Dispatch health and agent list disagree on active engineers, leaving stale starting records at capacity #1274

Merged
coilyco-ops merged 1 commit from issue-1174 into main 2026-07-14 10:50:56 +00:00
Member

Unifies the active-engineer model behind ward agent list, capacity checks, and ward agent dispatch-health. Cleanup-needed prelaunch records and failed-before-start records stay visible for diagnosis, but they no longer count against active capacity.

  • ward agent list now keeps stale starting records visible with status: cleanup-needed and surfaces failed-before-start rows separately.
  • dispatch-health reports cleanup-needed and failed-before-start counts separately and reuses the same active count for capacity.
  • ward agent reap and the launch capacity gate now consume the same active inventory model.
  • Tests: ward exec test, ward exec lint, pre-commit run --all-files.
  • closes #1174
  • ward.workflow: pull-request-and-merge
Unifies the active-engineer model behind `ward agent list`, capacity checks, and `ward agent dispatch-health`. Cleanup-needed prelaunch records and failed-before-start records stay visible for diagnosis, but they no longer count against active capacity. - `ward agent list` now keeps stale starting records visible with `status: cleanup-needed` and surfaces failed-before-start rows separately. - `dispatch-health` reports cleanup-needed and failed-before-start counts separately and reuses the same active count for capacity. - `ward agent reap` and the launch capacity gate now consume the same active inventory model. - Tests: `ward exec test`, `ward exec lint`, `pre-commit run --all-files`. - closes #1174 - ward.workflow: pull-request-and-merge
fix agent inventory convergence
All checks were successful
test / test (pull_request) Successful in 2m2s
e5d829edaf
Sign in to join this conversation.
No description provided.