Implements suggestion discussed in issue #6049 by adding raw_response field to CreateResult.

python/packages/autogen-agentchat/tests/test_assistant_agent.py

@@ -64,6 +64,7 @@ async def test_run_with_tools(monkeypatch: pytest.MonkeyPatch) -> None:
                 usage=RequestUsage(prompt_tokens=10, completion_tokens=5),
                 thought="Calling pass function",
                 cached=False,
+                raw_response={"id": "mock-id", "provider": "replay"},
             ),
             "pass",
             "TERMINATE",
@@ -142,18 +143,21 @@ async def test_run_with_tools_and_reflection() -> None:
                 content=[FunctionCall(id="1", arguments=json.dumps({"input": "task"}), name="_pass_function")],
                 usage=RequestUsage(prompt_tokens=10, completion_tokens=5),
                 cached=False,
+                raw_response={"id": "mock-id", "provider": "replay"},
             ),
             CreateResult(
                 finish_reason="stop",
                 content="Hello",
                 usage=RequestUsage(prompt_tokens=10, completion_tokens=5),
                 cached=False,
+                raw_response={"id": "mock-id", "provider": "replay"},
             ),
             CreateResult(
                 finish_reason="stop",
                 content="TERMINATE",
                 usage=RequestUsage(prompt_tokens=10, completion_tokens=5),
                 cached=False,
+                raw_response={"id": "mock-id", "provider": "replay"},
             ),
         ],
         model_info={
@@ -244,6 +248,7 @@ async def test_run_with_parallel_tools() -> None:
                 usage=RequestUsage(prompt_tokens=10, completion_tokens=5),
                 thought="Calling pass and echo functions",
                 cached=False,
+                raw_response={"id": "mock-id", "provider": "replay"},
             ),
             "pass",
             "TERMINATE",
@@ -329,6 +334,7 @@ async def test_run_with_parallel_tools_with_empty_call_ids() -> None:
                 ],
                 usage=RequestUsage(prompt_tokens=10, completion_tokens=5),
                 cached=False,
+                raw_response={"id": "mock-id", "provider": "replay"},
             ),
             "pass",
             "TERMINATE",
@@ -411,6 +417,7 @@ async def test_handoffs() -> None:
                 ],
                 usage=RequestUsage(prompt_tokens=42, completion_tokens=43),
                 cached=False,
+                raw_response={"id": "mock-id", "provider": "replay"},
             )
         ],
         model_info={
@@ -557,6 +564,7 @@ async def test_list_chat_messages(monkeypatch: pytest.MonkeyPatch) -> None:
                 content="Response to message 1",
                 usage=RequestUsage(prompt_tokens=10, completion_tokens=5),
                 cached=False,
+                raw_response={"id": "mock-id", "provider": "replay"},
             )
         ]
     )
@@ -759,6 +767,7 @@ async def test_model_client_stream_with_tool_calls() -> None:
                 finish_reason="function_calls",
                 usage=RequestUsage(prompt_tokens=10, completion_tokens=5),
                 cached=False,
+                raw_response={"id": "mock-id", "provider": "replay"},
             ),
             "Example response 2 to task",
         ]

python/packages/autogen-agentchat/tests/test_group_chat.py

@@ -255,6 +255,7 @@ async def test_round_robin_group_chat_with_tools(runtime: AgentRuntime | None) -
                 content=[FunctionCall(id="1", name="pass", arguments=json.dumps({"input": "pass"}))],
                 usage=RequestUsage(prompt_tokens=0, completion_tokens=0),
                 cached=False,
+                raw_response={"id": "mock-id", "provider": "replay"},
             ),
             "Hello",
             "TERMINATE",
@@ -848,6 +849,7 @@ async def test_swarm_handoff_using_tool_calls(runtime: AgentRuntime | None) -> N
                 content=[FunctionCall(id="1", name="handoff_to_agent2", arguments=json.dumps({}))],
                 usage=RequestUsage(prompt_tokens=0, completion_tokens=0),
                 cached=False,
+                raw_response={"id": "mock-id", "provider": "replay"},
             ),
             "Hello",
             "TERMINATE",
@@ -938,6 +940,7 @@ async def test_swarm_with_parallel_tool_calls(runtime: AgentRuntime | None) -> N
                 ],
                 usage=RequestUsage(prompt_tokens=0, completion_tokens=0),
                 cached=False,
+                raw_response={"id": "mock-id", "provider": "replay"},
             ),
             "Hello",
             "TERMINATE",

python/packages/autogen-core/src/autogen_core/models/_types.py

@@ -1,5 +1,5 @@
 from dataclasses import dataclass
-from typing import List, Literal, Optional, Union
+from typing import List, Literal, Optional, Union, Dict, Any
 
 from pydantic import BaseModel, Field
 from typing_extensions import Annotated
@@ -125,3 +125,8 @@ class CreateResult(BaseModel):
     thought: Optional[str] = None
     """The reasoning text for the completion if available. Used for reasoning models
     and additional text content besides function calls."""
+
+    raw_response: Optional[Dict[str, Any]] = None
+    """Raw response from the model API, useful for custom field access."""
+
+

python/packages/autogen-core/tests/test_tool_agent.py

@@ -113,13 +113,15 @@ async def create(
                     usage=RequestUsage(prompt_tokens=0, completion_tokens=0),
                     cached=False,
                     logprobs=None,
+                    raw_response={"id": "mock-id", "provider": "replay"},
                 )
             return CreateResult(
                 content="Done",
                 finish_reason="stop",
                 usage=RequestUsage(prompt_tokens=0, completion_tokens=0),
                 cached=False,
                 logprobs=None,
+                raw_response={"id": "mock-id", "provider": "replay"},
             )
 
         def create_stream(

python/packages/autogen-ext/src/autogen_ext/experimental/task_centric_memory/utils/chat_completion_client_recorder.py

