grafana

name	grafana
description	Grafana on the yai stack — datasource UIDs, file provisioning (not TF), dashboard conventions, scrape targets page, Grafana plugin setup. Auto-load on Grafana, dashboard, datasource, DS_*, panel, templating variable, provisioning questions.
when_to_use	Load when the user asks about Grafana dashboards, panels, datasources, Grafana API calls, provisioning files, plugin setup, or wants to view/create/update any dashboard or panel on the yai stack.

Grafana

URL in env: YAI_GRAFANA_URL (port 22000). API token: YAI_GRAFANA_TOKEN. Default admin login: admin / password set in grafana/.env.

Provisioning

Grafana in the yai stack uses file-based provisioning, not the Terraform provider. Datasources are provisioned via files mounted into the container at /etc/grafana/provisioning/datasources/. Dashboards are currently managed via the Grafana UI (first-time setup) — file provisioning for dashboards is a reasonable follow-up.

The Grafana UI is write-enabled — edits you make in the UI persist in grafana/data/grafana/. They are not committed to the repo.

Datasources

Datasource URLs use compose internal service names (resolved from inside the Grafana container, not from the host):

Datasource	Type	Internal URL
VictoriaMetrics	victoriametrics-metrics-datasource	http://victoriametrics:8428
VictoriaLogs	victoriametrics-logs-datasource	http://victorialogs:9428
VictoriaTraces	jaeger	http://victoriatraces:10428/select/jaeger/

Don't substitute localhost:<host_port> — those don't resolve from inside the Grafana container.

HTTP API quick-ref

source ./env.sh
H="Authorization: Bearer $YAI_GRAFANA_TOKEN"

# All dashboards
curl -s -H "$H" "$YAI_GRAFANA_URL/api/search?type=dash-db" \
  | jq '.[] | {uid, title, folderTitle}'

# One dashboard JSON
curl -s -H "$H" "$YAI_GRAFANA_URL/api/dashboards/uid/<UID>" | jq

# All datasources
curl -s -H "$H" "$YAI_GRAFANA_URL/api/datasources" \
  | jq '.[] | {uid, name, type}'

# Health check
curl -s "$YAI_GRAFANA_URL/api/health" | jq

Scrape targets page

The easiest way to see which scrape jobs are up (served by VictoriaMetrics built-in scraper):

open "$YAI_VMETRICS_URL/targets"
# or as JSON
curl -s "$YAI_VMETRICS_URL/api/v1/targets" | jq '.data.activeTargets[] | {job: .labels.job, health}'

Panel conventions

• Use PromQL/MetricsQL against VictoriaMetrics for metrics panels.
• Use LogsQL against VictoriaLogs for log panels (see vlogs skill).
• Trace panels use the Jaeger datasource against VictoriaTraces.
• Multi-select template variables in regex: use :pipe format.

Grafana plugins (auto-downloaded on start)

GF_INSTALL_PLUGINS in grafana/.env triggers download from grafana.com on container start. In an air-gapped setup, bake a custom Grafana image.

The two required Victoria* plugins:

• victoriametrics-metrics-datasource (VictoriaMetrics)
• victoriametrics-logs-datasource (VictoriaLogs)