Share
Report
2I am trying to use the relatively new "Evaluation" tab in Copilot Studio to automatically evaluate my agent based on a few test questions.
Besides the fact that there is currently a bug with the automatic csv-file import (although the csv-file has the right format as described in the documentation, it does not correctly import and separate the questions from the answers), I have another problem with the response from the agent. So I have created a test set manually.
My agent uses the "get items"-tool that connects to a SharePoint list and extracts data from it and answers the user query based on this knowledge. When I then start the evaluation with my test questions, I always end up with 0%. That is because the agent did not return any answers. I then had a look into the agent and the flow that should have been executed and found out, that the flow is waiting for a user interaction. More specific, the agent awaits the confirmation of the user to allow the connection to SharePoint.
Is this currently a bug and will be fixed soon or is there a workaround for this problem?
Categories:
I have the same question (0)
All responses (
Answers (
