hy,
when i use the trajectory evaluator, this error shows up : error:
Object reference not set to an instance of an object.
even if i use the defaut trajectory evaluator, or a personnalized trajectory,
when i search about it, it seems it can’t return the logs, but in the details of the execution, the response of the agent is there
i can’t understand the issue
please if you have any idea, share it with me, it can help!
This error is usually caused by the Trajectory Evaluator not the agent itself, especially if the agent response is visible in the execution details. Check that the expected output schema matches the actual output(including the correct Output Key), ensure no trajectory fields are null, and update to the latest UiPath Studio and GenAI Activities versions. If it still occurs with the default Trajectory Evaluator, it is likely a product bug, and raising a UiPath support ticket with the execution logs is recommended.
If it’s work for you please mark as solution
Cheers
to add a clarification,
i use the same evaluator for all the evaluation set, some of the evaluation works normaly, others dosen’t work, even if i already test it before and it worked for must of them, and i changed nothing,
i tried to test with the default evaluator, to see if the problem is from my evaluator, but still the same problem no trajectory fields are null
Since the agent execution completes successfully and the response is present in the execution details, the issue is unlikely to be with your custom evaluation. Because the same error occurs with the **default Trajectory Evaluator and the same evaluation sets worked previously without any changes, this points to a backend/product issue rather than a workflow issue. As a workaround, try republishing the evaluation set or creating a new one and rerun the tests. Also ensure you’re using the latest UiPath Studio and GenAI/Evaluation package versions. If the problem persists, collect the execution logs and open a UiPath Support ticket, as this appears to be a Trajectory Evaluator service issue rather than a configuration problem.
Cheers
I tried to re-create the evaluation set and the evaluator (i left it as default), still not stable, it changed the result everytime, i select the error ones, and re-run them, one of them worked, the other not, then i select the last one, and it worked, even if i changed nothing.
the same problem persists,
if i execute one by one, it worked but if i want to execute the full evaluation set, it doesn’t work
I think it’s not stable yet for the trajectory evaluations