Remove full_messages

app.py

@@ -110,11 +110,9 @@ def chat_with_model(messages, pid):
     pad_id = current_tokenizer.pad_token_id or current_tokenizer.unk_token_id or 0
     eos_id = current_tokenizer.eos_token_id
 
-    # FULL conversation
-    full_messages = system_messages + messages
 
     # --- Generate from full context
-    prompt = format_prompt(full_messages)
+    prompt = format_prompt(messages)
 
     device = torch.device("cuda")
     current_model.to(device).half()