[Bugfix] - Add Trace Headers to Beam Search Path (vllm-project#29100)

Signed-off-by: dsuhinin <suhinin.dmitriy@gmail.com>

Author: dsuhinin

vllm/entrypoints/openai/serving_chat.py

@@ -319,6 +319,7 @@ async def create_chat_completion(
                         request_id=request_id,
                         params=sampling_params,
                         lora_request=lora_request,
+                        trace_headers=trace_headers,
                     )
                 else:
                     engine_request, tokenization_kwargs = await self._process_inputs(

vllm/entrypoints/openai/serving_completion.py

@@ -216,6 +216,7 @@ async def create_completion(
                         request_id=request_id,
                         params=sampling_params,
                         lora_request=lora_request,
+                        trace_headers=trace_headers,
                     )
                 else:
                     engine_request, tokenization_kwargs = await self._process_inputs(

vllm/entrypoints/openai/serving_engine.py

@@ -343,6 +343,7 @@ async def beam_search(
         request_id: str,
         params: BeamSearchParams,
         lora_request: LoRARequest | None = None,
+        trace_headers: Mapping[str, str] | None = None,
     ) -> AsyncGenerator[RequestOutput, None]:
         beam_width = params.beam_width
         max_tokens = params.max_tokens
@@ -437,6 +438,7 @@ async def beam_search(
                             beam_search_params,
                             request_id_item,
                             lora_request=lora_req,
+                            trace_headers=trace_headers,
                         )
                     )
                 )