Bring your towel and join one of the fastest growing data communities. Welcome to our second-generation open source orchestration platform, a completely rethought approach to dataflow automation.

Prefect Community

Hi there, I'm trying to understand how I can deploy flows without having flow dependencies installed in my CI/CD environment. I am baking my flow and dependencies into a docker image, and trying to register the flow metadata with Prefect Cloud, but Prefect is trying to load my flow from its entrypoint and giving this error:
```prefect.exceptions.ScriptError: Script at 'main.py' encountered an exception: ModuleNotFoundError("No module named '&lt;flow dependency&gt;'")
An exception occurred.```

My flow is defined in a `prefect.yaml` file as follows:
```name: data-eng-pipelines/geocode
prefect-version: 2.14.5

build: null

push: null

pull:
- prefect.deployments.steps.set_working_directory:
    directory: /opt/prefect/

deployments:
- name: geocode-deployment
  version: null
  tags: []
  description: null
  schedule: {}
  flow_name: null
  entrypoint: main.py:main
  parameters: {}
  work_pool:
    name: '{{ prefect.variables.workPool }}'
    work_queue_name: null
    job_variables:
      image: '{{ prefect.variables.image }}:sample-{{ prefect.variables.tag }}'```
I strongly wish to separate CI/CD from dependency install, if possible, because the goal of this project is to have various flows which have their own separately managed dependencies. Are there any recommended solutions?

at the moment there's no way around this that I know of. I always install the requirements for my flows in my CI runner just before deploying them

Fair enough, thanks for letting me know. This is a major pain point I've had since Prefect 1, and I had mistakenly thought it was resolved.

mind upvoting this issue? <https://github.com/PrefectHQ/prefect/issues/9512>

if you find it relevant

Thanks for the link, <@U058FSVMBCN>, added my $0.02.

This is how I get around Prefect loading unnecessary dependencies on a deploy:

In my CICD pipeline: (we use CircleCI)

```- run:
    name: Deploy the Prefect Flow
    command: |
      export ENVIRONMENT='PREFECT_DEPLOY'
      prefect --no-prompt deploy --name <<parameters.environment>>```
Then, in my `main.py` where the deploy does a partial against the function in the entry point to grab the metadata off the flow decorator, (pulling in all the imports) I wrap the module imports unnecessary to the deploy within a conditional so they're skipped only during deploy.

```if not getenv(
    'ENVIRONMENT'
) == 'PREFECT_DEPLOY':
    from dependent_module.app import main

@flow(
    name='best-flow-evar'
)
def run_flow():
    main()


if __name__ == "__main__":
    run_flow()```