Skip to content

Analytics Guides#

Analytics is the SQL, metrics and dashboard layer over the feature store: Trino queries the offline data, the semantic layer defines metrics over it, Superset charts it. Start with a query, then define metrics or turn it into a dashboard.

  • Start here


    Open the Query Engine from the project sidebar and run SQL against a feature group. One catalog per table format, so pick delta, iceberg or hudi first.

    SELECT customer_id, avg(amount) AS avg_amount
    FROM delta.fraud_featurestore.transactions_1
    GROUP BY customer_id
    ORDER BY avg_amount DESC
    LIMIT 10
    

    Query engine · Superset

Query

  • Query engine Run interactive SQL over feature groups, with query history and cluster status.
  • Trino catalogs Expose a data source as a catalog so external tables join feature data.
  • External BI tools How a BI tool connects to the offline store through Trino.

Metrics

  • Semantic layer Metrics defined once in YAML over feature groups, queried by name.
  • Define metrics Semantic models, and simple, ratio, derived, cumulative and conversion metrics.
  • Query metrics From Python, the CLI, the UI or an agent, as SQL or rows, or by saved query.
  • Connect dbt clients and BI tools dbt-sl-sdk, JDBC, ADBC and GraphQL, and semantic projects as tables for BI tools.
  • Metric exports Saved queries kept in feature groups by scheduled jobs.

Visualize