Prometheus integration

Scrape Rime on-prem metrics with Prometheus.

Prometheus is a monitoring and alerting system commonly used to scrape and review operational metrics.

Prerequisites

Before proceeding, confirm:

1. You can reach the generator’s metrics port on your model container. Metrics are not served on the engine’s HTTP port (8080); that port serves only synthesis and the /livez and /readyz probes. Set GENERATOR_SERVICE_PORT explicitly on the model container (for example 30000) and scrape that port. If the variable is unset, the generator binds a random free port, which cannot be scraped reliably.
2. You have permissions to edit the main configuration file, typically prometheus.yml.
3. You have permissions to run a Docker container in your environment.

Running Prometheus

Step 1: Include a Prometheus target

Open prometheus.yml. Add a job pointing to your Rime Engine container:

global:
  scrape_interval: 5s  # Scrape metrics every 5 seconds

scrape_configs:
  - job_name: "model-rime-tts"
    static_configs:
      - targets: ["<instance_ip>:30000"]  # GENERATOR_SERVICE_PORT on the model container
    metrics_path: "/metrics"  # Default path for Prometheus metrics

Replace instance_ip with the relevant hostname or IP address, then save the file.

Step 2: Run Prometheus in a container

docker run -d \
  --rm \
  -p 9090:9090 \
  -v $(pwd)/prometheus.yml:/etc/prometheus/prometheus.yml \
  prom/prometheus:latest

Step 3: Validate the setup

Access the Prometheus UI (commonly at http://localhost:9090), navigate to Status → Targets, and look for the model-rime-tts job. Its status should be UP. That confirms Prometheus is collecting metrics.

Exploring metrics

Query the collected metrics from the Prometheus web interface or its API. Grafana and similar dashboarding tools can read from Prometheus to build dashboards and alerts.