Datasets

Creating Datasets

Walk through the dataset creation wizard from picking a source to naming the final result.

The wizard

From the sidebar open Datasets and click Create dataset. The wizard walks you through three steps: source, configuration and metadata.

Screenshot · Dataset creation wizard — step 1 of 3.
Dataset creation wizard — step 1 of 3.

1. Source

  • Pick a connection

    Choose the database the dataset reads from. Switching the connection later is supported but resets the query.

  • Write or paste the query

    Use the embedded SQL editor — same autocomplete and schema browsing as the full Query Editor. For document databases, write a find expression instead.

  • Preview

    Click Preview to run the query and confirm the shape of the result. You can adjust the SQL until it looks right.

2. Configure execution

  • Live — re-runs the query on every read. No cache.
  • Cached — refresh on a schedule (every 5 min, 1 hr, daily, custom cron). Reads in between hit the cached snapshot.

Pick the cheapest cadence

Long-running aggregates with hourly cached refreshes feel instant to readers and still cost the database just once per hour.

3. Name and describe

Give the dataset a clear name (it'll appear in every chart picker) and a one-line description so teammates know what it contains. Optional tags help with discovery once your library grows.

From a saved query

If you already have a saved query you like, open it from the editor and pick Promote to dataset. The wizard skips the source step and lands you straight on configuration.

With the dataset created, head to Editing Datasets to add columns, transforms and refinements, or jump straight to Creating a Report to visualise it.