Setting the file. One moment.
Subchapter 2.120
references/integrations/dagster-dbt/dependencies.mdMarkdown1 KBView on GitHub
Dagster parses the dependencies already present in your dbt project:
ref() calls create dependencies between modelssource() calls create dependencies on upstream assetsDefine a Dagster asset as a dbt source in sources.yml:
sources:
- name: dagster
tables:
- name: my_upstream_assetThen reference it in your dbt model:
select * from {{ source('dagster', 'my_upstream_asset') }}This creates a data dependency where the dbt model reads from the Dagster asset.
To add dependencies not encoded in dbt’s data lineage (e.g., scheduling constraints without data reading), use a Jinja comment in your model:
-- depends_on: {{ source('dagster', 'upstream') }}
SELECT ...When dbt compiles the project, it evaluates this Jinja template and adds the source to the model’s dependency list in the manifest. Dagster parses the manifest and creates the dependency edge. This creates a scheduling dependency without requiring the model to SELECT from that source.