v1.11.15

Features

• session.remote_sql() — server-side pushdown for Python models.
  Python models run client-side, so dbt.ref('big_table').filter(...)
  streams the entire upstream table across the wire before filtering
  locally. The new session.remote_sql(query) escape hatch runs arbitrary
  SQL on the GizmoSQL server over the existing ADBC connection and
  returns only the result as a local DuckDB relation, so filters and
  aggregations execute server-side and only the matching rows cross the
  network:

  def model(dbt, session):
      dbt.config(materialized="table")
      schema = dbt.this.schema
      return session.remote_sql(
          f"select * from {schema}.big_table where name = 'Joe'"
      )

  The session arg is now a _GizmoSQLSession proxy that delegates every
  attribute to the underlying local DuckDB connection (session.sql(),
  session.register(), etc. keep working unchanged) and adds
  remote_sql() on top — additive, not a replacement. remote_sql()
  returns a chainable relation you can combine with .filter(),
  .project(), .df(), pandas, etc., or return directly from the model.

Test suite

• New TestPythonRemoteSQLPushdown functional test: materializes a
  multi-row upstream table and a Python model that uses
  session.remote_sql() with a WHERE filter, then asserts the output
  table contains exactly the server-filtered rows.

🙏 Thanks to: @maartenbosteels for the feature idea!

Full Changelog: v1.11.14...v1.11.15