Skip to main content
SQL transformations are managed from Transformations, not from a canvas or Destination panel.

Create or open a model

  1. Open the pipeline’s Transformations tab.
  2. Select a destination.
  3. Click New transformation or open an existing row.
  4. Choose input streams, enter a stable key and output model, and write SQL.

Source references

Use source() references to read staged source data:
The source stream public.orders becomes public__orders. SaaS sources use names such as stripe__customers; Airtable uses the lowercase table-ID staging name shown by the editor.

Example

Draft and publish lifecycle

  1. Save draft stores the SQL without affecting runs.
  2. Validate draft checks the SQL and output model.
  3. Preview output executes the valid draft for inspection.
  4. Publish revision makes it the destination’s active run input.
The editor also shows revision history and the currently published version.

Delivery assignment

Publishing creates an output that can be assigned in the owning destination editor. Under Published output targets, enter the exact final table and Upsert keys. Airtable destinations map output columns to writable fields and select merge fields. The final destination table must exist before delivery.