Labs / General AI / browser-only monitor
AI Availability Monitor
A small incident-room exercise: refresh provider signals, compare freshness, and decide whether your application should fail over. The demo is safe to run on-site—it uses simulated checks, so it never sends your prompt or credentials anywhere.
No check run yet.
Key takeaway: availability is a decision, not a green badge. An enterprise harness should combine provider status, request error rate, latency, model-specific health, and the age of the last successful check before it fails over.
Try it: select an incident mode, run checks twice, then decide which signal belongs in an SLO and which belongs only on an operator dashboard.