Description
After completing a Merge Sorted Array interview session, the automatic evaluation failed to generate a report. The interface displayed “No evaluation”, and no scores were produced.
The error states that the session ran end-to-end, but the generated report failed schema validation after two repair attempts.
Steps to Reproduce
Start a merge-sorted-array interview session.
Select C++ and complete the interview.
End the session and wait for the automatic evaluation.
Observe the “No evaluation” message.
These steps describe the affected session; reproducibility has not yet been confirmed.
Expected Behavior
A completed interview session should produce an evaluation report and scores.
If report generation fails, the system should provide a clear recovery option, ideally allowing the report to be retried without repeating the interview.
Actual Behavior
No evaluation report or scores were produced. Report generation failed validation at:
$.improvementPlan[1].successCriterion
The reported validation reason was:
unsupported delivery or personality judgment
Error Message
The automatic evaluation could not be completed: Rust LiveKit runner ended (candidate_ended) but Gemini report generation failed for model gemini-3.1-flash-lite on merge-sorted-array. Final editor state: 413 bytes of cpp. Error: Gemini report failed schema validation after 2 repairs: $.improvementPlan[1].successCriterion: unsupported delivery or personality judgment. Your session ran end-to-end, but no scores were produced, so nothing here is an assessment of your work. Check the agent logs and GOOGLE_API_KEY, then try again.
Impact
The candidate completed the interview but could not receive feedback or scores.
Suggested Investigation
Inspect the generated successCriterion and the two repair attempts to determine why the content remained invalid.
Review the report-generation and repair instructions to ensure success criteria describe observable, task-related improvements.
Consider supporting report-only retries using the completed session data.
Make the recovery message specific to the failure. This error reports schema validation failure; it does not establish that GOOGLE_API_KEY is missing or invalid.
Description
After completing a Merge Sorted Array interview session, the automatic evaluation failed to generate a report. The interface displayed “No evaluation”, and no scores were produced.
The error states that the session ran end-to-end, but the generated report failed schema validation after two repair attempts.
Steps to Reproduce
Start a merge-sorted-array interview session.
Select C++ and complete the interview.
End the session and wait for the automatic evaluation.
Observe the “No evaluation” message.
These steps describe the affected session; reproducibility has not yet been confirmed.
Expected Behavior
A completed interview session should produce an evaluation report and scores.
If report generation fails, the system should provide a clear recovery option, ideally allowing the report to be retried without repeating the interview.
Actual Behavior
No evaluation report or scores were produced. Report generation failed validation at:
$.improvementPlan[1].successCriterion
The reported validation reason was:
unsupported delivery or personality judgment
Error Message
The automatic evaluation could not be completed: Rust LiveKit runner ended (candidate_ended) but Gemini report generation failed for model gemini-3.1-flash-lite on merge-sorted-array. Final editor state: 413 bytes of cpp. Error: Gemini report failed schema validation after 2 repairs: $.improvementPlan[1].successCriterion: unsupported delivery or personality judgment. Your session ran end-to-end, but no scores were produced, so nothing here is an assessment of your work. Check the agent logs and GOOGLE_API_KEY, then try again.
Impact
The candidate completed the interview but could not receive feedback or scores.
Suggested Investigation
Inspect the generated successCriterion and the two repair attempts to determine why the content remained invalid.
Review the report-generation and repair instructions to ensure success criteria describe observable, task-related improvements.
Consider supporting report-only retries using the completed session data.
Make the recovery message specific to the failure. This error reports schema validation failure; it does not establish that GOOGLE_API_KEY is missing or invalid.