Skip to main content

Provider Health Monitoring

The Provider Health page shows you the real-time status of every connected provider. If something is wrong with a provider, you will see it here before your users do.

What You See

📷 [Screenshot: Provider health page with 13 provider rows, each showing status badge, error rate bar, and latency]
For each provider:
FieldDescription
StatusGreen (healthy), Yellow (degraded), Red (down)
Error ratePercentage of requests that failed in the last 5 minutes
P50 latencyMedian response time in milliseconds
P95 latency95th percentile response time, shows worst-case performance
Last checkedWhen Routor last validated this provider

Status Definitions

StatusMeaning
HealthySuccess rate 95% or higher, and P95 latency at or below 5 seconds
DegradedSuccess rate between 70% and 95%, or P95 latency above 5 seconds - still in the routing pool but fallbacks are more likely
DownSuccess rate below 70%, or provider failed startup validation - removed from routing pool
All of the above is measured over a rolling 5-minute window, so status reflects recent behavior, not a long-term average.

Startup Validation

When Routor restarts, it pings every configured provider using their free /models endpoint. Providers that fail this check are marked as down immediately and removed from the routing pool before any traffic is accepted. This is why a misconfigured API key surfaces as a provider being down on this page, not as an error on your users’ requests.

What Happens When a Provider Goes Down

Routor does not wait for you to notice. When a provider’s error rate crosses the threshold, it starts routing around that provider automatically using the fallback chain. You will see:
  • The provider status badge turn red on this page
  • Fallback activity in the dashboard’s request logs (and X-Routor-Fallback: true on responses, if your deployment has DEBUG_ROUTING=1 set)
See Fallback and Auto-Failover for the full explanation.