dltHub's AI Workbench data quality toolkit bootstraps checks directly from dlt's existing schema metadata (primary keys, non-nullable columns, uniqueness hints), samples columns before writing any rules, and embeds checks as decorators inside pipeline.run() via @dq.with_checks and @dq.with_metrics. Results are stored in destination tables (_dlt_checks, _dlt_dq_metrics) for querying or alerting. When failures occur, an LLM routes the fix to the appropriate toolkit: ingestion issues go to rest-api-pipeline, modeling issues to transformations, anomalies to data-exploration. This contrasts with tools like Great Expectations, which only report failures without routing fixes. The toolkit is part of dltHub Pro with a free trial available.
Table of contents
Data quality to the rescue, but not YOLO Link iconBootstrapping from what dlt already knows Link iconMost DQ tools are the lab result. This one is the medical system Link icon110 Impressions