawled to create documents that mirror your public site or help centre. Each record carries a title, tags and a scope (support, product experience, etc.) according to your organisation.
What gets indexed
Useful text is segmented and stored for retrieval while answering. After major edits you trigger reindexing from the console so agents quote the latest wording.
Signals from conversations
The product can derive lessons from transcripts when auto-learning is enabled for the application. Those items feed agent memory under governance (review queues may apply) and do not replace your canonical documents.