Every response from the Knowledge Agent is transparent and verifiable. You can see exactly where the information came from and how the AI arrived at its answer.
The Sources panel
After each AI response, open the action menu (the three-dot icon on the message). Under "Sources," you will see a list of the files, tables, and data sources that informed the answer. Click any source to preview it - PDFs open in an inline viewer with the relevant pages highlighted, images display in a preview panel, and other file types download directly.
If more than five sources were used, click "View all" to see the complete list in a dedicated modal.
This matters because business decisions should never rely on unverified information. When the AI tells you that revenue grew 12% last quarter, you can confirm exactly which data tables produced that figure.
The Tracing panel
For users who want deeper technical insight, the Tracing Log provides performance metrics for each response:
Latency - how long the AI took to generate the answer (in seconds)
Token usage - input tokens, output tokens, and total tokens consumed
Requests - number of AI model calls made
Success status - whether the response completed without errors
Access the Tracing Log from the same action menu, under "View Tracing Log." This is primarily useful for administrators monitoring system performance.
When the AI does not have enough data
If the AI cannot find sufficient information to answer your question, it will say so rather than guessing. This typically happens when the relevant data source is not yet connected, or when the question falls outside the scope of your available data. The solution is usually to connect additional integrations or upload relevant documents.