Great question! In Prefect 3.x there are two recommended ways to orchestrate dbt for Snowflake, depending on whether you use dbt Core or dbt Cloud: - dbt Core (recommended if you run dbt from your own code repo/infra): Use prefect-dbt’s PrefectDbtRunner and point it at a Snowflake

profiles.yml

- dbt Cloud (recommended if you already manage jobs in dbt Cloud): Use prefect-dbt’s dbt Cloud tasks to trigger and wait on jobs Below are concise, production-ready patterns for both. 1) dbt Core + Snowflake (PrefectDbtRunner) - Install

pip install -U prefect prefect-dbt dbt-core dbt-snowflake
# Optional (for Snowflake Blocks if you want them elsewhere): pip install -U prefect-snowflake
# Optional (to create blocks from the UI): prefect block register -m prefect_dbt

- Flow example (runs dbt build and surfaces node-level status in Prefect)

from prefect import flow
from prefect_dbt import PrefectDbtRunner, PrefectDbtSettings

@flow
def dbt_build():
    runner = PrefectDbtRunner(
        settings=PrefectDbtSettings(
            project_dir="/path/to/your/dbt_project",
            profiles_dir="~/.dbt"  # adjust as needed
        ),
        raise_on_failure=True
    )
    runner.invoke(["build"])

if __name__ == "__main__":
    dbt_build()

- Snowflake credentials via dbt profiles.yml (most common) Create `~/.dbt/profiles.yml`:

my_dbt_profile:
  outputs:
    prod:
      type: snowflake
      account: "{{ env_var('DBT_SNOWFLAKE_ACCOUNT') }}"
      user: "{{ env_var('DBT_SNOWFLAKE_USER') }}"
      password: "{{ env_var('DBT_SNOWFLAKE_PASSWORD') }}"  # or use keypair/oauth options
      database: "{{ env_var('DBT_SNOWFLAKE_DATABASE') }}"
      schema: "{{ env_var('DBT_SNOWFLAKE_SCHEMA') }}"
      warehouse: "{{ env_var('DBT_SNOWFLAKE_WAREHOUSE') }}"
      threads: 4
  target: prod

Then set the env vars on your worker/agent or in your deployment:

export DBT_SNOWFLAKE_ACCOUNT="xy12345.us-east-1.snowflake"
export DBT_SNOWFLAKE_USER="dbt_user"
export DBT_SNOWFLAKE_PASSWORD="***"
export DBT_SNOWFLAKE_DATABASE="ANALYTICS"
export DBT_SNOWFLAKE_SCHEMA="TRANSFORM"
export DBT_SNOWFLAKE_WAREHOUSE="COMPUTE_WH"

Notes - This approach gives you dbt’s standard ergonomics and the best Prefect observability (node-level logs, failures, and retries at the Prefect task level if desired). - If you prefer to generate a profiles.yml at runtime from Prefect blocks, you can use the CLI profile helpers in prefect-dbt (DbtCliProfile + SnowflakeTargetConfigs). That’s handy when you don’t want a static profiles.yml on the image/host. Docs - prefect-dbt integration - dbt Snowflake profile reference 2) dbt Cloud jobs (trigger + wait + targeted retries) - Install

pip install -U prefect prefect-dbt
# Optional (to create blocks from the UI): prefect block register -m prefect_dbt

- Create a dbt Cloud credentials block

from prefect_dbt.cloud import DbtCloudCredentials
DbtCloudCredentials(api_key="YOUR_SERVICE_TOKEN", account_id=123456).save("dbt-cloud-creds")

- Flow to run a job and wait for completion (with targeted retries)

from prefect import flow
from prefect_dbt.cloud import DbtCloudCredentials
from <http://prefect_dbt.cloud.jobs|prefect_dbt.cloud.jobs> import run_dbt_cloud_job

@flow
def run_cloud_job():
    creds = DbtCloudCredentials.load("dbt-cloud-creds")
    result = run_dbt_cloud_job(
        dbt_cloud_credentials=creds,
        job_id=154217,       # your dbt Cloud job ID
        targeted_retries=3   # retry failed nodes automatically
    )
    return result

if __name__ == "__main__":
    run_cloud_job()

Notes - Use this if you already manage execution environments in dbt Cloud and want Prefect to coordinate runs with upstream/downstream tasks and alerting. Docs - dbt Cloud with Prefect Which should you choose?

- Use dbt Core + PrefectDbtRunner if you run dbt near your data warehouse (common for Snowflake), want full control of runtime, and want rich Prefect visibility. - Use dbt Cloud integration if you already rely on dbt Cloud for execution and job configuration. If you share whether you’re on dbt Core or dbt Cloud (and how you prefer to supply Snowflake creds: env vars vs. keypair vs. SSO), I can tailor a starter template for your setup.