From 60b77e146315f4f0f23b9aa83c07d51f38927a1e Mon Sep 17 00:00:00 2001
From: Sanghoon Yoon <128017610+sanghoon-yn@users.noreply.github.com>
Date: Tue, 13 Jan 2026 12:21:49 +0900
Subject: [PATCH] [Frontend] Add `reasoning_effort` to
 `OpenAIServing._preprocess_chat()` (#31956)

Signed-off-by: Sanghoon Yoon <seanyoon@kakao.com>
---
 vllm/entrypoints/openai/serving_chat.py      | 6 +++++-
 vllm/entrypoints/openai/serving_responses.py | 8 ++++++++
 2 files changed, 13 insertions(+), 1 deletion(-)

diff --git a/vllm/entrypoints/openai/serving_chat.py b/vllm/entrypoints/openai/serving_chat.py
index b9a7b07b0..e65dba2b8 100644
--- a/vllm/entrypoints/openai/serving_chat.py
+++ b/vllm/entrypoints/openai/serving_chat.py
@@ -306,6 +306,10 @@ class OpenAIServingChat(OpenAIServing):
                 )
                 if error_check_ret is not None:
                     return error_check_ret
+
+                chat_template_kwargs = request.chat_template_kwargs or {}
+                chat_template_kwargs.update(reasoning_effort=request.reasoning_effort)
+
                 conversation, engine_prompts = await self._preprocess_chat(
                     request,
                     tokenizer,
@@ -316,7 +320,7 @@ class OpenAIServingChat(OpenAIServing):
                     continue_final_message=request.continue_final_message,
                     tool_dicts=tool_dicts,
                     documents=request.documents,
-                    chat_template_kwargs=request.chat_template_kwargs,
+                    chat_template_kwargs=chat_template_kwargs,
                     default_chat_template_kwargs=self.default_chat_template_kwargs,
                     tool_parser=tool_parser,
                     add_special_tokens=request.add_special_tokens,
diff --git a/vllm/entrypoints/openai/serving_responses.py b/vllm/entrypoints/openai/serving_responses.py
index c62fa0f7d..f79dad8d9 100644
--- a/vllm/entrypoints/openai/serving_responses.py
+++ b/vllm/entrypoints/openai/serving_responses.py
@@ -589,6 +589,13 @@ class OpenAIServingResponses(OpenAIServing):
             prev_msg=self.msg_store.get(prev_response.id) if prev_response else None,
             prev_response_output=prev_response.output if prev_response else None,
         )
+
+        chat_template_kwargs = dict(
+            reasoning_effort=None
+            if request.reasoning is None
+            else request.reasoning.effort
+        )
+
         _, engine_prompts = await self._preprocess_chat(
             request,
             tokenizer,
@@ -597,6 +604,7 @@ class OpenAIServingResponses(OpenAIServing):
             tool_parser=self.tool_parser,
             chat_template=self.chat_template,
             chat_template_content_format=self.chat_template_content_format,
+            chat_template_kwargs=chat_template_kwargs,
         )
         return messages, engine_prompts