Hi there! I'm testing out

pydantic>2

(version

2.4.2

, to be exact) support released with

prefect==2.13.7

. Since the upgrade, we this error while deploying:

File "<snip>/.venv/lib/python3.11/site-packages/pydantic/v1/schema.py", line 996, in encode_default
    return pydantic_encoder(dft)
           ^^^^^^^^^^^^^^^^^^^^^
  File "<snip>/.venv/lib/python3.11/site-packages/pydantic/v1/json.py", line 90, in pydantic_encoder
    raise TypeError(f"Object of type '{obj.__class__.__name__}' is not JSON serializable")
TypeError: Object of type 'ClassThatInheritsFromBaseModel' is not JSON serializable

This flow always worked before, and still works when downgrading with

pip install 'pydantic<2'

. Is this something that breaks on the

prefect

end or the

pydantic

end? The context is that we have flow definitions that look like this:

@flow
def my_flow(complex_argument: ClassThatInheritsFromBaseModel = ClassThatInheritsFromBaseModel(some_value="a-sane-default")):
    ...

Just calling

ClassThatInheritsFromBaseModel(...).dict()

on all of the default args works, but adds clutter and makes type checkers unhappy.

Hi @Richard Sundqvist! 👋 ! I’m keeping an eye on pydantic issues as they come up, so I’ll try to reproduce on my end and see if I can provide any clarity whether it’s a prefect or a pydantic issue.

@Alexander Azzam I created an issue for this: https://github.com/PrefectHQ/prefect/issues/10965 Repro is trivial, at least with a server running 🙂. Probably without, too.

I went into the same issue trying to prefect deploy a flow with ShellOperation.

pip install 'pydantic<2'

worked.

Hey @Richard Sundqvist I believe this has finally been addressed in today's release (https://prefect-community.slack.com/archives/CKNSX5WG3/p1698965320644549) please let us know if you still see this issue coming up

I've been following urimandujano's PR in fact 🙂. Quite a quick turnaround considering the number of curve balls that seem to have been involved. Please extend our thanks to everyone who worked on this!

Such a pleasure. We’re all grateful to you for taking the time to surface and leave an issue.