fix(coordinator): the AI summary waits for every job of the run, and the verdict counts reports in words #118

Merged
HumanBean merged 1 commit from HumanBean/Beanstalk:fix/summary-waits into main 2026-11-20 18:12:21 +00:00
Owner

This follows up #117, from what it did on the farm right after deploying.

The summary came too early. On Longshot #1873 it was written as soon as every target had a verdict. Targets past their least reports still had jobs of their range queued, and the comment said "done in about 6 h 45 min". Those later reports would never be in the summary. results_summary.finished now also needs no job queued or running on any target.

"2/1 report" in the folded Measurements verdict now reads "2 reports", or "1 of 2 reports" while the target is short.

Tests:

  • The e2e flow checks the summary isn't asked while the target's second job is queued. Once that job ends, it's asked once.
  • finished is false for a judged target with a job still queued.
  • The verdict's wording is updated in test_timing and test_reports.

Nothing was run on this PC, so CI runs pytest.

🤖 Generated with Claude Code

This follows up #117, from what it did on the farm right after deploying. **The summary came too early.** On Longshot #1873 it was written as soon as every target had a verdict. Targets past their least reports still had jobs of their range queued, and the comment said "done in about 6 h 45 min". Those later reports would never be in the summary. `results_summary.finished` now also needs no job queued or running on any target. **"2/1 report"** in the folded Measurements verdict now reads "2 reports", or "1 of 2 reports" while the target is short. **Tests:** - The e2e flow checks the summary isn't asked while the target's second job is queued. Once that job ends, it's asked once. - `finished` is false for a judged target with a job still queued. - The verdict's wording is updated in `test_timing` and `test_reports`. Nothing was run on this PC, so CI runs pytest. 🤖 Generated with [Claude Code](https://claude.com/claude-code)
fix(coordinator): the AI summary waits for every job of the run, and the verdict counts reports in words
Some checks failed
CI / Conventional Commit title (pull_request) Successful in 17s
CI / Workflow lint (actionlint) (pull_request) Successful in 16s
CI / Coordinator tests (pull_request) Failing after 1m20s
CI / Agent build and checks (pull_request) Successful in 38s
CI / Coordinator image (pull_request) Successful in 11s
dd0fd9ccb0
The summary was written once every target was judged, while targets past
their least reports still had jobs of their range queued (PR #1873: the
summary came with 6 h 45 min of jobs to go), so the later reports were
never in it. It now waits until no job is queued or running.

The Measurements verdict said "2/1 report"; it says "2 reports", or
"1 of 2 reports" while short.

Assisted-by: Claude Code
Co-Authored-By: Claude Opus 5.5 <[email protected]>
HumanBean changed title from WIP: fix(coordinator): the AI summary waits for every job of the run, and the verdict counts reports in words to fix(coordinator): the AI summary waits for every job of the run, and the verdict counts reports in words 2026-11-20 18:11:46 +00:00
HumanBean deleted branch fix/summary-waits 2026-11-20 18:12:22 +00:00
Sign in to join this conversation.
No reviewers
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
MagicBeanDev/Beanstalk!118
No description provided.