Thanks for sharing the snippet — the error usually means the deploy() step is trying to package your source directory but can’t read the path you passed to

from_source

. What’s happening - In Prefect 3.x,

flow.from_source(...).deploy(...)

packages the directory you give as

source

and uploads it as part of the deployment. If that path isn’t readable from the process running the script, you’ll see “[Errno 2] No such file or directory”. - Your

source=str(Path(__file__).parent)

is resolving to

/mnt/c/Users/huyao/Documents/github/qcsched/prefect

, which is a WSL path. If you’re running the script with Windows Python (outside WSL), that path won’t exist. Conversely, if you’re in WSL but the directory actually isn’t present/mounted as expected, you’ll see the same error. Quick checks - Run this to confirm what

source

resolves to and whether it exists at runtime:

from pathlib import Path
print(Path(__file__).parent)
print(Path(__file__).parent.exists())

- Also share the output of:

prefect version
uname -a  # if in WSL

Ways to fix 1) Make the path and runtime match - If you’re running the script in WSL, ensure the repo is accessible under that path (consider moving the repo into your WSL home, e.g.

/home/<you>/repo

, to avoid Windows mount nuances). - If you’re running in Windows Python, your

Path(__file__).parent

should resolve to something like `C:\Users\...`; if you’re seeing

/mnt/c/...

, the script is likely running under WSL. 2) Prefer Git-based source (recommended for portability) - This avoids local path issues entirely:

from pathlib import Path
from prefect import flow

@flow(log_prints=True)
def my_flow(name: str = "world"):
    print(f"Hello {name}! I'm a flow from a Python script!")

if __name__ == "__main__":
    my_flow.from_source(
        source="<https://github.com/<owner>/<repo>>",
        entrypoint="prefect/flow_cloud.py:my_flow",  # path within the repo
    ).deploy(
        name="my-deployment",
        parameters={"name": "Marvin"},
        work_pool_name="test",
    )

Notes and docs - In Prefect 3.x,

from_source

supports local directories and Git URLs;

deploy()

packages and uploads your flow code so workers can fetch it. See Store your flow code. - General 3.x deployment info: Release notes and Deployment examples. If you can paste: - the exact value printed by

Path(__file__).parent

- whether

exists()

returns True - whether you’re running the script from Windows Python or inside WSL …I can give you a precise next step.

@Marvin the output of Path(file).parent is /mnt/c/Users/huyao/Documents/github/qcsched/prefect,

exists()

returns True, prefect version is 3.4.22, and the output of uname -a is Linux smallcat-pc 5.15.167.4-microsoft-standard-WSL2 #1 SMP Tue Nov 5 002155 UTC 2024 x86_64 x86_64 x86_64 GNU/Linux

Thanks for the details — super helpful. Two likely causes and fixes: 1) Your project folder is named “prefect” - Your script lives under …/qcsched/prefect, which collides with the installed Prefect package name. This can confuse import/resource resolution during packaging and surface as a FileNotFoundError on that directory. - Fix: rename that folder to something else (e.g., qcsched/flows or qcsched/prefect_flows) and update your entrypoint accordingly, then retry. - Quick sanity check:

python -c "import prefect, sys; print('prefect package file:', prefect.__file__); print('sys.path[0]:', sys.path[0])"

You want prefect.file to point to site-packages, not your repo. 2) WSL path/mount quirks under /mnt/c - Although

exists()

is True, packaging sometimes stumbles on Windows-mounted paths. Moving the repo into your WSL home usually resolves this: - Move to e.g.

/home/<you>/qcsched

and update

source

to that path (or use

Path(__file__).parent.resolve()

from the new location). - Alternatively, avoid local-path packaging by using a Git URL for `source`:

from prefect import flow

@flow(log_prints=True)
def my_flow(name: str = "world"):
    print(f"Hello {name}! I'm a flow from a Python script!")

if __name__ == "__main__":
    my_flow.from_source(
        source="<https://github.com/<owner>/<repo>>",
        entrypoint="path/in/repo/flow_cloud.py:my_flow",
    ).deploy(
        name="my-deployment",
        parameters={"name": "Marvin"},
        work_pool_name="test",
    )

Also try resolving to an absolute path explicitly if you keep local storage:

source=str(Path(__file__).parent.resolve())

Docs for context: Store your flow code If the error persists after renaming the folder and/or moving into your WSL home, please paste the full traceback from the deploy call — that will let me pinpoint exactly where it’s failing.