An unexpected model is generated when server returns dictionary

首先复现带有 response_model=Dict[str, CliponMetadata] 的 FastAPI 路由 get_clipon_metadata_by_clip_name_list，并检查生成的响应模型和解析后的 additional_properties。将其与 List[CliponMetadata] 响应情况进行比较。当 dictionary 响应生成预期的 CliponMetadata 模型，且没有不需要的长名称 wrapper 模型时，即表示完成。

Describe the bug
I have a web API implementer with fastAPI. I have a function that make a POST request to the server I work with. This function returns dictionary that maps string to an object called "Clipon".

@router.get("/clipon/{version}", tags=["clipon"], status_code=status.HTTP_200_OK, response_model=Dict[str, CliponMetadata])
async def get_clipon_metadata_by_clip_name_list(version: str, clip_list: List[str]):
    try:
        res = await db_client.get_by_clip_name_list(clip_list=clip_list, version=version)
    except AssertionError as ex:
        raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail=str(ex))
    return {clip.clip_name: clip for clip in res}

When generating the models with openapi-python-client, what is generated is a model names "clipon_metadata" (as expected), and also another unexpected model with a long name:

It's full name is "GetCliponMetadataByClipNameListCliponVersionPostResponseGetCliponMetadataByClipNameListCliponVersionPost".

I don't understand why this model is created. For example, when I change this function to return List[CliponMetadata], only the "clipon_metadata" model is created.

In this behavior, when the client make this POST request, what it gets is A response object (also generated automatically), which in it's "parsed" attribute contains the unwanted object, and the "additional_properties" attribute of "parsed" holds the desired dictionary:

Expected behavior
I'd like that only the "clipon"metadata" model will be generated.

Desktop:

• OS: Linux (Ubuntu)
• Python Version: 3.9.1
• openapi-python-client version: 0.11.1