Update app.py

app.py

@@ -38,10 +38,15 @@ def respond(
         n_batch=1024,
         n_ctx=8192,
     )
-    
     provider = LlamaCppPythonProvider(llm)
 
-    # Configure sampling settings
+    agent = LlamaCppAgent(
+        provider,
+        system_prompt=f"{system_message}",
+        predefined_messages_formatter_type=chat_template,
+        debug_output=True
+    )
+
     settings = provider.get_provider_default_settings()
     settings.temperature = temperature
     settings.top_k = top_k
@@ -50,34 +55,32 @@ def respond(
     settings.repeat_penalty = repeat_penalty
     settings.stream = True
 
-    # Prepare chat history
     messages = BasicChatHistory()
 
     for msn in history:
-        user_message = {'role': Roles.user, 'content': msn[0]}
-        assistant_message = {'role': Roles.assistant, 'content': msn[1]}
-        messages.add_message(user_message)
-        messages.add_message(assistant_message)
-
-    # Generate response using Llama model
-    try:
-        stream = llm(
-            prompt=message,
-            max_tokens=max_tokens,
-            temperature=temperature,
-            top_p=top_p,
-            top_k=top_k,
-            stop=["\n"],  # Adjust stop conditions as needed
-            echo=False  # Ensure only generated text is returned
-        )
-
-        outputs = ""
-        for output in stream:
-            outputs += output
-            yield outputs.strip()
-
-    except Exception as e:
-        yield f"Error: {str(e)}"
+        user = {
+            'role': Roles.user,
+            'content': msn[0]
+        }
+        assistant = {
+            'role': Roles.assistant,
+            'content': msn[1]
+        }
+        messages.add_message(user)
+        messages.add_message(assistant)
+
+    stream = agent.get_chat_response(
+        message,
+        llm_sampling_settings=settings,
+        chat_history=messages,
+        returns_streaming_generator=True,
+        print_output=False
+    )
+
+    outputs = ""
+    for output in stream:
+        outputs += output
+        yield outputs
 
 
 def create_interface(model_name):