Mage Recipes¶

This page collects a few high-signal patterns that map well to the current package surface.

Validate A Dataset Before A Load¶

from truthound_mage import CheckBlockConfig, CheckTransformer


def transform(df, *args, **kwargs):
    result = CheckTransformer(
        config=CheckBlockConfig(
            rules=[
                {"column": "id", "check": "not_null"},
                {"column": "email", "check": "email_format"},
            ]
        )
    ).execute(df)
    return result.result_dict

Use this pattern when a pipeline should fail before data is written downstream.

Generate A Profile For Operator Review¶

Use ProfileTransformer for baselining, onboarding, and change review.

This is especially useful for pipelines that ingest third-party feeds whose quality shape changes over time.

Gate Downstream Work With A Sensor¶

Use QualityGateSensor or DataQualitySensor when downstream blocks should run only after a check result meets pass-rate or failure-rate thresholds.

This keeps quality gate semantics explicit instead of scattering them across block code.

Route Hard Failures And Soft Warnings Differently¶

Combine:

strict CheckTransformer configs for blocking datasets
condition or sensor blocks for soft rollout paths
SLA hooks for operator visibility

Keep Shared Config In One Place¶

Create one helper module that returns prebuilt CheckBlockConfig, ProfileBlockConfig, and SensorBlockConfig objects. This prevents configuration drift across pipelines.