Use a profile mapping

Profile mappings are utilities provided by Cosmos that translate Apache Airflow® connections to dbt profiles. This means that you can use the same connection objects you use in Airflow to authenticate with your database in dbt. To do so, there’s a class in Cosmos for each Airflow connection to dbt profile mapping.

Note

Cosmos natively supports all adapters compatible with dbt Core and dbt Fusion without any reliance on dbt Platform. This includes both official Trusted Adapters and Community Adapters. While some of these adapters lack a built-in automated profile mapping class in Cosmos, including AWS Glue, Dremio, Azure Synapse, and IBM Netezza, among others, they remain fully operational out of the box. In these scenarios, you can simply bypass automated connection mappings entirely and pass a user-provided profiles.yml file configuration instead. For a detailed implementation guide, see Use your own profiles.yml file.

You can find the available profile mappings on the left-hand side of this page. Each profile mapping is imported from cosmos.profiles and takes two arguments:

• conn_id: the Airflow connection ID to use.

• profile_args: a dictionary of additional arguments to pass to the dbt profile. This is useful for specifying values that are not in the Airflow connection. This also acts as an override for any values that are in the Airflow connection but should be overridden.

Below is an example of using the Snowflake profile mapping, where we take most arguments from the Airflow connection but override the database and schema values:

from cosmos.profiles import SnowflakeUserPasswordProfileMapping

profile_config = ProfileConfig(
    profile_name="my_profile_name",
    target_name="my_target_name",
    profile_mapping=SnowflakeUserPasswordProfileMapping(
        conn_id="my_snowflake_conn_id",
        profile_args={
            "database": "my_snowflake_database",
            "schema": "my_snowflake_schema",
        },
    ),
)

dag = DbtDag(profile_config=profile_config, ...)

Note that when using a profile mapping, the profiles.yml file gets generated with the profile name and target name you specify in ProfileConfig.