One command

[xzrderek]

Note

Adds --force to create rft, passes it through to upload, and polls evaluator status until ACTIVE before proceeding, failing with a dashboard link on timeout.

• CLI:
  • Add --force flag to create rft to allow overwriting an existing evaluator.
• RFT creation flow:
  • Pass --force through to upload and log when used; include env_file=None in upload args.
  • Poll evaluator via GET /v1/accounts/.../evaluators/{id} with User-Agent header until ACTIVE (or timeout/BUILD_FAILED).
  • On timeout/failure, print evaluator dashboard URL and exit with non-zero status before creating the job.

^{Written by Cursor Bugbot for commit 68c7a2d. This will update automatically on new commits. Configure here.}

[cursor]

Bug: Evaluator polling errors allow unconfirmed readiness progression

If the evaluator status polling (lines 406-421) raises an exception, it will be caught by the exception handler at line 424, which only prints a warning and allows execution to continue. This means the RFT job creation will proceed even though the evaluator's readiness was never confirmed. The polling logic should either be outside the try-except block, or the exception handler should check whether polling completed successfully before continuing.

 

[cursor]

Bug: Evaluator Upload Failure: Stop With Error Code

When evaluator upload fails (rc != 0), the function prints a warning but continues execution instead of returning an error code. This causes the function to proceed with RFT job creation using a potentially non-existent or failed evaluator, which will likely fail. The code should return 1 (or the actual error code) when upload fails, similar to the error handling at line 421 when the evaluator is not active.