Upgrading to Snowflake SQLAlchemy 2.0 and Python Connector 5.0: a checklist

Snowflake SQLAlchemy 2.0.0 was released on October 19, 2026. It is a major release with breaking changes. The dialect requires SQLAlchemy >=2.0.44,<2.1, Snowflake Python Connector 5.0 or newer, and Python 3.11 to 3.14. Python Connector 5.0.0 requires Python 3.11 or later. That work still sits on sound data infrastructure: Terraform, CI/CD, dbt, agile delivery, Azure DevOps, GitHub Actions, and Airflow. Governed data stays in the customer's account.

Who is affected

Check unpinned Python stacks, dbt-adjacent tooling, Superset, and Alembic migrations. If snowflake-sqlalchemy is not pinned, the next resolve can install 2.0.0.

Breaking changes

• JSON-type columns return dicts or lists by default.
• Integer // now floors.
• force_div_is_floordiv and legacy_url_params are removed.

New in 2.0.0

• Async engine
• Native UUID
• COLLATE
• InsertMulti
• Row access policies
• RELY constraints

Checklist

1. Inventory where snowflake-sqlalchemy is installed. Search the Python environments that talk to the warehouse, including dbt-adjacent tooling, Superset, and Alembic migrations.
2. Pin snowflake-sqlalchemy below 2.0.0 now. Hold the current line while you test the major release on a non-prod branch.

   snowflake-sqlalchemy<2.0.0

3. Check the Python version. SQLAlchemy 2.0.0 supports Python 3.11 to 3.14. Python Connector 5.0.0 requires Python 3.11 or later.
4. Upgrade the Python Connector to 5.0. Do this in a non-prod branch, against a non-prod Snowflake account.
5. Test JSON column handling and integer division. JSON-type columns return dicts or lists by default, and integer // now floors.
6. Remove the deprecated parameters. Delete force_div_is_floordiv and legacy_url_params from engine configuration. Both are removed in 2.0.0.
7. Run CI against a non-prod Snowflake account. Use Azure DevOps Pipelines or GitHub Actions. The account under test should not be production.
8. Promote through your normal release process. Move the tested upgrade on the same release path as the rest of the data infrastructure.
9. Smoke-test Airflow and dbt jobs. Run the jobs that import this stack after the promotion, and confirm they still reach the warehouse in the customer's account.