A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 13, 17:00 UTC
[Chore] eliminate duplicated and unconditional object serialization in anthropic messages api (#27792)
Signed-off-by: Vico Chu <vico24826@gmail.com>
Vico Chu committed
Nov 7, 2025 at 03:09 UTC
d4aa65c9980fb9287bfc2567a3b0c3efaae68255
2 files changed
+7
-6
vllm/entrypoints/anthropic/serving_messages.py
+4
-2
@@ -231,9 +231,11 @@ class AnthropicServingMessages(OpenAIServingChat):
231
See https://docs.anthropic.com/en/api/messages
232
for the API specification. This API mimics the Anthropic messages API.
233
"""
234
- logger.debug("Received messages request %s", request.model_dump_json())
234
+ if logger.isEnabledFor(logging.DEBUG):
235
+ logger.debug("Received messages request %s", request.model_dump_json())
236
chat_req = self._convert_anthropic_to_openai_request(request)
236
- logger.debug("Convert to OpenAI request %s", request.model_dump_json())
237
+ if logger.isEnabledFor(logging.DEBUG):
238
+ logger.debug("Convert to OpenAI request %s", chat_req.model_dump_json())
239
generator = await self.create_chat_completion(chat_req, raw_request)
240
241
if isinstance(generator, ErrorResponse):
vllm/entrypoints/openai/api_server.py
+3
-4
@@ -648,10 +648,9 @@ async def create_messages(request: AnthropicMessagesRequest, raw_request: Reques
648
return translate_error_response(generator)
649
650
elif isinstance(generator, AnthropicMessagesResponse):
651
- logger.debug(
652
- "Anthropic Messages Response: %s", generator.model_dump(exclude_none=True)
653
- )
654
- return JSONResponse(content=generator.model_dump(exclude_none=True))
651
+ resp = generator.model_dump(exclude_none=True)
652
+ logger.debug("Anthropic Messages Response: %s", resp)
653
+ return JSONResponse(content=resp)
654
655
return StreamingResponse(content=generator, media_type="text/event-stream")
656