API reference
GET /healthz
Liveness and readiness probe used by Railway and external uptime monitors.
Public, unauthenticated probe. Returns a compact JSON document and a 2xx status when the worker is ready to accept work. Used by Railway's health-check job and by our internal uptime monitor.
Endpoint
GET https://api.prom.codes/healthz
No headers required.
Response
{
"ok": true,
"service": "prometheus-indexer-worker",
"version": "0.4.2",
"commit": "8f3a1c2",
"uptimeSec": 12384,
"checks": {
"supabase": "ok",
"voyage": "ok",
"memoryMb": 218
}
}
Status semantics
| HTTP | Body ok | Meaning |
|---|---|---|
| 200 | true | All dependencies reachable. |
| 200 | false | Worker is up but degraded. Detail in checks. Treated as healthy by Railway. |
| 503 | false | Worker is starting up or in shutdown. Treated as unhealthy. |
The checks block reports the last successful probe of each
dependency. Probes run every 30 s in the background, so values are
near-realtime but not strictly per-request — the endpoint itself
is intentionally side-effect free.
Related
POST /index— the actual work endpoint.MCP tools— the stdio surface used by agent hosts.