Skip to main content

Create dbt assets

When dbt executes commands, it does so at the project level. For example, running dbt run executes all models in the project. You can pass flags like --select to limit execution, but the default behavior treats the project as a single unit.

In Dagster, we take a different approach: each dbt model becomes its own asset. This lets us build a fine-grained dependency graph where each model is linked to its upstream and downstream assets across both dbt and non-dbt layers within the Dagster project.

1. Scaffold the dbt component definition​

The best way to integrate a dbt project with Dagster is through the dbt component. This component turns a dbt project directory into a set of Dagster assets, one per model. You can scaffold it with dg:

dg scaffold defs dagster_dbt.DbtProjectComponent transform --project-path src/project_dbt/analytics
Creating defs at <YOUR PATH>/src/dagster_tutorial/defs/transform.

The generated YAML already points to the correct project path, so no edits are needed:

type: dagster_dbt.DbtProjectComponent

attributes:
  project: '{{ project_root }}/src/project_dbt/analytics'

2. Customize the dbt model assets​

By default, all dbt models are represented as assets, but the asset keys may not align with the rest of your Dagster project.

To fix this, we can add a translation section to the scaffolding. Translations let us customize the names of the assets created by the component:

type: dagster_dbt.DbtProjectComponent

attributes:
  project: '{{ project_root }}/src/project_dbt/analytics'
  translation:
    key: 'taxi_{{ node.name }}'
    group_name: dbt

This preserves lineage and ensures proper dependency tracking. In this example, the translator prepends taxi_ to all dbt source names, so source (zones) maps cleanly to the existing taxi_zones asset in Dagster.

Next steps​