Skip to content

Operations, Health, and Metrics

  • Health

    /health and /ready for liveness and readiness.

  • Metrics

    /metrics for Prometheus. Plus Postgres exporter for DB metrics.

  • Runtime Control

    Inspect and restart containers via Docker endpoints.

Get started Configuration API

Readiness Gate

Deployments should route traffic only after /ready returns success. This ensures PostgreSQL and Neo4j are reachable and responsive.

Scrape Interval

Metrics scraping interval can be 10–30 seconds depending on traffic and budget.

High-Cardinality Labels

Avoid per-query labels in Prometheus if cardinality explodes. Aggregate at the corpus or retriever level.

Endpoints

Endpoint Description
/health Process liveness
/ready Readiness including DB checks
/metrics Prometheus metrics
/docker/status Container status
/docker/{container}/restart Restart container
/docker/{container}/logs Tail logs 
flowchart LR
    Scrape["Prometheus"] --> API_METRICS["/metrics"]
    API_METRICS --> APP["TriBridRAG"]
    APP --> PG[("Postgres")]
    APP --> NEO[("Neo4j")]
    Scrape --> PExp["postgres-exporter"]

Examples

import httpx
print(httpx.get("http://localhost:8000/health").json())
print(httpx.get("http://localhost:8000/ready").json())
print(httpx.get("http://localhost:8000/metrics").text.splitlines()[:5])

curl -sS http://localhost:8000/health | jq .
curl -sS http://localhost:8000/ready | jq .
curl -sS http://localhost:8000/metrics | head -n 20

await fetch('/health').then(r => r.ok || Promise.reject('down'))
await fetch('/ready').then(r => r.ok || Promise.reject('not ready'))
const sample = await (await fetch('/metrics')).text();
console.log(sample.split('\n').slice(0, 5));
  • Gate traffic with readiness
  • Alert on 5xx and slow search
  • Monitor DB connection pool saturation and timeouts
flowchart TB
    Alert["Alerts"] --> OnCall["On-Call"]
    Metrics["Metrics"] --> Alert
    OnCall --> Mitigate["Mitigation"]
Log Access

Use /docker/{container}/logs for quick log retrieval. For long-term retention, integrate with a centralized logging solution. Loki is included in the compose stack.