From fdfa0a1659a4a75de607219f9a7004d09928d572 Mon Sep 17 00:00:00 2001 From: Taneem Ibrahim Date: Mon, 7 Sep 2026 18:55:20 -0500 Subject: [PATCH] [Pooling] Honor request_id from request bodies (#55665) Signed-off-by: Taneem Ibrahim --- tests/entrypoints/pooling/token_embed/test_online.py | 3 +++ vllm/entrypoints/pooling/base/serving.py | 5 ++++- 2 files changed, 7 insertions(+), 1 deletion(-) diff --git a/tests/entrypoints/pooling/token_embed/test_online.py b/tests/entrypoints/pooling/token_embed/test_online.py index 42713ae852..2c807a5b52 100644 --- a/tests/entrypoints/pooling/token_embed/test_online.py +++ b/tests/entrypoints/pooling/token_embed/test_online.py @@ -54,6 +54,7 @@ def server(): @pytest.mark.parametrize("model_name", [MODEL_NAME]) async def test_pooling_token_embed(server: RemoteOpenAIServer, model_name: str): task = "token_embed" + request_id = "client-request-id" response = requests.post( server.url_for("pooling"), json={ @@ -61,11 +62,13 @@ async def test_pooling_token_embed(server: RemoteOpenAIServer, model_name: str): "input": input_text, "encoding_format": "float", "task": task, + "request_id": request_id, }, ) poolings = PoolingResponse.model_validate(response.json()) + assert poolings.id == f"pooling-{request_id}" assert len(poolings.data) == 1 assert len(poolings.data[0].data) == len(input_tokens) assert len(poolings.data[0].data[0]) == 384 diff --git a/vllm/entrypoints/pooling/base/serving.py b/vllm/entrypoints/pooling/base/serving.py index 0b9239e622..cfda3d6563 100644 --- a/vllm/entrypoints/pooling/base/serving.py +++ b/vllm/entrypoints/pooling/base/serving.py @@ -109,7 +109,10 @@ class PoolingBaseServing(ABC, BaseServing): raw_request: Request | None = None, ): model_name = self.models.model_name() - request_id = f"{self.request_id_prefix}-{self._base_request_id(raw_request)}" + base_request_id = self._base_request_id( + raw_request, getattr(request, "request_id", None) + ) + request_id = f"{self.request_id_prefix}-{base_request_id}" await self._check_model(request) pooling_params = io_processor.create_pooling_params(request)