Operations

Monitoring

Use health checks, metrics, and traces to keep Orbit predictable in production.

Health check

GET/v1/health

Use as liveness probe. Keep it cheap and frequent.

Prometheus metrics

GET /v1/metricsPrometheus format output

Prometheus UIhttp://localhost:9090

Status endpoint

GET/v1/status

Returns usage, storage footprint, and quota counters for the caller.

Rate limit and replay headers

X-RateLimit-LimitMaximum requests allowed

X-RateLimit-RemainingRequests remaining in current window

X-RateLimit-ResetWindow reset timestamp

Retry-AfterRetry delay in seconds (on 429)

X-Idempotency-ReplayedWhether response came from idempotency replay

OpenTelemetry

ORBIT_OTEL_SERVICE_NAMELogical service name

ORBIT_OTEL_EXPORTER_ENDPOINTOTLP exporter destination

Practical alerting starter pack

Alert on sustained 5xx rate, p95 latency regressions, and spikes in 401/429 responses. If your memory API is noisy, your chatbot quality will be noisy too.

Troubleshooting ->