I am creating a simple REST API for a blog website. I'm writing an endpoint for creating posts, I'm using Forms to define the schema instead of Pydantic's models because one of the fields is a user uploaded image which is much easier to handle as a form. I'm trying to add a "Tags" system for posts where a post can have multiple tags, Each tag is an integer value that corresponds to that tag's ID in the DB:

Sending that request yields this response:

{
  "detail": [
    {
      "type": "int_parsing",
      "loc": [
        "body",
        "tags",
        0
      ],
      "msg": "Input should be a valid integer, unable to parse string as an integer",
      "input": "3,5"
    }
  ]
}

The request is processed normally if the tags array is empty or if it contains exactly 1 element, But 2 or more causes that parsing error.

Here's my endpoint's signature:

@router.post('/', response_model=PostRead, status_code=status.HTTP_201_CREATED)
def create_post(
    db: Annotated[Session, Depends(get_db)],
    author: Annotated[User, Depends(get_current_user)],
    title: Annotated[Optional[str], Form()],
    body: Annotated[Optional[str], Form()] = None,
    image_file: Annotated[Optional[UploadFile], File()] = None,
    tags: Annotated[List[int], Form()] = [],
):

Here's the schema from the openapi.json, it's using version 3.1.0:

Body_create_post_api_posts__post    
properties  
title   
anyOf   
0   
type    "string"
1   
type    "null"
title   "Title"
body    
anyOf   
0   
type    "string"
1   
type    "null"
title   "Body"
image_file  
anyOf   
0   
type    "string"
contentMediaType    "application/octet-stream"
1   
type    "null"
title   "Image File"
tags    
items   
type    "integer"
type    "array"
title   "Tags"
default []
type    "object"
required    
0   "title"
title   "Body_create_post_api_posts__post"

I tried changing the tags parameter annotation to these but nothing changed:

tags: List[int] = Form([])

tags: List[int] = Form(...)

tags: Annotated[List[int], Form(...)] = []

Also tried changing int to str.

Somebody experienced the same issue here.