Test data quality with AI in GX
Turn plain-language data rules into checks against a known batch and find out which conditions passed, failed or could not be evaluated.
Teams can test data quality with AI in GX when the rules already have a clear business meaning. If a shipment reference must be unique within each delivery, for instance, the agent needs to know that delivery boundary before checking a whole table. The gx-configure-expectations skill matches the requirement to expectation types available in your installed GX package.
It runs the registered suite against a batch you can already read. A broken calculation calls for a different response from a genuine duplicate, so the report keeps those outcomes separate and relates each one to its own expectation.
What it does
- Check that a column contains the values your team expects.
- Validate a range without adding thresholds the business never supplied.
- Use a known partition to check a rule against the correct scope.
- Review the result of each condition in a multi-rule suite.
- Keep a named suite for the next batch of files.
Install and use
- Prepare current GX Core in Python, a supported coding agent and a working GX batch definition. Use data whose results may be shown in the session.
- State each condition in ordinary language, including the field and any permitted exceptions.
- Read the results individually and confirm where to persist a suite created in memory.
FAQ
Source
Made by Great Expectations. Original skill: gx-configure-expectations, published on GitHub under the Apache-2.0 licence.
View on GitHub