Dashdown Docs

Cube connector

Connects to a Cube deployment — a standalone semantic-layer server where a team models measures and dimensions, served over a structured JSON API. Unlike every other connector, Cube isn't queried with SQL: the connector is a thin HTTP+JWT client that powers the backend: cube semantic layer, so you reference its model with the metric={…} by={…} grammar rather than :::query SQL.

Experimental — preview

The Cube integration is preview — fully unit-tested with fakes but not yet verified against a live Cube deployment, and it covers the measures/dimensions/timeDimensions/filters subset of Cube's query shape. Treat it as a preview to try, not a production guarantee.

# sources.yaml
cube:
  type: cube
  url: https://cube.example.com
  secret: ${CUBE_API_SECRET}      # HS256 signing secret (env-expanded)
  token_ttl: 300                  # JWT lifetime in seconds (default)
  security_context:               # optional — embedded in every JWT (the RLS rail)
    tenant_id: acme
Key Purpose
url Required. Base URL of the Cube deployment.
secret HS256 signing secret — Dashdown mints a short-lived JWT per request.
private_key RS256 private key, as an alternative to secret.
token A static pre-minted JWT (escape hatch — no minting).
algorithm Signing algorithm (default HS256).
security_context Claims embedded in every JWT — Cube applies them server-side for row-level security.
api_path API path (default /cubejs-api/v1).
token_ttl JWT lifetime in seconds (default 300); re-minted before expiry and once on a 401.
timeout Request timeout in seconds (default 60).

One of secret, private_key, or token is required.

How it's used #

Because Cube speaks a structured JSON query (not SQL), this connector is paired with a semantic/ model — Cube's GET /meta auto-fills the catalogue, so a model is as small as orders: { connector: cube }:

# semantic/orders.yml
orders:
  connector: cube
<LineChart metric={orders.revenue} by={orders.createdAt} grain="month" />
<BarChart  metric={orders.count}   by={orders.status} />

See the semantic layer → Cube backend for the full modeling, filter mapping, JWT lifecycle, and grain= behaviour. There is no ${param} injection surface — filter values are JSON data, never assembled into a query string.

Just want SQL access to Cube?

Cube also exposes a Postgres-wire-compatible SQL API. You don't need this connector for that — point a postgres connector at Cube's SQL port and write queries/*.sql. This connector is for the first-class metric={…} by={…} grammar.

Install: uv add 'dashdown-md[cube]'.

Generated