metrics: support customer labels specified in request header (#10143)
This commit is contained in:
@@ -96,6 +96,7 @@ class OpenAIServingChat(OpenAIServingBase):
|
||||
def _convert_to_internal_request(
|
||||
self,
|
||||
request: ChatCompletionRequest,
|
||||
raw_request: Request = None,
|
||||
) -> tuple[GenerateReqInput, ChatCompletionRequest]:
|
||||
reasoning_effort = (
|
||||
request.chat_template_kwargs.pop("reasoning_effort", None)
|
||||
@@ -127,6 +128,9 @@ class OpenAIServingChat(OpenAIServingBase):
|
||||
else:
|
||||
prompt_kwargs = {"input_ids": processed_messages.prompt_ids}
|
||||
|
||||
# Extract customer labels from raw request headers
|
||||
customer_labels = self.extract_customer_labels(raw_request)
|
||||
|
||||
adapted_request = GenerateReqInput(
|
||||
**prompt_kwargs,
|
||||
image_data=processed_messages.image_data,
|
||||
@@ -145,6 +149,7 @@ class OpenAIServingChat(OpenAIServingBase):
|
||||
bootstrap_room=request.bootstrap_room,
|
||||
return_hidden_states=request.return_hidden_states,
|
||||
rid=request.rid,
|
||||
customer_labels=customer_labels,
|
||||
)
|
||||
|
||||
return adapted_request, request
|
||||
|
||||
Reference in New Issue
Block a user