This is an example project for running fal-serverlesss with dbt using the dbt-fal adapter. We will run a sentiment analysis model on some fake Zendesk data by building and running a Python model on fal-serverless on a GPU machine.

git clone [email protected]:fal-ai/dbt-fal-serverless-demo.git

pip install fal-serverless dbt-fal[snowflake]

fal-serverless auth login

Follow the link that's generated and login using GitHub. Come back to the shell, when ready. (Reach out to the fal team for access if you don't already have it)

fal-serverless key generate

This will print a message containing values for KEY_ID and KEY_SECRET. We will need these for setting up the dbt profile.

In order to run your Python models in fal-serverless, you should update the profiles.yml to include the newly generated credentials. Here's an example of how it should look like:

fal_profile:
  target: fal_serverless
  outputs:
    fal_serverless:
      type: fal
      db_profile: db
      host: <ask the fal team>
      key_secret: MY_KEY_SECRET_VALUE
      key_id: MY_KEY_ID_VALUE
    db:
      type: snowflake
      username: USERNAME
      password: PASSWORD

That's it. Doing a dbt run against this profile will execute your Python models in fal-serverless.

dbt seed will create the seed data. dbt run will run the SQL model and then run the Python model on fal Serverless on a GPU machine.