`ninja.streaming.SSE` and `JSONL` are not `Generic[T]` - mypy reports "type is not generic and not indexable"

[adonig]

Describe the bug

The streaming type containers introduced in v1.6 (ninja.streaming.SSE,
ninja.streaming.JSONL) support SSE[Schema] syntax at runtime via
__class_getitem__, but the underlying StreamFormat base class does not
inherit from typing.Generic. As a result, mypy (and any other type checker
that doesn't special-case __class_getitem__) rejects the documented usage:

error: The type "type[SSE]" is not generic and not indexable  [misc]

This forces users with strict mypy configurations to add # type: ignore[misc]
to every streaming endpoint, which weakens the type-safety story that
django-ninja otherwise has.

Versions

• django-ninja: 1.6.2
• mypy: 1.20.0
• Python: 3.14

Reproduction

from ninja import NinjaAPI, Schema
from ninja.streaming import SSE

api = NinjaAPI()


class Event(Schema):
    message: str


@api.get("/events", response=SSE[Event])  # mypy error here
async def stream(request):
    yield {"message": "hello"}

Run mypy on the file:

error: The type "type[SSE]" is not generic and not indexable  [misc]

Runtime works correctly - only static type checking fails.

Root cause

In ninja/streaming.py:

class StreamFormat:
    media_type: str

    def __class_getitem__(cls, item_type: type) -> "_StreamAlias":
        return _StreamAlias(cls, item_type)

__class_getitem__ enables the runtime syntax, but mypy only treats a class
as subscriptable when it inherits from typing.Generic[T] (or one of the
special-cased stdlib generics).

Suggested fix

Make StreamFormat generic over the item type. The change is small and
shouldn't affect runtime behavior:

from typing import Generic, TypeVar

T = TypeVar("T")


class StreamFormat(Generic[T]):
    media_type: str
    # ... rest unchanged

SSE and JSONL then inherit the generic parameter automatically, and
SSE[Event] becomes a valid parameterized type for mypy.

Happy to send a PR if this approach is acceptable.