Dashboards

Grafana dashboards are auto-provisioned from configs/grafana/provisioning/dashboards/, so a deployment gets a working set the moment the stack starts — no manual import. The Grafana file provider has foldersFromFilesStructure: true, so dashboards placed in a subdirectory land in a folder named after that subdirectory.

The 9 dashboards are grouped into two folders:

• General folder — 5 service dashboards at the root of dashboards/. (The file provider creates no application/ subdirectory, so these land in the default General folder.)
• Observability folder — 4 stack-health dashboards in dashboards/observability/.

First time opening Grafana

You do this once after deploy.

1. Browse to https://<your-domain>/grafana/ (the Kong route at /grafana/).
2. Keycloak SSO redirects you to the realm login — sign in with your normal realm user (e.g. an admin realm role is required to see admin-only dashboards).
3. The default landing dashboard is Service health (set by GF_DASHBOARDS_DEFAULT_HOME_DASHBOARD_PATH).

The local GRAFANA_ADMIN_USER / GRAFANA_ADMIN_PASSWORD form is a break-glass login (login form disabled by GF_AUTH_DISABLE_LOGIN_FORM=true). To enable it for emergency access, set GF_AUTH_DISABLE_LOGIN_FORM=false in docker-compose.yaml and restart Grafana.

General folder — service dashboards

Operational views of the GENIE.AI services themselves.

[Screenshot placeholder — Service health overview]

Observability folder — stack-health dashboards

Health of the observability stack itself — so the monitoring does not fail silently.

[Screenshot placeholder — Observability stack health]

Pivoting between dashboards

The dashboards share a common tag set (service.name, service.namespace, component) so pivots work out of the box:

• A metric spike on Application metrics → the matching trace on Trace explorer (or RAG pipeline trace waterfall for chat queries).
• A trace span on Trace explorer → its logs on Service logs (by trace_id) or its metrics on Application metrics (by service name + time window).
• A panel showing “no data” on Application metrics → the Observability stack health dashboard to confirm the collector is still ingesting.

How to actually pivot

Click any data point in a panel → Inspect or Drilldown → pick the linked datasource (Jaeger, VictoriaLogs, or VictoriaMetrics). The tag set is preserved so the linked query opens with the same service/namespace/time window pre-filled.

Empty panels — the most common new-deploy symptom

Empty panels usually mean the metric has not been emitted yet. Send a few chat queries through the UI, then reload. If panels stay empty after an hour, check the Observability stack health dashboard for ingestion errors and the Alerting page for firing rules.

The backend metrics http_requests_total and http_request_duration_seconds start at zero on a fresh deploy and only accumulate once Express middleware has recorded requests.

Customising

Dashboards are JSON files under configs/grafana/provisioning/dashboards/. The Grafana file provider scans the directory every 30 seconds (updateIntervalSeconds: 30 in dashboards.yml) — edits to the bind-mounted JSON appear in Grafana without a restart. To add a net-new dashboard, drop a JSON file in the folder (or a subdirectory for a custom folder); the dashboards.yml provider picks it up on the next scan.

If your edit does not appear after a minute, the JSON is invalid — check Grafana’s container logs (docker service logs genieai_grafana --tail 50) for parse errors.

Adding a new dashboard

1. Export or author a JSON file matching the Grafana schema.
2. Save it under configs/grafana/provisioning/dashboards/ (or a subdirectory to control the folder name).
3. Within 30 seconds it appears in Grafana → Dashboards → Manage.

Next steps

Once dashboards look healthy, wire up alerting — see Alerting for the built-in rules and how to add a contact point.