Skip to main content
Use the Filter node when the job is only to keep some records and drop others. If you also need to rename, flatten, or derive fields, use transform(record) instead.

​
Two live filter modes

​
Python Expression

Use a record-level expression such as: 
record.get("status") != "deleted"
Return True to keep the record and False to drop it.

​
SQL WHERE

Use a source-side predicate such as: 
status != 'deleted' AND created_at >= '2026-01-01'
The current builder stores that as a WHERE clause pattern for the source query path.

​
Step by step

  1. Add or open a Filter node on the branch.
  2. Choose Python Expression or SQL WHERE.
  3. Enter the condition.
  4. Save the node.
  5. Validate with source preview, transform preview, and one manual run before scheduling.

​
When to use each mode

  • Use Python Expression when the logic depends on the shaped record fields you already know from the payload.
  • Use SQL WHERE when the condition is simple and source-side filtering is clearer.

​
Real-world examples

  • Python: record.get("financial_status") == "paid"
  • Python: record.get("email") and "@example.com" not in record.get("email")
  • SQL: updated_at >= '2026-04-01'
  • SQL: status != 'deleted'

​
Current validation guidance

The filter panel is primarily a configuration surface today. Treat source preview, transform preview, destination preview, and a manual pipeline run as the reliable validation path.