feat: skip response cache for /chat-dev endpoint

backend/agent/controller.py

@@ -42,6 +42,7 @@ async def chat_controller(
     return_user_insight: bool = False,
     is_authenticated: bool = False,  # New: để track user vs guest
     device_id: str | None = None,  # Device ID từ request (có cả khi authenticated)
+    skip_cache: bool = False,  # Skip response cache (for dev endpoint)
 ) -> dict:
     """
     Controller main logic for non-streaming chat requests.
@@ -63,7 +64,7 @@ async def chat_controller(
     logger.info("chat_controller start: model=%s, key=%s", model_name, identity_key)
 
     # ====================== CACHE LAYER ======================
-    if REDIS_CACHE_TURN_ON:
+    if REDIS_CACHE_TURN_ON and not skip_cache:
         cached_response = await redis_cache.get_response(user_id=identity_key, query=query)
         if cached_response:
             logger.info("CACHE HIT key=%s", identity_key)
@@ -303,7 +304,7 @@ async def chat_controller(
     if user_insight_dict is not None:
         response_payload["user_insight"] = user_insight_dict
 
-    if REDIS_CACHE_TURN_ON:
+    if REDIS_CACHE_TURN_ON and not skip_cache:
         await redis_cache.set_response(user_id=identity_key, query=query, response_data=response_payload, ttl=300)
 
     background_tasks.add_task(

backend/agent/tools/product_mapping.py

@@ -64,6 +64,7 @@ PRODUCT_LINE_MAP: dict[str, list[str]] = {
     "Bộ quần áo": ["bộ quần áo", "đồ bộ"],
     "Bộ mặc nhà": ["bộ mặc nhà", "đồ ngủ", "đồ mặc nhà"],
     "Blazer": ["blazer"],
+    "Tất": ["tất", "vớ", "bao chân", "vớ chân", "tất chân"],
 }
 
 # ==============================================================================

backend/api/chatbot_route.py

@@ -107,6 +107,7 @@ async def fashion_qa_chat_dev(request: Request, req: QueryRequest, background_ta
             return_user_insight=False,
             is_authenticated=is_authenticated,  # Pass auth status for Langfuse metadata
             device_id=device_id,  # Luôn truyền device_id để lưu vào Langfuse metadata
+            skip_cache=True,  # Dev endpoint: luôn gọi LLM mới, không cache
         )
 
         usage_info = await message_limit_service.increment(