From 13bc171e9dd4d18cb7e1831ce9bc41608df257d4 Mon Sep 17 00:00:00 2001
From: fuzhongyun <15339891972@163.com>
Date: Thu, 15 Jan 2026 18:13:43 +0800
Subject: [PATCH] =?UTF-8?q?fix:=201.=E6=96=B0=E5=A2=9Ethink=E5=8F=82?=
 =?UTF-8?q?=E6=95=B0=202.=E7=A7=BB=E9=99=A4enable=5Fcot=E6=84=8F=E5=A4=96?=
 =?UTF-8?q?=E5=8F=82=E6=95=B0?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 app/api/routes.py | 2 ++
 app/core/rag.py   | 4 +++-
 2 files changed, 5 insertions(+), 1 deletion(-)

diff --git a/app/api/routes.py b/app/api/routes.py
index d221903..cba5952 100644
--- a/app/api/routes.py
+++ b/app/api/routes.py
@@ -34,6 +34,7 @@ class QueryRequest(BaseModel):
     mode: str = "hybrid"  # 可选: naive, local, global, hybrid
     top_k: int = 5
     stream: bool = False
+    think: bool = False
 
 class IngestResponse(BaseModel):
     filename: str
@@ -111,6 +112,7 @@ async def query_knowledge_base(
                     request.query, 
                     system_prompt=sys_prompt, 
                     stream=True,
+                    think=request.think,
                     hashing_kv=rag.llm_response_cache
                 )
                 
diff --git a/app/core/rag.py b/app/core/rag.py
index 8968f46..9581cd6 100644
--- a/app/core/rag.py
+++ b/app/core/rag.py
@@ -19,11 +19,13 @@ async def llm_func(prompt, system_prompt=None, history_messages=[], **kwargs) ->
     # 移除可能存在的 model 参数，避免冲突
     kwargs.pop('model', None)
     kwargs.pop('hashing_kv', None)
+    kwargs.pop('enable_cot', None)  # 移除不支持的参数
     keyword_extraction = kwargs.pop("keyword_extraction", False)
     if keyword_extraction:
         kwargs["format"] = "json"
     
     stream = kwargs.pop("stream", False)
+    think = kwargs.pop("think", False)
     # Debug: 检查流式参数
     if stream:
         logging.info("LLM called with stream=True")
@@ -40,7 +42,7 @@ async def llm_func(prompt, system_prompt=None, history_messages=[], **kwargs) ->
     if stream:
         async def inner():
             # 使用 **kwargs 透传参数，确保 format 等顶级参数生效
-            response = await client.chat(model=settings.LLM_MODEL, messages=messages, stream=True, **kwargs)
+            response = await client.chat(model=settings.LLM_MODEL, messages=messages, stream=True, think=think, **kwargs)
 
             async for chunk in response:
                 msg = chunk.get("message", {})