Skip to main content

Superset Fronting

Superset demonstrates governed analytics behavior over a local BI system.

The package is:

What It Proves

Superset is a strong ANIP example because analytics systems often expose powerful primitives: SQL, datasets, charts, dashboards, and publishing. ANIP should not make raw SQL the agent-facing product interface.

The contract exposes governed analytics capabilities:

  • Discovery is bounded to allowed datasets, charts, dashboards, databases, metrics, and dimensions.
  • Analytics answers use provider-owned semantic execution, chart data, dataset metadata, or saved-query behavior, not arbitrary agent-supplied SQL.
  • Chart creation defaults to preview.
  • Chart publishing, dashboard updates, and dataset drafts require approval posture.
  • Raw export, unrestricted SQL, protected database access, and publication bypass are denied.

Capability Surface

CapabilityIntent
superset.analytics.discover_contextDiscover allowed analytics context.
superset.analytics.answer_questionAnswer a bounded analytics question without exposing raw SQL as the agent interface.
superset.chart.preview.createPrepare a chart preview.
superset.chart.publish.requestRequest chart publication through approval posture.
superset.dashboard.draft.preparePrepare a dashboard draft.
superset.dataset.draft.preparePrepare a dataset draft.

Backend Boundary

The backend is native Superset REST/API behavior. Superset MCP can be useful as a comparison surface, but this showcase does not expose MCP tools as the ANIP product contract.

The important difference:

Agent-facing: governed analytics capability
Backend-facing: Superset-native execution binding

Artifacts

ArtifactPath
Source specdocs/examples/superset-fronting-showcase/source-spec.md
Packageexamples/showcase/superset_fronting/registry-packages/superset-fronting-showcase-0.2.0.anip-package.json
Service definitionexamples/showcase/superset_fronting/registry-packages/superset-fronting-showcase-0.2.0-service-definition.json
Custom bundlesexamples/showcase/superset_fronting/custom-code-bundles/
Local composeexamples/showcase/superset_fronting/compose/
Generated servicesexamples/showcase/superset_fronting/generated/

Local Validation

Superset is designed to be runnable locally because it should not require a paid SaaS account to understand the showcase.

Start the local Superset 6.1 stack:

examples/showcase/superset_fronting/compose/setup.sh

Then run the live smoke:

python3 examples/showcase/superset_fronting/scripts/live_smoke.py

The smoke should validate bounded discovery and preview behavior without exposing raw SQL as an agent input.