fix：修复任务失败不进入重试的逻辑。feat：优化快慢任务创建逻辑

app/server/ChatbotAgent/agent_server.py

@@ -195,33 +195,26 @@ class LCAgent(ls.LitAPI):
         task_map = {}
 
         stylist_agent_kwages = self.stylist_agent_kwages.copy()
+        tasks_mapping = {}
         if num_outfits == 1:
-            logger.info(f"fast request outfit_id is : {outfit_ids[0]}")
-            # 通过请求数量判断 num == 1 单个outfit刷新
-            stylist_agent_kwages['outfit_id'] = outfit_ids[0]
-            stylist_agent_kwages['stylist_name'] = stylist_name
-            stylist_agent_kwages['gender'] = gender
-            agent = AsyncStylistAgent(**stylist_agent_kwages)
-            task = agent.run_quick_batch_styling(
-                request_summary=request_summary,
-                occasions=occasions,
-                start_outfit=start_outfit,
-                batch_sources=batch_sources,
-                user_id=user_id,
-                callback_url=callback_url,
-            )
-            tasks.append(task)
-            task_map[task] = {"outfit_id": outfit_ids[0], "retries": 0}
-        elif num_outfits > 1:
-            # 通过请求数量判断 num > 1 四套搭配推荐 (1快 , num-1慢)
+            tasks_mapping[outfit_ids[0]] = "fast"
+        else:
             for i in range(num_outfits):
-                stylist_agent_kwages['outfit_id'] = outfit_ids[i]
+                if i == 0:
+                    tasks_mapping[outfit_ids[i]] = "fast"
+                else:
+                    tasks_mapping[outfit_ids[i]] = "slow"
+
+        for k, v in tasks_mapping.items():
+            logger.info(f"fast request outfit_id is : {k}")
+            # 通过请求数量判断 num == 1 单个outfit刷新
+            stylist_agent_kwages['outfit_id'] = k
             stylist_agent_kwages['stylist_name'] = stylist_name
             stylist_agent_kwages['gender'] = gender
+            stylist_agent_kwages['callback_url'] = callback_url
+
             agent = AsyncStylistAgent(**stylist_agent_kwages)
-                if i == 0:
-                    # 第一套搭配使用快速方法 一次跑出所有单品
-                    logger.info(f"fast request outfit_id is : {outfit_ids[i]}")
+            if v == "fast":
                 task = agent.run_quick_batch_styling(
                     request_summary=request_summary,
                     occasions=occasions,
@@ -231,7 +224,6 @@ class LCAgent(ls.LitAPI):
                     callback_url=callback_url,
                 )
             else:
-                    # 后续
                 task = agent.run_iterative_styling(
                     request_summary=request_summary,
                     occasions=occasions,
@@ -241,7 +233,7 @@ class LCAgent(ls.LitAPI):
                     callback_url=callback_url,
                 )
             tasks.append(task)
-                task_map[task] = {"outfit_id": outfit_ids[i], "retries": 0}
+            task_map[task] = {"outfit_id": k, "retries": 0}
         logger.info(f"--- Starting {num_outfits} concurrent outfit generation tasks. ---")
 
         # 2. 任务执行与重试循环
@@ -252,7 +244,8 @@ class LCAgent(ls.LitAPI):
         retry_limit = 1  # 允许重试一次
         while tasks_to_run:
             try:
-                results = await asyncio.gather(*tasks, return_exceptions=True)
+                results = await asyncio.gather(*tasks_to_run, return_exceptions=True)
+
                 next_tasks_to_run = []
                 for task, result in zip(tasks_to_run, results):
                     task_info = task_map[task]
@@ -279,8 +272,20 @@ class LCAgent(ls.LitAPI):
                             stylist_agent_kwages['outfit_id'] = outfit_id
                             stylist_agent_kwages['stylist_name'] = stylist_name
                             stylist_agent_kwages['gender'] = gender
+                            stylist_agent_kwages['callback_url'] = callback_url
+
                             agent = AsyncStylistAgent(**stylist_agent_kwages)
+                            if tasks_mapping[outfit_id] == "fast":
                                 new_task = agent.run_quick_batch_styling(
+                                    request_summary=request_summary,
+                                    occasions=occasions,
+                                    start_outfit=start_outfit,
+                                    batch_sources=batch_sources,
+                                    user_id=user_id,
+                                    callback_url=callback_url,
+                                )
+                            else:
+                                new_task = agent.run_iterative_styling(
                                     request_summary=request_summary,
                                     occasions=occasions,
                                     start_outfit=start_outfit,
@@ -298,11 +303,12 @@ class LCAgent(ls.LitAPI):
                             # 达到重试上限，最终失败 ， 并通知前端
                             object_data = {
                                 'outfit_id': outfit_id,
-                                "status": "failed",
+                                "status": "retry_failed",
                                 "path": "",
                             }
-                            post_request(url=f'{callback_url}/api/style/callback', data=json.dumps(object_data))
+                            response = post_request(url=f'{callback_url}/api/style/callback', data=json.dumps(object_data))
                             failed_outfits.append(f"Outfit {outfit_id} ultimately failed after {retry_limit} retries: {result}")
+                            logger.info(f"request data ：{json.dumps(object_data, ensure_ascii=False, indent=2)} | JAVA callback info -> status:{response.status_code} | message:{response.text}")
                             del task_map[task]
 
