BentoML

Hey <@U055ZT013PD>, welcome to the community. :congaparrot:
I'm still learning use of bentoml, I think internally bentoml has validation around the inputs and output, it must not support primitive types of the box.

Try something like this as a workaround till I can dig more and find out what the problem is

```class Model(BaseModel):
    __root__: list[str]

@svc.api(input=JSON(pydantic_model=Model), output=NumpyNdarray())
def predict_bentoml(input_data: Model) -&gt; np.ndarray:
    pass```

From Inference API definition looks like the input and output defined needs to be type of `IODescriptor` , which is why I think using primitive type might not work.
```class InferenceAPI:
    def __init__(
        self,
        user_defined_callback: t.Callable[..., t.Any] | None,
        input_descriptor: IODescriptor[t.Any],
        output_descriptor: IODescriptor[t.Any],
        name: Optional[str],
        doc: Optional[str] = None,
        route: Optional[str] = None,
    ):```