Data modeling

After skippr sync lands bronze data, the Data Agent analyses schemas and produces business-ready models — dimensions, facts, and metrics — as a validated dbt project you can inspect, edit, and run.

AI assists with scaffolding. Schema metadata is the default input; data samples are optional and off by default. See Vision.

Pipeline

1. Analyse — schemas, joins, grain, and patterns across landed sources.
2. Model — author staging, dimension, and fact SQL as standard dbt.
3. Validate — compile and run against your warehouse (skippr test for dbt tests).
4. Publish — materialise tables you query in the warehouse you already use.

When validation fails, the agent reads the error, adjusts the model, and re-validates for common issues. You still review the SQL.

Related

• skippr model
• skippr test
• How it works