Allow documenting alternative response content-type

[alechoey]

I'm interested in using quart-schema as a drop-in replacement for apiflask to generate OpenAPI documentation for my application.

So far, I've been able to use validate_response to generate relevant documentation for all of my application/json endpoints.

When it comes to my other content-types, such as text/event-stream, I've been able to successfully document the headers with Pydantic models with the signature:

from pydantic import BaseModel, Field, ConfigDict, RootModel

def to_kebab(field_name: str) -> str:
    return field_name.replace("_", "-")

class StreamHeaders(BaseModel):
    model_config = ConfigDict(alias_generator=to_kebab)
    content_type: Literal["text/event-stream"] = Field(default="text/event-stream")

class StreamResponse(RootModel):
    root: bytes

@app.route("/")
@validate_response(StreamResponse, 200, StreamHeaders)
def index():
    async def async_generator():
        yield "hello"
        yield "world"
    return async_generator(), 200, StreamHeaders().model_dump(by_alias=True)

However, this yields me the following OpenAPI docs:

/:
  get:
    operationId: get_index
    parameters: []
    ...
    responses:
      '200':
        content:
          application/json:
            schema:
              format: binary
              title: StreamResponse
              type: string
          headers:
            content-type:
              schema:
                const: text/event-stream
                default: text/event-stream
                title: Content-Type
        description: ''
      ...

Instead of documenting the response as application/json, I would like it to use the relevant literal from the Headers class that I've provided. I've added a simple test case to demonstrate.

[pgjones]

I'm going to give this more consideration. I think Quart-Schema might benefit from a way to consider alternative/multiple content-types in a general sense.