Share of all missing jobs by root cause — click a band to focus
Where in the pipeline jobs are lost
ATS / source carrying the gap
Which list(s) a company came from (its CSV source column) — click to filter
| # | Company | Live | HC | DB | Missing | Gap % | Category | System | Boards / tokens | Last checked |
|---|
Spawns the full codex batch — exactly like a no-parameter command-line run, with optional overrides
Launching a batch needs the live dashboard server. Start it with node google-sheet-finder/runBoardInventory.js, then open http://localhost:8787/ and use this tab.
Live console output from the running batch
The live run log is only available through the dashboard server (http://localhost:8787/), not when opening this file directly.
| Row | Company | Result | Live → HC | Tokens | Time | Log |
|---|---|---|---|---|---|---|
| No companies processed yet — start a run from the Run tab. | ||||||
Codex ranks the open gaps by missing jobs recovered per unit of fix effort — quickest, highest-yield fixes first (a ratio, not raw job count)
The priority analysis runs codex on the server. Start it with node google-sheet-finder/runBoardInventory.js, open http://localhost:8787/, then use this tab.
For every implemented ATS, sample 5 live jobs and confirm the page matches Firestore (title & description via embedding similarity, plus posted date). Runs the complete sweep once; after that, re-check a single ATS with its ↻ button (others are kept). “Run checks” re-runs everything.
Vertical quality checks run on the server (Playwright + minimal codex) — available only when the hub is served, not on a static file:// page.
Permanently bans this company — it disappears from the dashboard, and the reason below is given to codex any time it is re-run, so it stays banned.
That's a lot of accumulated gap data to throw away in one click — there's no good reason to bulk-delete this much from the dashboard. If you truly mean it, delete it manually on the box. Otherwise… good luck catching Yes. 🙂
Hidden from this dashboard. Un-ban to bring one back on the next refresh.
Leave everything untouched to run exactly like a plain no-parameter run. Blank number fields fall back to the built-in defaults (shown as placeholders).