Release Notes

Python API

0.9.18

September 24, 20255:00 PM UTC

Version 0.9.18 of the relationalai Python package is now available. This release includes improvements to logging and observability to aid in support and debugging.

To upgrade, activate your virtual environment and run the following command:

pip install --upgrade relationalai

New Features and Enhancements

Application telemetry now include details about transactions executed when direct access is turned on. This enhancement improves observability to help diagnose issues and enable improved support.

Version 0.9.17 of the relationalai Python package is now available. This release includes enhanced engine provisioning error messages and improved Snowflake JWT (Javascript Web Token) authentication for better security and reliability.

To upgrade, activate your virtual environment and run the following command:

pip install --upgrade relationalai

Breaking Changes

The old EngineAutoCreateFailed exception has been replaced with EngineProvisioningFailed, which surfaces the original error details for clearer diagnostics during engine provisioning failures.

New Features and Enhancements

Engine provisioning errors now include the original underlying cause to help you troubleshoot more effectively. When engine creation fails—due to issues like connection timeouts or service problems—the error message will display the specific reason alongside guidance on manual engine management.
Improved security and reliability for Snowflake key-pair JWT authentication in the Python client:
- The JWT generated from your private key is now exchanged at Snowflake for a new access token scoped to your configured role and the specific endpoint being accessed, enhancing security by limiting token permissions.
- Token retrieval is thread-safe and cached, preventing redundant token requests when creating multiple resources concurrently.
- Both Snowflake account name and account locator formats are supported in configuration, resolving previous SSL mismatch errors.

0.9.16

September 19, 20254:12 PM UTC

Version 0.9.16 of the relationalai Python package is now available. This release includes the ability to skip invalid rows during data export or query with detailed warnings to improve transparency and troubleshooting.

To upgrade, activate your virtual environment and run the following command:

pip install --upgrade relationalai

New Features and Enhancements

You can now skip invalid rows during data export or query by using the new skip_invalid_data=True option in the Python API’s query and export methods. When enabled, any rows with invalid data (such as type mismatches or constraint violations) are automatically skipped instead of causing failures.

Importantly, the system will provide detailed warnings showing which rows were rejected, which columns caused errors, and the specific error messages. This helps you quickly identify and fix data quality issues without silent failures or guesswork.

Example Usage:

# Skip invalid data during export
@model.export("<my_db>.<my_schema>", skip_invalid_data=True)
def my_export():
    ...

# Skip invalid data during query:
with model.query(skip_invalid_data=True) as select:
    ...

Example warning output:

Your data has been loaded but 2 rows were skipped due to erroneous data. Here are the first 2 rejected rows:
Rejected record: {"id": "abc", "amount": "not_a_number"}
- Erroneous column: amount
  Error message: Could not convert value to numeric

Rejected record: {"id": 123, "amount": null}
- Erroneous column: amount
  Error message: Null value not allowed