[Bug] generating response_format for JSONAdapter triggers an exception

[poudro]

What happened?

When generating the response format in JSONAdapter, an exception is systematically triggered and thus, under the hood, the response_format parameter sent is always {"type": "json_object"}

Steps to reproduce

Can be reproduced with

import dspy
from dspy.adapters.json_adapter import _get_structured_outputs_response_format

cot = dspy.Predict('question: str -> answer: str')
predictor = cot.predictors()[0]
response_format = _get_structured_outputs_response_format(predictor.signature)

Which triggers exception

Traceback (most recent call last):
  File "/Users/poudro/dspy/tt.py", line 6, in <module>
    response_format = _get_structured_outputs_response_format(predictor.signature)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/poudro/.pyenv/versions/dspy/lib/python3.12/site-packages/dspy/adapters/json_adapter.py", line 210, in _get_structured_outputs_response_format
    pydantic_model = pydantic.create_model(
                     ^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/poudro/.pyenv/versions/dspy/lib/python3.12/site-packages/pydantic/main.py", line 1761, in create_model
    return meta(
           ^^^^^
  File "/Users/poudro/.pyenv/versions/dspy/lib/python3.12/site-packages/pydantic/_internal/_model_construction.py", line 110, in __new__
    config_wrapper = ConfigWrapper.for_model(bases, namespace, kwargs)
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/poudro/.pyenv/versions/dspy/lib/python3.12/site-packages/pydantic/_internal/_config.py", line 138, in for_model
    config_new.update(config_from_namespace)
TypeError: 'type' object is not iterable

There is an easy fix

diff --git a/dspy/adapters/json_adapter.py b/dspy/adapters/json_adapter.py
index 018bd5f4..eb324e90 100644
--- a/dspy/adapters/json_adapter.py
+++ b/dspy/adapters/json_adapter.py
@@ -210,7 +210,7 @@ def _get_structured_outputs_response_format(signature: SignatureMeta) -> type[py
     pydantic_model = pydantic.create_model(
         "DSPyProgramOutputs",
         **fields,
-        __config__=type("Config", (), {"extra": "forbid"}),
+        __config__={"extra": "forbid"}
     )

     # Generate the initial schema.

DSPy version

2.6.24

[cbowdon]

I'm also struggling with this issue and would appreciate a fix.

[poudro]

it appears to have been fixed in latest iteration of JSONAdapter (see 63020fd), so next release should fix this