Merge branch 'develop_fifa_2026_replay_match' of ssh://gitlab.cmvideo.cn:7999/pr…

…oduction_capacity/clip/aigc-group/aigc-embedding-service into develop_fifa_2026_replay_match

# Conflicts:
#	src/football_replay_match/cfg/config.prod.yaml
#	src/football_replay_match/core/api.py
#	src/football_replay_match/core/llm_image.py

@@ -21,8 +21,8 @@ output_kafka:
   password: ${kafka_password}
 
 llm:
-  base_url: http://192.168.1.59:11434
-  model_name: qwen3.6:35b-a3b-q8_0
+  base_url: http://192.168.1.59:11434/v1
+  model_name: Qwen3.6-35B-A3B-UD-Q8_K_XL.gguf
   temperature: 0.7
 
 common:

@@ -3,7 +3,7 @@ import os.path
 from pathlib import Path
 
 from langchain_core.messages import SystemMessage, HumanMessage
-from langchain_ollama import ChatOllama
+from langchain_openai import ChatOpenAI
 
 try:
     from .llm_video_content import contents as video_contents
@@ -79,7 +79,8 @@ class FootballReplayMatchLive:
         #                    keep_alive=-1, reasoning=False)
         # self.model = ChatOpenAI(base_url="http://192.168.1.59:11434/v1", model="qwen3.6:35b-a3b-q8_0", temperature=0.7,
         #                        api_key='no_key')
-        self.model = ChatOllama(base_url=base_url, model=model, temperature=temperature, keep_alive=-1, reasoning=False)
+        self.model = ChatOpenAI(base_url=base_url, model=model, temperature=temperature, api_key=api_key,
+                                extra_body={"chat_template_kwargs": {"enable_thinking": False}})
 
     def _match_once(self, replay_video_contents: list, live_videos: list[dict], cache_path=None, record: list = None):
 

@@ -3,7 +3,7 @@ import os.path
 from pathlib import Path
 
 from langchain_core.messages import SystemMessage, HumanMessage
-from langchain_ollama import ChatOllama
+from langchain_openai import ChatOpenAI
 
 try:
     from .llm_video_content import contents as video_contents
@@ -74,7 +74,8 @@ class FootballReplayVideoEvent:
         #                    keep_alive=-1, reasoning=False)
         # self.model = ChatOpenAI(base_url="http://192.168.1.59:11434/v1", model="qwen3.6:35b-a3b-q8_0", temperature=0.7,
         #                        api_key='no_key')
-        self.model = ChatOllama(base_url=base_url, model=model, temperature=temperature, keep_alive=-1, reasoning=False)
+        self.model = ChatOpenAI(base_url=base_url, model=model, temperature=temperature, api_key=api_key,
+                                extra_body={"chat_template_kwargs": {"enable_thinking": False}})
 
     def video_event(self, video_path: str, asr_text: str = '无', cache_path=None):
         if cache_path is not None and os.path.exists(cache_path):

@@ -6,9 +6,9 @@ import json
 
 
 def batch_match():
-    replay_match_live = FootballReplayMatchLive(base_url="http://192.168.1.59:11434", model="qwen3.6:35b-a3b-q8_0")
+    replay_match_live = FootballReplayMatchLive(base_url="http://192.168.1.59:11434/v1", model="Qwen3.6-35B-A3B-UD-Q8_K_XL.gguf")
     qwen_asr = QwenAsr(base_url="http://192.168.1.59:8101/v1", model="Qwen/Qwen3-ASR-1.7B")
-    fbrv = FootballReplayVideoEvent(base_url="http://192.168.1.59:11434", model="qwen3.6:35b-a3b-q8_0", temperature=0.7)
+    fbrv = FootballReplayVideoEvent(base_url="http://192.168.1.59:11434/v1", model="Qwen3.6-35B-A3B-UD-Q8_K_XL.gguf", temperature=0.7)
 
     videos_dir = rf"R:\wdx\202604\20260420_download_football_video\finished"
     for video_name in sorted(os.listdir(videos_dir)):

@@ -4,7 +4,7 @@ import os
 from football_replay_video_event_by_llm import FootballReplayVideoEvent
 from qwen_asr_util import QwenAsr
 def batch_with_asr():
-    fbrv = FootballReplayVideoEvent(base_url="http://192.168.1.59:11434", model="qwen3.6:35b-a3b-q8_0", temperature=0.7)
+    fbrv = FootballReplayVideoEvent(base_url="http://192.168.1.59:11434/v1", model="Qwen3.6-35B-A3B-UD-Q8_K_XL.gguf", temperature=0.7)
     qwen_asr = QwenAsr(base_url="http://192.168.1.59:8101/v1", model="Qwen/Qwen3-ASR-1.7B",temperature=0.7,
                            api_key='no_key')
 
@@ -27,7 +27,7 @@ def batch_with_asr():
             print(event_json)
 
 def batch_without_asr():
-    fbrv = FootballReplayVideoEvent(base_url="http://192.168.1.59:11434", model="qwen3.6:35b-a3b-q8_0", temperature=0.7)
+    fbrv = FootballReplayVideoEvent(base_url="http://192.168.1.59:11434/v1", model="Qwen3.6-35B-A3B-UD-Q8_K_XL.gguf", temperature=0.7)
     # qwen_asr = QwenAsr(base_url="http://192.168.1.59:8101/v1", model="Qwen/Qwen3-ASR-1.7B", temperature=0.7,
     #                    api_key='no_key')
 
@@ -51,7 +51,7 @@ def batch_without_asr():
             print(event_json)
 
 def one_video_test(video_path):
-    fbrv = FootballReplayVideoEvent(base_url="http://192.168.1.59:11434", model="qwen3.6:35b-a3b-q8_0", temperature=0.7)
+    fbrv = FootballReplayVideoEvent(base_url="http://192.168.1.59:11434/v1", model="Qwen3.6-35B-A3B-UD-Q8_K_XL.gguf", temperature=0.7)
     event_json = fbrv.video_event(video_path, None)
     print(event_json)
 

@@ -3,7 +3,7 @@ import json
 from pathlib import Path
 
 from aabd.base.time_util import vms2str_auto
-from langchain_ollama import ChatOllama
+from langchain_openai import ChatOpenAI
 
 
 class ClipByEvent:
@@ -19,7 +19,7 @@ class ClipByEvent:
         #                    keep_alive=-1, reasoning=False)
         # self.model = ChatOpenAI(base_url="http://192.168.1.59:11434/v1", model="qwen3.6:35b-a3b-q8_0", temperature=0.7,
         #                        api_key='no_key')
-        self.model = ChatOllama(base_url=base_url, model=model, temperature=temperature, keep_alive=-1, reasoning=False)
+        self.model = ChatOpenAI(base_url=base_url, model=model, temperature=temperature, api_key=api_key)
 
     def get_video_match_time(self, video_path, video_time):
         import cv2