@@ -142,6 +142,7 @@ async def create(
                 finish_reason=data.get("finish_reason", "stop"),
                 usage=data.get("usage", RequestUsage(prompt_tokens=0, completion_tokens=0)),
                 cached=True,
+                raw_response=data.get("raw_response", {"id": "mock-id", "provider": "replay"}),
             )
             return result
 

python/packages/autogen-ext/src/autogen_ext/models/anthropic/_anthropic_client.py

@@ -557,6 +557,7 @@ async def create(
             usage=usage,
             cached=False,
             thought=thought,
+            raw_response=result,
         )
 
         # Update usage statistics
@@ -775,6 +776,7 @@ async def create_stream(
             usage=usage,
             cached=False,
             thought=thought,
+            raw_response=result,
         )
 
         # Emit the end event.

python/packages/autogen-ext/src/autogen_ext/models/azure/_azure_ai_client.py

@@ -438,6 +438,7 @@ async def create(
             usage=usage,
             cached=False,
             thought=thought,
+            raw_response=result,
         )
 
         self.add_usage(usage)

python/packages/autogen-ext/src/autogen_ext/models/llama_cpp/_llama_cpp_completion_client.py

@@ -357,7 +357,7 @@ async def create(
         if not response_tool_calls and not response_text:
             logger.debug("DEBUG: No response text found. Returning empty response.")
             return CreateResult(
-                content="", usage=RequestUsage(prompt_tokens=0, completion_tokens=0), finish_reason="stop", cached=False
+                content="", usage=RequestUsage(prompt_tokens=0, completion_tokens=0), finish_reason="stop", cached=False, raw_response=response
             )
 
         # Create a CreateResult object
@@ -373,6 +373,7 @@ async def create(
             usage=cast(RequestUsage, response["usage"]),
             finish_reason=normalize_stop_reason(finish_reason),  # type: ignore
             cached=False,
+            raw_response=response,
         )
 
         # If we are running in the context of a handler we can get the agent_id

python/packages/autogen-ext/src/autogen_ext/models/ollama/_ollama_client.py

@@ -602,6 +602,7 @@ async def create(
             usage=usage,
             cached=False,
             logprobs=None,
+            raw_response=result,
         )
 
         self._total_usage = _add_usage(self._total_usage, usage)
@@ -737,6 +738,7 @@ async def create_stream(
             usage=usage,
             cached=False,
             logprobs=None,
+            raw_response=result,
         )
 
         # Emit the end event.

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

@@ -653,6 +653,7 @@ async def create(
             cached=False,
             logprobs=logprobs,
             thought=thought,
+            raw_response=result,
         )
 
         self._total_usage = _add_usage(self._total_usage, usage)
@@ -879,6 +880,7 @@ async def create_stream(
             cached=False,
             logprobs=logprobs,
             thought=thought,
+            raw_response=result,
         )
 
         # Log the end of the stream.

python/packages/autogen-ext/src/autogen_ext/models/replay/_replay_chat_completion_client.py

@@ -176,7 +176,7 @@ async def create(
             _, output_token_count = self._tokenize(response)
             self._cur_usage = RequestUsage(prompt_tokens=prompt_token_count, completion_tokens=output_token_count)
             response = CreateResult(
-                finish_reason="stop", content=response, usage=self._cur_usage, cached=self._cached_bool_value
+                finish_reason="stop", content=response, usage=self._cur_usage, cached=self._cached_bool_value, raw_response=response
             )
         else:
             self._cur_usage = RequestUsage(
@@ -221,7 +221,7 @@ async def create_stream(
                 else:
                     yield token
             yield CreateResult(
-                finish_reason="stop", content=response, usage=self._cur_usage, cached=self._cached_bool_value
+                finish_reason="stop", content=response, usage=self._cur_usage, cached=self._cached_bool_value, raw_response=response
             )
             self._update_total_usage()
         else:

python/packages/autogen-ext/src/autogen_ext/models/semantic_kernel/_sk_chat_completion_adapter.py

@@ -521,6 +521,7 @@ async def create(
             usage=RequestUsage(prompt_tokens=prompt_tokens, completion_tokens=completion_tokens),
             cached=False,
             thought=thought,
+            raw_response=result
         )
 
     @staticmethod
@@ -676,6 +677,7 @@ async def create_stream(
                         finish_reason="function_calls",
                         usage=RequestUsage(prompt_tokens=prompt_tokens, completion_tokens=completion_tokens),
                         cached=False,
+                        raw_response=result,
                     )
                     return
 
@@ -698,6 +700,7 @@ async def create_stream(
             usage=RequestUsage(prompt_tokens=prompt_tokens, completion_tokens=completion_tokens),
             cached=False,
             thought=thought,
+            raw_response=result,
         )
 
         # Emit the end event.