Skip to content

GET /metrics

Prometheus exposition. The gateway emits a small, stable set of counters and histograms. No authentication is required; the endpoint is public.

Request

No parameters, no body.

curl -fsS https://smart-openrounter.bee1x.one/metrics

The response uses the Prometheus text exposition format and the standard Content-Type: text/plain; version=0.0.4.

Metrics

Name Type Labels Notes
sgw_requests_total Counter model, status One increment per request reaching the gateway. status is 200, 403, 500, etc. model is the real OpenRouter ID.
sgw_routing_decisions_total Counter action, model Routing decision counter. action is ROUTE, SCOUT, STOP_FOR_HUMAN, FALLBACK_TRY, FALLBACK_GLM_INVOKE, FALLBACK_GLM_FAIL.
sgw_request_latency_seconds Histogram — End-to-end request latency.

There is no gauge for open connections, current model health state, or budget spend. Those are reconstructed from logs and from /v1/route/decision evidence when needed.

Example output (abridged)

# HELP sgw_requests_total Total requests
# TYPE sgw_requests_total counter
sgw_requests_total{model="deepseek/deepseek-v4-flash-0731",status="200"} 17
sgw_requests_total{model="deepseek/deepseek-v4-pro-0813",status="200"} 4
sgw_requests_total{model="z-ai/glm-5.2",status="200"} 1
sgw_requests_total{model="smart-router",status="403"} 1

# HELP sgw_routing_decisions_total Routing decisions
# TYPE sgw_routing_decisions_total counter
sgw_routing_decisions_total{action="ROUTE",model="deepseek_v4_flash"} 17
sgw_routing_decisions_total{action="ROUTE",model="deepseek_v4_pro"} 4
sgw_routing_decisions_total{action="SCOUT",model="nemotron_ultra"} 2
sgw_routing_decisions_total{action="FALLBACK_TRY",model="deepseek_v4_pro"} 1
sgw_routing_decisions_total{action="FALLBACK_GLM_INVOKE",model="z-ai/glm-5.2"} 1

# HELP sgw_request_latency_seconds Request latency
# TYPE sgw_request_latency_seconds histogram
sgw_request_latency_seconds_bucket{le="0.5"} 5
sgw_request_latency_seconds_bucket{le="1.0"} 14
sgw_request_latency_seconds_bucket{le="2.5"} 19
sgw_request_latency_seconds_bucket{le="5.0"} 21
sgw_request_latency_seconds_bucket{le="+Inf"} 22
sgw_request_latency_seconds_sum 38.7
sgw_request_latency_seconds_count 22

Status codes

Status When
200 Metrics emitted.

Scrape configuration

A minimal Prometheus scrape config:

scrape_configs:
  - job_name: smart-gateway
    metrics_path: /metrics
    scheme: https
    static_configs:
      - targets: ['smart-openrounter.bee1x.one']

For host-local scraping:

scrape_configs:
  - job_name: smart-gateway-loopback
    metrics_path: /metrics
    scheme: http
    static_configs:
      - targets: ['127.0.0.1:9600']

Suggested alerts

Alert Query Meaning
High error rate rate(sgw_requests_total{status=~"5.."}[5m]) > 0.05 More than 5% of requests are 5xx.
Privacy denials rate(sgw_requests_total{status="403"}[15m]) > 0 Non-zero 403 rate; possibly a misconfigured client.
All-fallback path rate(sgw_routing_decisions_total{action="FALLBACK_GLM_INVOKE"}[5m]) > 0 Senior solver is failing; GLM fallback in use.
Senior unreachable rate(sgw_routing_decisions_total{action="FALLBACK_GLM_FAIL"}[5m]) > 0 Even the fallback is failing.

These are starting points. Tune thresholds to your traffic.