Skip to main content
A MantrixFlow pipeline combines one shared source configuration with one or more independent destinations. Each destination owns its SQL transformations, published revisions, output assignments, and checkpoints.

​
Workspace tabs

The current pipeline detail page replaces the former canvas builder. The legacy /workspace/data-pipelines/[id]/builder route only redirects old links into the appropriate workspace tab.

​
Lifecycle

  1. Create a pipeline from Data Pipelines.
  2. Configure and save streams in Source.
  3. Add a delivery target in Destinations.
  4. Select that destination in Transformations and create its SQL draft.
  5. Save, validate, preview, and publish the transformation revision.
  6. Return to the destination editor and assign the published output table and Upsert keys.
  7. Validate readiness in Overview.
  8. Click Run all in the header.
  9. Inspect the execution in Runs.
  10. Configure automation from Settings after manual validation.

​
Execution model

During a run:
  1. MantrixFlow extracts selected source streams into isolated temporary staging.
  2. Published destination-owned SQL revisions read relations such as raw.public__orders.
  3. Each destination receives only its assigned model output.
  4. Upsert delivery writes to the configured existing table.
  5. Successful checkpoints are recorded and temporary staging is cleaned up.

​
Destination contract

Final database tables use schema.table or database.table. Airtable uses a selected existing base/table plus explicit field mappings. The delivery runner does not create final tables. Use the destination editor’s explicit Create or verify destination table action when appropriate, or create the reviewed table yourself.

​
Write behavior

The current write mode is Upsert. Configure stable keys for every published output assignment. Database keys should match a primary or unique key; Airtable merge fields must be writable, mapped, and non-empty for every output row.