ml-api

Run modes

Server mode is controlled through .env.

Example:

APP_ENV=dev
APP_PORT=8000
API_TOKEN=your-token
LOKI_PORT=3100
GRAFANA_PORT=3000
GRAFANA_ADMIN_USER=admin
GRAFANA_ADMIN_PASSWORD=admin

Values:

• APP_ENV=dev - starts uvicorn with --reload, and code changes in api/ are applied automatically.
• APP_ENV=prod - starts the server without auto-reload.

Logging

The Docker stack includes Loki, Grafana, and Alloy.

• Alloy reads container logs from Docker and ships them to Loki.
• Grafana is provisioned with Loki as the default datasource.
• Grafana preinstalls the grafana-lokiexplore-app plugin required for Logs Drilldown.
• Application logs already go to stdout, so api and iris logs appear in Loki automatically.
• Loki is configured with volume_enabled: true, which is required for Grafana Logs Drilldown.

After startup:

• Grafana: http://localhost:${GRAFANA_PORT}
• Loki: http://localhost:${LOKI_PORT}

Open Grafana and query logs by labels such as service, container, and stream. In Logs Drilldown, logs should appear grouped by service_name for each Compose service/container.