Files
pipelines/app/notes/runs-tab-status-indicators.md
T
alan fa95c2fcae Remove the Runs error counter for now
Keep the running indicator and Needs you flow, but remove the failed-count badge and its tab count plumbing until unseen-state behavior is addressed. Preserve the unfinished seen-state work separately. Verify application, Runs, attention and shell regressions.
2026-11-11 23:18:30 +00:00

2.1 KiB
Raw Blame History

Independent Runs tab indicators

Current behavior

The user requested temporary removal of the error counter. The failed badge, its tab label/tooltip count and its tab-publication plumbing are removed. The running indicator stays. The unfinished seen-state change was preserved outside the checkout rather than included in this release.

Regression checks for this removal: 734 application, Runs, attention and shell tests passed; analysis was clean. The earlier dual-counter description below is historical and no longer describes the visible tab.

Earlier dual-counter implementation (superseded)

The Runs tab now shows the executing count at the lower edge in the primary theme colour and the failed count at the upper edge in the error colour. Neither suppresses the other. Zero hides its indicator; the failed label caps at 99+ while the tooltip and accessible label retain the actual count.

Counts use known, unfiltered server summaries and update from existing pushed frames without extra polling. Failed status is independent of notification dismissals and the daily attention window. Terminal statuses override a stale execution flag. Older summaries without that flag can still name running work. Waiting decisions remain announced by the attention label and available in the Needs you bell; they are not mislabelled as failed runs.

The app-level observer now republishes status changes even when total attention is unchanged (for example a waiting run becoming failed). Standalone Runs lists also publish both counts through the existing shell registry.

Focused validation: 24 tests passed, covering both colours/positions in light and dark at narrow/wide widths, zero states, live updates from another tab, unchanged-attention transitions, and terminal/older status summaries. Analysis was clean. Full-suite validation then passed: 3,230 tests, zero failures.

Inspected real-browser captures at 1440×1000 light and 560×1000 dark with running=1&chats=2: both counts are distinct, fit their tab, and remain visible after switching to Chat. This is fake-gateway UI evidence, not live-provider or on-device verification.