Add vllm id to the response

Signed-off-by: Mert Toslali <[email protected]>

Author: toslali-ibm

src/guidellm/backends/response_handlers.py

@@ -145,6 +145,7 @@ def compile_non_streaming(
             request_args=str(
                 request.arguments.model_dump() if request.arguments else None
             ),
+            response_id=response.get("id"), # use vLLM ID if available
             text=text,
             input_metrics=input_metrics,
             output_metrics=output_metrics,

src/guidellm/schemas/response.py

@@ -41,6 +41,10 @@ class GenerationResponse(StandardBaseModel):
     request_id: str = Field(
         description="Unique identifier matching the original GenerationRequest."
     )
+    response_id: str | None = Field(
+        default=None,
+        description="Unique identifier matching the original vLLM Response ID."
+    )
     request_args: str | None = Field(
         description="Arguments passed to the backend for request processing."
     )