Tutorial

Note

Placeholder. The tutorial will be written once the self-service layer (notebooks and interactive SQL) is built, so it can run end to end.

The planned path, using the demo domain:

1. Land a file. Drop a day of returns into landing/ and watch the sensor load it into raw/ with dlt.

2. Stage it. Clean and type a raw table in Python (pandas or Polars), written as a Delta table by the I/O manager.

3. Model it. Add a dbt model, with its column declarations and tests, and materialize it from Dagster.

4. Check it. See the dbt tests and the schema-drift check as asset checks, and the lineage in the asset graph.

5. Query it. Open a notebook and query the mart with DuckDB, straight from the lake.

6. Swap the side. Point the same graph at S3 and compare the results.

Until then, the dbt Models in the Lake page and the Lake Layout and Naming Convention show the pieces individually.