Datasource Routing

Queries identify a datasource with kind, name, and an optional scope. The runtime resolves that reference against registrations built by the host.

const registry = new DatasourceRegistry()
  .register({
    name: "primary",
    kind: "ClickHouseDatasource",
    scope: "workspace",
    config: primaryConfig,
    adapter: new ClickHouseAdapter(primaryConfig),
  })
  .register({
    name: "archive",
    kind: "ClickHouseDatasource",
    scope: "workspace",
    config: archiveConfig,
    adapter: new ClickHouseAdapter(archiveConfig),
  })
  .register({
    name: "tempo-prod",
    kind: "TempoDatasource",
    scope: "workspace",
    config: tempoConfig,
    adapter: new TempoAdapter(tempoConfig),
  });

Resolution priority is dashboard, then workspace, then global unless the query specifies a scope. A matching name with the wrong kind fails with DATASOURCE_KIND_MISMATCH; it never falls through to an unrelated adapter.

Selected connectors

A product such as OpenLIT can store a selected connector ID per signal and map that record to the named registration. The actual traces, logs, or metrics then come from the selected instance. OpenPlait does not select a product connector or read its database—it executes the datasource reference supplied by the host.