Hi everyone! Is it possible to define a deployment to have an unspecified number of parameters? Basically the same like you would define such a function in python

def fcn(**kwargs)

. Is it possible?

I think I would reach for a pydantic model here what are you trying to do?

I would like to have a flexible way to input parameters in order to avoid having to change multiple dags when I need to add a new parameter. So that every flow has the same interface

> I would like each flow exposing one single parameter which is a dict > dict[str, anything] if this is the case, why not just do

@flow
def f(data: dict[str, Any]):

I tried this but it does not work Any from

from typing import Any

?

yea using

Any

isnt exactly true bc the values would need to be JSON serializable I'd be inclined to do this

from pydantic import BaseModel

from prefect import flow


class Inputs(BaseModel):
    a: int
    b: str
    c: bool
    d: dict[str, str | dict[str, str]]


@flow
def something(data: Inputs):
    print(data)


something({"a": 1, "b": "2", "c": True, "d": {"e": "f", "g": {"h": "i"}}})

because you'd be free to change the impl of

Inputs

as needed, without changing the signature of your functions that use it, and then you get the nice form

I want to avoid to list all the parameters because I want to have parameter like

{"a": {"b": {"c": 0, "d": 1}}}

you can, this is just a matter of writing a pydantic model that suits your use case if you want to default certain things, then set defaults so users dont have to provide them

from pydantic import BaseModel, Field

from prefect import flow


class Inputs(BaseModel):
    a: int
    b: str
    c: bool
    d: dict[str, str | dict[str, str]] = Field(default_factory=dict)


@flow
def something(data: Inputs):
    print(data)


something({"a": 1, "b": "2", "c": True})

or if you want to allow arbitrary extras, you can

from pydantic import BaseModel

from prefect import flow


class Inputs(BaseModel, extra="allow"):
    a: int


@flow
def something(data: Inputs):
    print(data)


something({"a": 1, "b": "2", "c": True})

https://docs.pydantic.dev/latest/concepts/models/

would this allow me to have multiple nested dicts?

yes! i would encourage you to read the pydantic docs, you can have any json serializable type on your inputs model

great! I would give it a read, thank you Nate!

👍