Fix extra_body silently dropped during load_component()

python/packages/autogen-ext/src/autogen_ext/models/openai/config/__init__.py

@@ -1,4 +1,4 @@
-from typing import Awaitable, Callable, Dict, List, Literal, Optional, Union
+from typing import Any, Awaitable, Callable, Dict, List, Literal, Optional, Union
 
 from autogen_core import ComponentModel
 from autogen_core.models import ModelCapabilities, ModelInfo  # type: ignore
@@ -57,6 +57,9 @@ class CreateArguments(TypedDict, total=False):
     - 'low': Faster responses with less reasoning
     - 'medium': Balanced reasoning and speed
     - 'high': More thorough reasoning, may take longer"""
+    extra_body: Optional[Dict[str, Any]]
+    """Extra JSON body fields to include in the API request.
+    Useful for provider-specific parameters (e.g., ``enable_thinking``)."""
 
 
 AsyncAzureADTokenProvider = Callable[[], Union[str, Awaitable[str]]]
@@ -108,6 +111,8 @@ class CreateArgumentsConfigModel(BaseModel):
     parallel_tool_calls: bool | None = None
     # Controls the amount of effort the model uses for reasoning (reasoning models only)
     reasoning_effort: Literal["minimal", "low", "medium", "high"] | None = None
+    # Extra JSON body fields to include in the API request (e.g., enable_thinking)
+    extra_body: Dict[str, Any] | None = None
 
 
 class BaseOpenAIClientConfigurationConfigModel(CreateArgumentsConfigModel):

python/packages/autogen-ext/tests/models/test_openai_model_client.py

@@ -3378,3 +3378,50 @@ async def test_reasoning_effort_validation() -> None:
         }
 
         ChatCompletionClient.load_component(config)
+
+
+@pytest.mark.asyncio
+async def test_extra_body_load_component() -> None:
+    """Test that extra_body survives load_component() round-trip (issue #7418)."""
+    from autogen_core.models import ChatCompletionClient
+
+    # Direct instantiation should work (baseline)
+    client = OpenAIChatCompletionClient(
+        model="gpt-4o",
+        api_key="fake_key",
+        extra_body={"enable_thinking": False},
+    )
+    assert client._create_args["extra_body"] == {"enable_thinking": False}  # pyright: ignore[reportPrivateUsage]
+
+    # load_component should preserve extra_body
+    openai_config = {
+        "provider": "OpenAIChatCompletionClient",
+        "config": {
+            "model": "gpt-4o",
+            "api_key": "fake_key",
+            "extra_body": {"enable_thinking": False},
+        },
+    }
+    loaded_client = ChatCompletionClient.load_component(openai_config)
+    assert loaded_client._create_args["extra_body"] == {"enable_thinking": False}  # type: ignore[attr-defined] # pyright: ignore[reportPrivateUsage, reportUnknownMemberType, reportAttributeAccessIssue]
+    assert loaded_client._raw_config["extra_body"] == {"enable_thinking": False}  # type: ignore[attr-defined] # pyright: ignore[reportPrivateUsage, reportUnknownMemberType, reportAttributeAccessIssue]
+
+    # dump_component -> load_component round-trip
+    config_dict = client.dump_component()
+    reloaded_client = OpenAIChatCompletionClient.load_component(config_dict)
+    assert reloaded_client._create_args["extra_body"] == {"enable_thinking": False}  # pyright: ignore[reportPrivateUsage]
+
+    # Azure variant
+    azure_config = {
+        "provider": "AzureOpenAIChatCompletionClient",
+        "config": {
+            "model": "gpt-4o",
+            "azure_endpoint": "https://fake.openai.azure.com",
+            "azure_deployment": "gpt-4o",
+            "api_version": "2024-06-01",
+            "api_key": "fake_key",
+            "extra_body": {"enable_thinking": False},
+        },
+    }
+    loaded_azure = ChatCompletionClient.load_component(azure_config)
+    assert loaded_azure._create_args["extra_body"] == {"enable_thinking": False}  # type: ignore[attr-defined] # pyright: ignore[reportPrivateUsage, reportUnknownMemberType, reportAttributeAccessIssue]