Monitoring

The service exposes Prometheus metrics at /metrics. See Service Metrics for a full reference of available metrics.

Configuration

METRICS_SCRAPE_KEY is a required environment variable. The service will refuse to start without it. Generate the key and store it safely.

The /metrics endpoint is protected by authentication. Every request must include the header:

Authorization: Bearer <METRICS_SCRAPE_KEY>

Example: Prometheus Scrape Job

Add a scrape job to your Prometheus configuration:

scrape_configs:
  - job_name: 'ai-inference'
    static_configs:
      - targets: ['localhost:8000']
    authorization:
      type: Bearer
      credentials: <METRICS_SCRAPE_KEY>

The easiest way is a separate docker-compose.yml on the same host. Use network_mode: "host" so Prometheus sees the service on localhost:

services:
  prometheus:
    image: prom/prometheus:latest
    container_name: prometheus
    network_mode: "host"
    volumes:
      - ./prometheus.yml:/etc/prometheus/prometheus.yml:ro
      - ./data:/prometheus
    command:
      - '--config.file=/etc/prometheus/prometheus.yml'
      - '--storage.tsdb.path=/prometheus'

Place your prometheus.yml (with the scrape job above) in the same directory and start with docker compose up -d. Prometheus will be reachable on localhost:9090.

Example 2: Grafana Dashboard

Point a Grafana instance at the Prometheus data source to build dashboards for request rate, latency histograms, error ratios, and Go runtime metrics.

A sample dashboard JSON can be exported from Grafana after setting up the desired panels — the Prometheus label structure (method, path, status) is stable and can be relied on for alerts and dashboards.