FAQ

Frequently Asked Questions (FAQ)#

Q: Can I test the agent without deploying it?
A: Yes, absolutely. You can freely test the agent in the development environment before deployment.
We recommend deploying it only after verifying that it functions as intended.

Q: If I edit or delete knowledge, is it immediately applied to the live environment?
A: No. Any changes to knowledge require re-testing and redeployment for the updates to take effect in the live environment.
This ensures stable and consistent service operation.

Q: What should I do if the agent gives an incorrect response during testing?
A: You can use the “👎 Not quite” feedback feature to record unintended responses.
Optionally, you can register an exemplary answer directly or update the related knowledge to retrain the agent.

Q: Can the RAG search model be fine-tuned?
A: Yes. Positive “👍 Good job” feedback is used to collect correct responses, which can later be leveraged for RAG embedding model fine-tuning.
Currently, fine-tuning is available upon request through the Sales Team.

Q: What do the numbers shown in the answer sources mean?
A: Each cited knowledge source in the answer is labeled with a number. Clicking the label opens the referenced knowledge directly.
You can also view the original document (PDF) using the built-in viewer.
For a deeper analysis, detailed call logs can be reviewed in the Detailed Log page.

Q: How are sensitive data and confidential information handled in registered knowledge?
A: The Training Data Quality Management feature automatically scans knowledge for predefined sensitive information patterns.
Detected data will trigger either a “Review Required” alert or automatic learning failure, depending on settings—helping prevent security risks in advance.

Q: Can I check what users actually asked?
A: Yes. The Log Page allows you to review real user queries and corresponding agent responses.
This helps identify missing knowledge areas or frequently asked topics.
You can also use the feedback feature to manage and improve cases that require additional training.