Observability

How to watch Smart Router in production and diagnose what it's doing. Four pillars:

Topic	What's in it
Metrics	Every Prometheus metric exposed, with types and labels — plus PromQL recipes and suggested alerts.
Traces	OpenTelemetry spans over the full relay lifecycle, configured via standard OTel env vars.
Profiling	Continuous CPU / memory / mutex / block profiles to Pyroscope (or one-off pprof).
Error codes	The internal error taxonomy seen in logs and the smartrouter_errors_total metric, with retryability.

The prebuilt dashboard overlay graphs the metric series out of the box. Most signals are off until you point the router at a collector / Pyroscope / Prometheus — each page below covers how to turn it on.