Add token usage to messages

python/packages/autogen-agentchat/src/autogen_agentchat/agents/_assistant_agent.py

@@ -266,8 +266,8 @@ async def on_messages_stream(
         while isinstance(result.content, list) and all(isinstance(item, FunctionCall) for item in result.content):
             event_logger.debug(ToolCallEvent(tool_calls=result.content, source=self.name))
             # Add the tool call message to the output.
-            inner_messages.append(ToolCallMessage(content=result.content, source=self.name))
-            yield ToolCallMessage(content=result.content, source=self.name)
+            inner_messages.append(ToolCallMessage(content=result.content, source=self.name, model_usage=result.usage))
+            yield ToolCallMessage(content=result.content, source=self.name, model_usage=result.usage)
 
             # Execute the tool calls.
             results = await asyncio.gather(
@@ -303,7 +303,8 @@ async def on_messages_stream(
 
         assert isinstance(result.content, str)
         yield Response(
-            chat_message=TextMessage(content=result.content, source=self.name), inner_messages=inner_messages
+            chat_message=TextMessage(content=result.content, source=self.name, model_usage=result.usage),
+            inner_messages=inner_messages,
         )
 
     async def _execute_tool_call(

python/packages/autogen-agentchat/src/autogen_agentchat/messages.py

@@ -1,7 +1,7 @@
 from typing import List
 
 from autogen_core.components import FunctionCall, Image
-from autogen_core.components.models import FunctionExecutionResult
+from autogen_core.components.models import FunctionExecutionResult, RequestUsage
 from pydantic import BaseModel
 
 
@@ -11,6 +11,9 @@ class BaseMessage(BaseModel):
     source: str
     """The name of the agent that sent this message."""
 
+    model_usage: RequestUsage | None = None
+    """The model client usage incurred when producing this message."""
+
 
 class TextMessage(BaseMessage):
     """A text message."""

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

@@ -78,7 +78,7 @@ async def test_run_with_tools(monkeypatch: pytest.MonkeyPatch) -> None:
             created=0,
             model=model,
             object="chat.completion",
-            usage=CompletionUsage(prompt_tokens=0, completion_tokens=0, total_tokens=0),
+            usage=CompletionUsage(prompt_tokens=10, completion_tokens=5, total_tokens=0),
         ),
         ChatCompletion(
             id="id2",
@@ -88,7 +88,7 @@ async def test_run_with_tools(monkeypatch: pytest.MonkeyPatch) -> None:
             created=0,
             model=model,
             object="chat.completion",
-            usage=CompletionUsage(prompt_tokens=0, completion_tokens=0, total_tokens=0),
+            usage=CompletionUsage(prompt_tokens=10, completion_tokens=5, total_tokens=0),
         ),
         ChatCompletion(
             id="id2",
@@ -100,7 +100,7 @@ async def test_run_with_tools(monkeypatch: pytest.MonkeyPatch) -> None:
             created=0,
             model=model,
             object="chat.completion",
-            usage=CompletionUsage(prompt_tokens=0, completion_tokens=0, total_tokens=0),
+            usage=CompletionUsage(prompt_tokens=10, completion_tokens=5, total_tokens=0),
         ),
     ]
     mock = _MockChatCompletion(chat_completions)
@@ -113,9 +113,17 @@ async def test_run_with_tools(monkeypatch: pytest.MonkeyPatch) -> None:
     result = await tool_use_agent.run("task")
     assert len(result.messages) == 4
     assert isinstance(result.messages[0], TextMessage)
+    assert result.messages[0].model_usage is None
     assert isinstance(result.messages[1], ToolCallMessage)
+    assert result.messages[1].model_usage is not None
+    assert result.messages[1].model_usage.completion_tokens == 5
+    assert result.messages[1].model_usage.prompt_tokens == 10
     assert isinstance(result.messages[2], ToolCallResultMessage)
+    assert result.messages[2].model_usage is None
     assert isinstance(result.messages[3], TextMessage)
+    assert result.messages[3].model_usage is not None
+    assert result.messages[3].model_usage.completion_tokens == 5
+    assert result.messages[3].model_usage.prompt_tokens == 10
 
     # Test streaming.
     mock._curr_index = 0  # pyright: ignore
@@ -158,7 +166,7 @@ async def test_handoffs(monkeypatch: pytest.MonkeyPatch) -> None:
             created=0,
             model=model,
             object="chat.completion",
-            usage=CompletionUsage(prompt_tokens=0, completion_tokens=0, total_tokens=0),
+            usage=CompletionUsage(prompt_tokens=42, completion_tokens=43, total_tokens=85),
         ),
     ]
     mock = _MockChatCompletion(chat_completions)
@@ -173,9 +181,17 @@ async def test_handoffs(monkeypatch: pytest.MonkeyPatch) -> None:
     result = await tool_use_agent.run("task")
     assert len(result.messages) == 4
     assert isinstance(result.messages[0], TextMessage)
+    assert result.messages[0].model_usage is None
     assert isinstance(result.messages[1], ToolCallMessage)
+    assert result.messages[1].model_usage is not None
+    assert result.messages[1].model_usage.completion_tokens == 43
+    assert result.messages[1].model_usage.prompt_tokens == 42
     assert isinstance(result.messages[2], ToolCallResultMessage)
+    assert result.messages[2].model_usage is None
     assert isinstance(result.messages[3], HandoffMessage)
+    assert result.messages[3].content == handoff.message
+    assert result.messages[3].target == handoff.target
+    assert result.messages[3].model_usage is None
 
     # Test streaming.
     mock._curr_index = 0  # pyright: ignore