                     else:

app/server/ChatbotAgent/core/stylist_agent_server.py

@@ -27,7 +27,7 @@ logger = logging.getLogger(__name__)
 
 
 class AsyncStylistAgent:
-    def __init__(self, local_db: str, gemini_model_name: str, outfit_id: str, stylist_name: str, gender: str):
+    def __init__(self, local_db: str, gemini_model_name: str, outfit_id: str, stylist_name: str, gender: str, callback_url: str):
         # self.outfit_items: List[Dict[str, str]] = []
         self.outfit_id = outfit_id
         self.stylist_name = stylist_name
@@ -52,6 +52,7 @@ class AsyncStylistAgent:
         )
         self.gcs_bucket = "lc_stylist_agent_outfit_items"
         self.minio_bucket = "lanecarford"
+        self.callback_url = f'{callback_url}/api/style/callback'
 
     def _load_style_guide(self, stylist_name: str):
         """加载 markdown 风格指南内容。"""
@@ -182,8 +183,13 @@ class AsyncStylistAgent:
             results = []
 
         if not results:
-            print(f"数据库中未找到符合 '{category}' 和描述的单品。")
-            return None
+            self.post_operation(
+                status="failed",
+                message=f"数据库中未找到符合 '{category}' 和描述的单品。",
+                callback_url=self.callback_url,
+                img_path="",
+            )
+            raise Exception(f"数据库中未找到符合 '{category}' 和描述的单品。")
 
         # 3. 模拟 Agent 审核(实际应用中,你需要将图片发回给 Agent进行审核)
         best_meta = results[0]  # 第一个 batch 的第一个 metadata
@@ -295,14 +301,13 @@ class AsyncStylistAgent:
             gemini_data = self._parse_gemini_response(gemini_response_text)
 
             if not gemini_data:
-                print("Agent 返回无效响应,终止流程。")
                 self.post_operation(
                     status="failed",
                     message="Agent returned invalid response, terminating process.",
                     callback_url=url,
                     img_path=merged_image_path,
                 )
-                break
+                raise Exception("Agent 返回无效响应,终止流程。")
 
             # 处理推荐单品
             if gemini_data.get('action') == 'recommend_item':
@@ -385,13 +390,14 @@ class AsyncStylistAgent:
         failed_found_item_count = 0
 
         if not recommended_items or not isinstance(recommended_items, List):
-            print("No recommended item from Gemini, terminating process.")
             self.post_operation(
                 status="failed",
                 message="Agent returned invalid response, terminating process.",
                 callback_url=url,
                 img_path=merged_image_path
             )
+            raise Exception("No recommended item from Gemini, terminating process.")
+
         else:
             for idx, rec_item in enumerate(recommended_items):
                 subcategory = rec_item.get('subcategory')
@@ -428,7 +434,7 @@ class AsyncStylistAgent:
                     callback_url=url,
                     img_path=merged_image_path
                 )
-                print(f"There are {failed_found_item_count} items (total {len(recommended_items)}) are not found in the database")
+                raise Exception(f"There are {failed_found_item_count} items (total {len(recommended_items)}) are not found in the database")
         return reason
 
     async def run_iterative_styling(self, request_summary, occasions, start_outfit: Optional[List] = None, batch_sources: List = [], user_id="test", callback_url=""):