Pandera Models with FastAPI

[vilmar-hillow]

Is there any detailed documentation on the interactions of pandera Models with FastAPI?
For example, let's take this sample:

from fastapi import FastAPI
import pandera as pa

app = FastAPI()


class InputSpec(pa.DataFrameModel):
    some_int: pa.typing.Series[int] = pa.Field(ge=0, lt=100)
    some_float: pa.typing.Series[float] = pa.Field(ge=0, lt=100)

    class Config:
        coerce = True


@app.post("/path")
async def foo(
    request_body: pa.typing.DataFrame[InputSpec],
) -> pa.typing.DataFrame[InputSpec]:
    return request_body

Calling /path endpoint with {"some_int": [1], "some_float": [2.0]} payload will return a 500 due to serialization issues: ValueError: [TypeError("'numpy.int64' object is not iterable"), TypeError('vars() argument must have __dict__ attribute')]
This can be circumvented by using a float instead of int in the spec, or specifying to_format = "dict" in the Config, but both behaviours are not expected.

I'm using pandera 0.16.1 and fastapi 0.99.1 (0.100 and higher seem to conflict with pandera)