fal configurations

Adapter no longer maintained

The dbt-fal adapter is no longer actively maintained. This means although the adapter is still operational, there is no further development or bug fixes planned and it may not be compatible with future versions of dbt. dbt-fal was test until dbt v1.5.

Documentation for dbt-fal are kept for reference purposes only and will eventually be removed from the site in the future.

Setting the db_profile

The fal profile configuration needs the db_profile property set to the profile configuring your database for SQL models.

fal will wrap around adapter and just handle Python models while letting all the SQL needs to the underlying database adapter.

fal will inherit the threads configuration from the db_profile unless explicitly specified.

Example:

profiles.yml

jaffle_shop:
  target: dev_with_fal
  outputs:
    dev_with_fal:
      type: fal
      db_profile: dev_pg # This points to your main adapter
    dev_pg:
      type: postgres
      ...

Using fal_environment model configuration

By creating a fal_project.yml in the same location as your dbt_project.yml and adding environment definitions in there:

fal_project.yml

environments:
  - name: clustering
    type: conda
    packages:
      - kmodes==0.12.2

  - name: predict
    type: venv
    requirements:
      - prophet

You can now reference any of these environments in your dbt Python models:

my_clustering_model.py

def model(dbt, fal):
    dbt.config({
        "fal_environment": "clustering"
    })

    import pandas as pd
    # kmodes is available because of the `fal_environment` being used
    from kmodes.kmodes import KModes

    df: pd.DataFrame = dbt.ref("order_detailed")
    df_train = df[["size", "is_vegan", "is_vegetarian", "is_keto", "shape"]]

    km_2 = KModes(n_clusters=3, init="Huang")
    km_2.fit_predict(df_train)
    df["cluster_label"] = km_2.labels_

    return df