Update app.py

app.py

@@ -189,33 +189,34 @@ def generate(
 """
         yield warning_message
         is_first_interaction = False
+
     try:
-    response_type = determine_response_type(message)
+        response_type = determine_response_type(message)
         
-    if response_type == "short":
-        max_new_tokens = max(70, max_new_tokens)
-    elif response_type == "long":
-        max_new_tokens = min(max(150, max_new_tokens), 250)
-    else:  # medium
-        max_new_tokens = min(max(70, max_new_tokens), 150)
+        if response_type == "short":
+            max_new_tokens = max(70, max_new_tokens)
+        elif response_type == "long":
+            max_new_tokens = min(max(150, max_new_tokens), 250)
+        else:  # medium
+            max_new_tokens = min(max(70, max_new_tokens), 150)
         
-    chat_history = chat_history[-MAX_HISTORY_LENGTH:]
-    
-    conversation = []
-    
-    # Ajout du system prompt et du LUCAS_KNOWLEDGE_BASE
-    enhanced_system_prompt = f"{system_prompt}\n\n{LUCAS_KNOWLEDGE_BASE}"
-    conversation.append({"role": "system", "content": enhanced_system_prompt})
-    
-    # Ajout des 5 derniers inputs utilisateur uniquement
-    for user, _ in chat_history:
-        conversation.append({"role": "user", "content": user})
-    
-    # Ajout du message actuel de l'utilisateur
-    conversation.append({"role": "user", "content": message})
+        chat_history = chat_history[-MAX_HISTORY_LENGTH:]
+        
+        conversation = []
+        
+        # Ajout du system prompt et du LUCAS_KNOWLEDGE_BASE
+        enhanced_system_prompt = f"{system_prompt}\n\n{LUCAS_KNOWLEDGE_BASE}"
+        conversation.append({"role": "system", "content": enhanced_system_prompt})
+        
+        # Ajout des 5 derniers inputs utilisateur uniquement
+        for user, _ in chat_history:
+            conversation.append({"role": "user", "content": user})
+        
+        # Ajout du message actuel de l'utilisateur
+        conversation.append({"role": "user", "content": message})
 
-    input_ids = tokenizer.apply_chat_template(conversation, return_tensors="pt")
-    attention_mask = input_ids.ne(tokenizer.pad_token_id).long()
+        input_ids = tokenizer.apply_chat_template(conversation, return_tensors="pt")
+        attention_mask = input_ids.ne(tokenizer.pad_token_id).long()
 
         if input_ids.shape[1] > MAX_INPUT_TOKEN_LENGTH:
             input_ids = input_ids[:, -MAX_INPUT_TOKEN_LENGTH:]
@@ -261,7 +262,7 @@ def generate(
     finally:
         # Nettoyage de la mémoire GPU
         torch.cuda.empty_cache()
-        
+
 def vote(data: gr.LikeData, history):
     user_input = history[-1][0] if history else ""