@Marvin I am using a pydantic BaseModel for my flow parameters and it is deployed. I want to start a run of this deployment from the command line, and I supply the parameter

pydantic_param

, that is defined with pydantic. I get an error, saying my provided parameter "is not of type object". If I know the type that would be valid, is there a workaround for this?

Your issue with the pydantic parameter likely stems from Prefect's handling of pydantic models in flow parameters. In Prefect 3.x, which uses Pydantic v2, you might encounter compatibility issues if your parameters are not correctly serialized or if there's a mix of Pydantic v1 and v2 models. ### Workaround: 1. Serialization: Ensure that the parameter you are passing is serialized correctly. You might need to convert your pydantic model to a dictionary or JSON-compatible format before passing it as a parameter. 2. Use JSON or Dict: Convert your

pydantic_param

to a JSON string or dictionary before passing it to the command line. You can then parse this back into a pydantic model within your flow. 3. Ensure Compatibility: Make sure that you are not mixing Pydantic v1 and v2 models within your flow parameters. ### Example: Suppose your pydantic model is

MyModel

. You can convert it to a dictionary like so:

python
# Assuming `my_model_instance` is an instance of `MyModel`
param_dict = my_model_instance.dict()

When running your deployment from the command line, pass the serialized dictionary or JSON:

bash
prefect deployment run <deployment-name> --params '{"pydantic_param": {"key": "value"}}'

Then, within your flow, you can reconstruct the pydantic model:

python
from pydantic import BaseModel

class MyModel(BaseModel):
    key: str

@flow
def my_flow(pydantic_param: dict):
    model_instance = MyModel(**pydantic_param)
    # Use model_instance as needed

For more insights, you can refer to the Prefect 3.x documentation and ensure that your deployment configurations align with the requirements. If you continue to face issues, reviewing the specific error messages and ensuring that your parameter types are consistent with the expected input types would be beneficial.