Floom batch-b · full dashboard page tour

Editorial hero ‘Work done’ (3 runs/7d), 4 stat tiles with sparklines, Worker activity + Coming-up columns, Emily panel right. Real data, renders clean.

1 worker (NovaSearch V5, ‘needs attention’). Filter chips + grid/list toggle + Add. Renders clean.

Opens as a right-side detail panel (not a full route). Tabs: Overview / History 10 / Source / Versions / Config. Latest output + WHAT-IT-DOES flow diagram. Renders clean.

Source tab (Config is a sibling tab). File tabs worker.yml / SKILL.md / run.py with the real worker.yml YAML. Renders clean.

22 runs (12 failed, 1 running). Card grid with status pills (Completed/Running/Failed). Renders clean.

Completed run. Tabs Output / Trace / Inputs / Raw. RESULT JSON (real shortlist: Rebekka Katzur 72, full stats) in a scrollable pane. Renders clean.

No dedicated ‘Logs’ tab — logs live under the Trace tab. Duration 5m18s, Tokens —, STEPS ‘No steps recorded’, real LOGS stream (e2b sandbox spawn etc.). Renders clean.

The ‘Run’ button opens a manual-run input modal (Job description / Job title / Mode / Sourcing / Operation). This is the actual run-input flow. NOTE: there is NO separate two-pane /app/run/ route — see #8b.

FINDING: the full-page two-pane run route /app/run/ returns 404 ‘Page not found’ on batch-b. The run experience is the modal in #8, not a standalone page.

Library is labelled ‘Brain’. Empty state: 0 folders / 0 files, ‘No folders yet’. Renders clean (empty).

FINDING: page shows ‘Something went wrong — Could not load approvals’, header reads ‘0 pending’, but the sidebar badge says ‘Approvals 1’. Retry does not recover. Count mismatch + data load failure.

37 total / 37 active / 0 error. Secret + app cards grid (AI_MATCH_*, APIFY_API_KEY, AWS_* etc.). Renders clean.