Skip to content
Grafana

Serialization

The grafana lexicon writes Grafana’s own files, keyed by path under the directory of the -o output:

  • dashboards/<uid>.json, or dashboards/<folder>/<uid>.json for a dashboard with a folder: the dashboard JSON, one file per dashboard.
  • provisioning/datasources/chant.yaml: every declared datasource.
  • provisioning/dashboards/chant.yaml: a file provider for the dashboards directory; a default one unless a DashboardProvider is declared.
  • provisioning/alerting/chant.yaml: rule groups, contact points, the notification policy tree, mute timings and templates, when the build declares any.

The primary output (the -o file itself) is a JSON index of the dashboards, datasources, alerting and files that were built.

  • Dashboard JSON carries the pinned schema’s schemaVersion, no numeric id, and every field Grafana’s schema requires (link defaults, an empty annotations.list, fieldConfig.overrides).
  • Panels without a position are laid out in declaration order on the 24-column grid. Ids and query refIds are assigned in order unless set.
  • A declared Datasource is written as { type, uid } wherever it is referenced, and as its uid inside another datasource’s jsonData.
  • chant build rewrites JSON files with sorted keys; Grafana does not care about key order.
  • The dashboard JSON carries no ownership marker. Grafana keeps who manages a dashboard in the resource’s metadata on its dashboard.grafana.app API, not in the JSON: for a provisioned dashboard, the name of the provider that loaded it, which is chant unless a DashboardProvider names another. chant lifecycle diff --live reads that back (see Drift and Live Export).