Every live AI app on one board: usage, spend, speed, reliability, accuracy and security, with the rules that raise an alert.
One row per live app. Click a row for its drill-down.
| App | Model in production | AI requests | Spend | Slow answer (p95) | Failed requests | Accuracy (latest) | Security | Cheaper setting |
|---|
AI requests from each app's own log line; other requests and scanner probes from the Cloud Run request logs. Scanners are automated probes for files like stored passwords and backups.
From each app's per-request log line. The dashed line is the daily budget; going over it raises an alert.
Fixed and checked every time the board loads. Alerts show on this page only.
The live figures are private. Enter the access code to see them.
Answered; declined (the app said it couldn't answer from its sources); refused at the front door (limits, access code); or failed.