[Bug] GEPAFeedbackMetric Protocol missing 'self' parameter causes type checker errors

[seehuhn]

What happened?

I am getting unexpected type checker errors for my GEPA objective functions:

voss@porkpie [~] uv run pyright repro.py
/Users/voss/sources/dspy-bug/repro.py
  /Users/voss/sources/dspy-bug/repro.py:9:12 - error: Argument of type "(gold: Unknown, pred: Unknown, trace: Unknown | None = None, pred_name: Unknown | None = None, pred_trace: Unknown | None = None) -> float" cannot be assigned to parameter "metric" of type "GEPAFeedbackMetric" in function "__init__"
    "FunctionType" is incompatible with protocol "GEPAFeedbackMetric"
      Could not bind method "__call__" because "FunctionType" is not assignable to parameter "gold"
        "FunctionType" is not assignable to "Example" (reportArgumentType)
1 error, 0 warnings, 0 informations

I believe that my function is correct, and the the error reported here is a false positive. Test file is below.

Steps to reproduce

View this test file in vscode, or run the type checker as shown above: repro.py.

Note that the problem does not occur for version 3.0.4. Sadly the reason for this is not that the problem has been fixed, but instead the new @experimental(version="3.0.0") decorator on GEPAFeedbackMetric (file "teleprompt/gepa/gepa.py") confuses the type checker.

DSPy version

3.0.3

[seehuhn]

I believe that the problem is that the Protocol is missing the self parameter.

Existing code in "teleprompt/gepa/gepa.py":

class GEPAFeedbackMetric(Protocol):
    def __call__(
        gold: Example,
        pred: Prediction,
        trace: Optional["DSPyTrace"],
        pred_name: str | None,
        pred_trace: Optional["DSPyTrace"],
    ) -> Union[float, "ScoreWithFeedback"]:
    ...

The examples shown in PEP 544 all seem to have the initial self argument in Protocol methods. I believe the correct fix is to add self as follows:

class GEPAFeedbackMetric(Protocol):
    def __call__(
        self,
        gold: Example,
        pred: Prediction,
        trace: Optional["DSPyTrace"],
        pred_name: str | None,
        pred_trace: Optional["DSPyTrace"],
    ) -> Union[float, "ScoreWithFeedback"]:
    ...