Bug: Deserialization of list[Model[T]] not working in v5

[chrisgoddard]

Describe the bug
I just upgraded to v5 and all of my subscribers that consume from a redis list broke. I use a Pydantic model with a generic, kind of like this:

T = typing.TypeVar('T', bound=BaseModel | dict)
class Event(BaseModel, typing.Generic[T]):
    uid: str
    data: T

I now for a consumer like this:

@broker.subscriber(
    list=ListSub(
        QUEUE_OBJECTS_WRITE,
        batch=True,
        max_records=10000,
        polling_interval=30
    )
)
async def write_object_stream(
    events: list[Event[ObjectRecord]],
    ...

I get an error like this:

events.0
  Input should be a valid dictionary or instance of Event[ObjectRecord] [type=model_type, input_value='{"uid":"01HVMGPZC8S6RVP4...e.\\"}"},"metadata":{}}', input_type=str]
    For further information visit https://errors.pydantic.dev/2.7/v/model_type
events.1
  Input should be a valid dictionary or instance of Event[ObjectRecord] [type=model_type, input_value='{"uid":"01HVMGPZC812XW2C...false}"},"metadata":{}}', input_type=str]
    For further information visit https://errors.pydantic.dev/2.7/v/model_type
events.2
  Input should be a valid dictionary or instance of Event[ObjectRecord] [type=model_type, input_value='{"uid":"01HVMGPZCGBZ5EPW...l.\\"}"},"metadata":{}}', input_type=str]
    For further information visit https://errors.pydantic.dev/2.7/v/model_type
events.3

The messages are coming through as strings so it's not correctly parsing the JSON before passing them to the Pydantic model.

Has anything changed in the way batch consumers are configured?

I've had to roll back to 0.4.7 but I'd quite like to leverage the StopApplication exception so I'd love to resolve this.

Thanks!

[Lancetnik]

Sorry, our tests were missed this behavior for some reason. Now it is fixed and tests are added. Wait for patch a bit.

[chrisgoddard]

All good! Thanks for the quick turnaround!

[chrisgoddard]

Confirm it's working as expected.