@Marvin - if my code files are under a sub-directory in the git repo. How should I make sure prefect-deployment is able to see all the files in that sub-directory? Do I need to include certain parameters in "flow.from_source"

To ensure that Prefect can see all files in a sub-directory when deploying, you should specify the correct path in the

from_source

method. Here's a concise approach: 1. Use

from_source

with `GitRepository`: If your code is in a sub-directory of a Git repository, you can use the

GitRepository

storage option to specify the repository URL and the entry point to your flow. 2. Specify the Entry Point: The

entrypoint

parameter should include the relative path to your flow within the repository. For example, if your flow is located in a sub-directory, you might specify it as

"subdir/your_flow.py:flow_function"

. 3. Using Local Source: If deploying from a local directory, ensure you provide the correct path to the sub-directory where your flow resides. Here's an example:

python
from pathlib import Path
from prefect import flow

@flow
def my_flow():
    pass

if __name__ == "__main__":
    my_flow.from_source(
        source=str(Path(__file__).parent / "your_sub_directory"),
        entrypoint="your_sub_directory/your_flow.py:my_flow"
    ).deploy()

For more details, you can refer to the Prefect documentation on deployment configurations.

hi @marqeu you can just put your url to your repo as source and then the relative path to your function from repo root, like from_source( source=https://github.com/org/repo.git entrypoint=path/to/dir/file.py:fn )

repo is private and it works fine if I move the files to parent dir; just fails if I move the files in a sub-dir weird.

oh another gotcha that i’ve ran into sometimes is i change the structure locally and then forget to push my code, and then i accidentally expect to pull what i have instead of what’s in the repo

i checked all that - seems OK. code works fine if all files are in parent but fails if I move the starting point under sub-dir "db_inverta"

hrm not sure. if you can share the actual trace i can probably look at it later