Recreated from a production system-health page: live dependency probes (DB, storage, cache) with latency, config-presence integration checks, an overall status only live failures can drop, and an incident log with a 60-minute per-source alert throttle so a flapping dependency can't spam on-call. Hit 'Simulate incident' to trip a fault.
System operational
3 live probes · 4 integrations
Database (Postgres)
Connected
Object storage
Connected
Cache (Redis)
Connected
OpenAI (LLM + images)
Key configured
ElevenLabs (voice)
Key configured
OpenRouter (LLM + video)
Key configured
Resend (email)
Key configured
No incidents. Hit "Simulate incident" a few times — every third one trips a live fault.
Recreated from the system-health page shipped in a production field-service platform. The key discipline: monitoring must never take down what it monitors, and config presence is not health — a missing API key is a feature toggle, not an outage. Hit 'Simulate incident' a few times to trip a live fault and watch the throttled alert log.
