Great question — env vars only help dbt if they’re present in the environment of the process that actually runs dbt. With Prefect’s dbt CLI tasks (like

dbt_build

/

run_dbt_build

), there isn’t a special parameter to pass env vars; the dbt subprocess inherits

os.environ

from your flow process. That means: - If you run the flow locally, calling

load_dotenv()

before

run_dbt_build(...)

is usually enough. - If you “invoke” a flow/deployment so it runs on a worker (Docker, Kubernetes, Process/Universal), the env vars must exist in that worker’s runtime environment (or be injected via work pool job variables, Dockerfile, Kubernetes env, etc.). Loading a

.env

file locally will not affect a remotely executed flow. Concrete options 1) Local run (single process) - Load your

.env

before calling dbt:

from prefect import flow
from prefect_dbt.cli.commands import run_dbt_build
from dotenv import load_dotenv

@flow
def dbt_flow():
    load_dotenv(override=True)  # ensures os.environ is populated
    return run_dbt_build(
        project_dir="/path/to/project",
        profiles_dir="/path/to/profiles",
        stream_output=True,
    )

if __name__ == "__main__":
    dbt_flow()

2) Deployment on a worker (recommended approaches) - Put the env vars in the worker’s environment: - Process/Universal worker: export them before starting the worker - Docker worker: add

ENV

lines in your Dockerfile or pass env via job variables - Kubernetes worker: set env or secrets on the pod - Or pass them via deployment job variables when you deploy from source:

flow = flow.from_source(
    source="<https://github.com/your-org/your-repo>",
    entrypoint="flows/dbt_flow.py:dbt_flow",
)
flow.deploy(
    name="dbt-build",
    work_pool_name="your-pool",
    job_variables={
        "env": {
            "database": "my_db",
            # Optionally, if needed by dbt:
            # "DBT_PROFILES_DIR": "/opt/dbt/profiles"
        }
    }
)

3) Pull from Prefect Variables or Blocks, then set os.environ before running dbt

from prefect.variables import get
import os

os.environ["database"] = get("database")  # make it visible to dbt

A few gotchas to check - The key you use in

env_var('...')

must exist in the dbt process’s environment at runtime. If you use

{{ env_var('database') }}

, make sure the variable name matches exactly (case sensitive on most systems). - Ensure you call

load_dotenv()

before invoking

run_dbt_build

. - If you’re running remotely, confirm your worker/container actually has those env vars (print them in the flow just before calling dbt to verify). - You can also set

DBT_PROFILES_DIR

to point dbt to your profiles directory if that’s easier than passing

profiles_dir

. Docs for reference - Prefect + dbt integration: Prefect dbt integration - Prefect Shell (alternative if you want explicit control over env via

ShellOperation(env=...)

): Prefect Shell integration - Prefect settings/profiles (for environment handling patterns): Prefect settings and profiles If you can share: - How you’re running this (local vs deployment), which worker type, and how you’re invoking the flow - A minimal snippet showing where you call

load_dotenv()

and

run_dbt_build(...)

- The dbt error output (e.g., “env var not found”) …I can pinpoint the exact fix.

don't judge the conditional logic too harshly for finding the dbt path or the seemingly random print statements. This is not even close to a half-baked solution 😆

fwiw - I was able to get this to run in a docker container with this deployment from yaml.

- name: Test Run DBT Build
    flow_name: Run DBT build
    entrypoint: flows/dbt_flow.py:run_dbt_build_flow
    work_pool: 
      name: local-machine-work-pool
      job_variables:
        env:
          database: "{{ $database }}"
          server: "{{ $server }}"
          user: "{{ $username }}"
          password: "{{ $password }}"
    schedule: null

Just wondering how to do this when invoking runs locally for quicker development

if you set the env vars in your terminal rather than using dotenv, does that work?