The dagster-dbt library turns each model in a dbt project into a Dagster asset, read from dbt's manifest.json. @dbt_assets(manifest=project.manifest_path) decorates a function that runs dbt through the DbtCliResource: yield from dbt.cli(["build"], context=context).stream(). Each model becomes an asset with its dependencies; dbt sources become external assets upstream; dbt tests become asset checks.
select="tag:daily" limits the assets to part of the project, using dbt's selection syntax. Because the function receives the context, Dagster passes only the selected models to dbt, so you can materialize one model from the asset graph.
dbt build runs tests right after each model and skips models downstream of a failing error-severity test, then exits with an error; models it already built stay materialized. To make a Python asset depend on a model, look up its key: deps=[get_asset_key_for_model([shop_dbt], "fct_revenue")].
In this workspace dbt does not run: the project fixture gives each model's row count and each test's result, which differ in the hidden copy.
Every model in the manifest becomes an asset; dbt build runs them and their tests.