Skip to content
northstar
Docs menu

SQL Executor

Read-only queries on your warehouses, within your dataset access.

In short

The SQL Executor runs read-only SQL on your warehouses, with paginated results and AI assistance. Each query is checked table by table against your dataset access: you can only read the tables of datasets you can see. It is a tool for the data team; business users get the same answers through Chat, KPIs and dashboards.

Read-only queries

  1. Pick a storage system, and optionally a dataset to prefill the query and power autocompletion.
  2. Write the query. Each run accepts a single SELECT statement: nothing can be created, changed or deleted.
  3. Run it. Results are paginated; add an ORDER BY to keep a stable order from one page to the next.

Qualify table names the way your warehouse expects, for example schema.table on PostgreSQL or database.schema.table on Snowflake. The editor shows a hint for the selected warehouse.

What you can query

  • A table can be queried only if it belongs to a dataset you can see. The check runs on every query, table by table.
  • Platform administrators can query every table of the storage system.
  • Some roles can draft queries but not run them: running needs the execute permission of the SQL Executor.

AI help

  • Generate with AI writes a query from your description, for example "monthly revenue by country over the last 12 months". It replaces the editor's content; review it before running.
  • Fix with AI appears when a query fails. It proposes a corrected query in the editor; review it, then run again.
  • Generate AI summary reads the first rows of the page shown and explains them. It only runs when you ask, and those rows are sent to the AI.

FAQ

Why is a table refused although it exists in the warehouse? It is not registered as a dataset you can see. Ask for access to the dataset, or for the table to be registered.

Can I save a query? Turn it into a governed asset instead: a KPI defined in SQL, or a tile of a dashboard.

Why is the storage system list empty? You need access to at least one dataset stored in a SQL warehouse.

See also