pydantic base model in JSON output adaptor

[chlee1016]

Hi BentoML team! When I take a look at new feature in BentoML 1.0, found the pydantic model can be used in JSON I/O adaptor.

• https://docs.bentoml.org/en/latest/api/api_io_descriptors.html#bentoml.io.JSON

As I know, the Input can be checked whether it is valid using Custom pydantic model.
However, in case of the Output, even if I set the pydantic model, no action is taken.

I wonder how can the pydantic model work in JSON Output adaptor. Please let me know if I'm missing anything.

• BentoML/bentoml/_internal/io_descriptors/json.py

  Lines 171 to 180 in c6a50e6

  	async def to_http_response(self, obj: JSONType) -> Response:
  	json_str = json.dumps(
  	obj,
  	cls=self._json_encoder,
  	ensure_ascii=False,
  	allow_nan=False,
  	indent=None,
  	separators=(",", ":"),
  	)
  	return Response(json_str, media_type=MIME_TYPE_JSON)

• related issue : Add pydantic model schema #1480

[withsmilo]

Hi, @chlee1016 😄
I definitely think this is a bug. The JSON IO Descriptor should validate the output json data using a pydantic model for the outputs. Please contribute to solving this before releasing v1.0 if you have time.

[aarnphm]

Thanks for spotting this out. I will take a look at this