fix bug LLM

app/main.py

@@ -150,7 +150,7 @@ async def process_message(message_data: Dict[str, Any]):
     if not message_data or not isinstance(message_data, dict):
         logger.error(f"[ERROR] Invalid message_data: {message_data}")
         return
-    required_fields = ["sender_id", "page_id", "text"]
+    required_fields = ["sender_id", "page_id", "text", "timestamp"]
     for field in required_fields:
         if field not in message_data:
             logger.error(f"[ERROR] Missing field {field} in message_data: {message_data}")
@@ -158,6 +158,7 @@ async def process_message(message_data: Dict[str, Any]):
     sender_id = message_data["sender_id"]
     page_id = message_data["page_id"]
     message_text = message_data["text"]
+    timestamp = message_data["timestamp"]
     attachments = message_data.get('attachments', [])
     logger.bind(user_id=sender_id, page_id=page_id, message=message_text).info("Processing message")
 
@@ -166,7 +167,19 @@ async def process_message(message_data: Dict[str, Any]):
         logger.info(f"[DEBUG] Không có message_text và attachments, không xử lý...")
         return
 
-    # Get page access token (sync)
+    # Get conversation history (run in thread pool)
+    loop = asyncio.get_event_loop()
+    history = await loop.run_in_executor(
+        executor, lambda: sheets_client.get_conversation_history(sender_id, page_id)
+    )
+    logger.info(f"[DEBUG] history: {history}")
+
+    if history:
+        if history["timestamp"] == timestamp and history["recipient_id"] == sender_id and history["page_id"] == page_id:
+            logger.info(f"[DEBUG] Message bị duplicate (trùng sender, page, timestamp), không xử lý...")
+            return 
+
+ # Get page access token (sync)
     page_token = supabase_client.get_page_token(page_id)
     if page_token:
         logger.info(f"[DEBUG] page_token: {page_token[:10]} ... {page_token[-10:]}")
@@ -178,7 +191,7 @@ async def process_message(message_data: Dict[str, Any]):
         logger.error(f"No access token found for page {page_id}")
         return
 
-    # Extract command and keywords
+# Extract command and keywords
     command, remaining_text = extract_command(message_text)
     # Sử dụng LLM để phân tích message_text và extract keywords, mục đích, hành vi vi phạm
     llm_analysis = await llm_client.analyze(message_text)
@@ -198,13 +211,6 @@ async def process_message(message_data: Dict[str, Any]):
 
     logger.info(f"[DEBUG] Phương tiện: {keywords} - Hành vi: {hanh_vi_vi_pham} - Mục đích: {muc_dich}")
 
-    # Get conversation history (run in thread pool)
-    loop = asyncio.get_event_loop()
-    history = await loop.run_in_executor(
-        executor, lambda: sheets_client.get_conversation_history(sender_id, page_id)
-    )
-    logger.info(f"[DEBUG] history: {history}")
-
     response = ""
     log_needed = True
     log_kwargs = {
@@ -218,6 +224,7 @@ async def process_message(message_data: Dict[str, Any]):
         'vehicle': ','.join(keywords),
         'action': hanh_vi_vi_pham,
         'purpose': muc_dich,
+        'timestamp': timestamp,
         'is_done': False
     }
 

app/sheets.py

@@ -90,8 +90,8 @@ class SheetsClient:
             history = []
 
             for row in values:
-                # Bổ sung cột rỗng cho đủ 11 cột
-                row = row + [""] * (11 - len(row))
+                # Bổ sung cột rỗng cho đủ 12 cột
+                row = row + [""] * (12 - len(row))
                 if row[4] == user_id and row[5] == page_id and row[8].lower() == 'false':
                     history.append({
                         'conversation_id': row[0],
@@ -104,7 +104,8 @@ class SheetsClient:
                         'originalvehicle': row[7],
                         'originalaction': row[8],
                         'originalpurpose': row[9],
-                        'isdone': row[10].lower() == 'true'
+                        'timestamp': row[10],
+                        'isdone': row[11].lower() == 'true'
                     })
 
             return history
@@ -125,6 +126,7 @@ class SheetsClient:
         vehicle: str = "",
         action: str = "",
         purpose: str = "",
+        timestamp: str = "",
         is_done: bool = False
     ) -> bool:
         """
@@ -138,11 +140,11 @@ class SheetsClient:
             if not self.service:
                 raise RuntimeError("Google Sheets service not initialized")
 
-            timestamp = datetime.now().isoformat()
+            ts = datetime.now().isoformat()
 
             if not conversation_id:
                 # Create new conversation
-                conversation_id = generate_conversation_id(user_id, page_id, timestamp)
+                conversation_id = generate_conversation_id(user_id, page_id, ts)
                 values = [[
                     conversation_id,
                     command,
@@ -154,6 +156,7 @@ class SheetsClient:
                     vehicle,
                     action,
                     purpose,
+                    timestamp,
                     str(is_done).lower()
                 ]]
 
@@ -188,8 +191,8 @@ class SheetsClient:
                 if row_index is not None:
                     # Lấy dữ liệu dòng hiện tại
                     current_row = values[row_index]
-                    # Đảm bảo đủ 11 cột
-                    while len(current_row) < 11:
+                    # Đảm bảo đủ 12 cột
+                    while len(current_row) < 12:
                         current_row.append("")
                     # Tạo dòng mới với giá trị mới nếu có, giữ nguyên nếu không
                     new_row = [
@@ -203,7 +206,8 @@ class SheetsClient:
                         vehicle if vehicle else current_row[7],
                         action if action else current_row[8],
                         purpose if purpose else current_row[9],
-                        str(is_done).lower() if is_done is not None else current_row[10]
+                        timestamp if timestamp else current_row[10],
+                        str(is_done).lower() if is_done is not None else current_row[11]
                     ]
                     update_range = f"{SHEET_RANGE.split('!')[0]}!A{row_index + 1}"
                     body = {