Description
In openagent_eval/cli/commands/run.py at line 146, completed=True is passed to Progress.update(). In Rich's progress bar, True is coerced to 1 in a numeric context, causing the progress bar to drop from 100% to 1/total_items when the evaluation completes.
Location
openagent_eval/cli/commands/run.py, line 146
Current Code
progress.update(task, description="Complete!", completed=True)
Why This Is Wrong
completed=True evaluates to 1 when used numerically
- The task's
total is set to total_items (e.g., 50 items) at line 126
- Before this line, the bar correctly shows
50/50 (100%)
- After this line, the bar drops to
1/50 (2%) — it looks like the progress reset
Expected Fix
# Option A: Advance to completion properly
progress.update(task, description="Complete!", completed=total_items)
# Option B: Use advance + update
progress.update(task, description="Complete!")
Impact
- Visual bug: The progress bar visibly drops from 100% to near-zero right at the end
- This is especially noticeable for large evaluations with many items
- Creates a poor user experience — looks like something went wrong
Severity
High — Visible UI bug that affects all evaluation runs.
Steps to Reproduce
- Create a config file with a dataset of 10+ items
- Run
oaeval run config.yaml
- Observe the progress bar drops to a near-zero percentage when the evaluation completes and the description changes to "Complete!"
Description
In
openagent_eval/cli/commands/run.pyat line 146,completed=Trueis passed toProgress.update(). In Rich's progress bar,Trueis coerced to1in a numeric context, causing the progress bar to drop from 100% to 1/total_items when the evaluation completes.Location
openagent_eval/cli/commands/run.py, line 146Current Code
Why This Is Wrong
completed=Trueevaluates to1when used numericallytotalis set tototal_items(e.g., 50 items) at line 12650/50 (100%)1/50 (2%)— it looks like the progress resetExpected Fix
Impact
Severity
High — Visible UI bug that affects all evaluation runs.
Steps to Reproduce
oaeval run config.yaml