Pass cached tokens to RequestUsage

Author: sukeesh

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

@@ -86,6 +86,7 @@ class FunctionExecutionResultMessage(BaseModel):
 class RequestUsage:
     prompt_tokens: int
     completion_tokens: int
+    cached_tokens: int
 
 
 FinishReasons = Literal["stop", "length", "function_calls", "content_filter", "unknown"]

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

@@ -685,6 +685,7 @@ async def create(
         usage = RequestUsage(
             prompt_tokens=result.usage.input_tokens,
             completion_tokens=result.usage.output_tokens,
+            cached_tokens=result.usage.cache_read_input_tokens or 0,
         )
         serializable_messages: List[Dict[str, Any]] = [self._serialize_message(msg) for msg in anthropic_messages]
 

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

@@ -711,6 +711,7 @@ async def create(
             # TODO backup token counting
             prompt_tokens=getattr(result.usage, "prompt_tokens", 0) if result.usage is not None else 0,
             completion_tokens=getattr(result.usage, "completion_tokens", 0) if result.usage is not None else 0,
+            cached_tokens=getattr(result.usage.prompt_tokens_details, "cached_tokens", 0) if result.usage is not None else 0,
         )
 
         logger.info(