Operations API

Operations endpoints are used for liveness checks, routing visibility, and incident triage.

Audience Guidance

• SRE/ops: integrate these routes into health checks and dashboards.
• Developers: use them when debugging routing/performance behavior.

Core Endpoints

• GET /health for liveness/readiness style checks.
• GET /v1/metrics/providers for rolling provider-level performance/usage stats.

Monitoring Examples

Basic liveness check:

curl -sS -f http://localhost:8317/health

Provider metrics snapshot:

curl -sS http://localhost:8317/v1/metrics/providers | jq

Prometheus-friendly probe command:

curl -sS -o /dev/null -w '%{http_code}\n' http://localhost:8317/health

Suggested Operational Playbook

1. Check /health first.
2. Inspect /v1/metrics/providers for latency/error concentration.
3. Correlate with request logs and model-level failures.
4. Shift traffic (prefix/model/provider) when a provider degrades.

Failure Modes

• Health endpoint flaps: resource saturation or startup race.
• Provider metrics stale/empty: no recent traffic or exporter initialization issues.
• High error ratio on one provider: auth expiry, upstream outage, or rate-limit pressure.

Related Docs

• Routing and Models Reference
• Troubleshooting
• Management API