Update app.py

app.py

@@ -22,6 +22,7 @@ def respond(
     from llama_cpp_agent import LlamaCppAgent
     from llama_cpp_agent import MessagesFormatterType
     from llama_cpp_agent.providers import LlamaCppPythonProvider
+    from llama_cpp_agent.chat_history import BasicChatHistory
     
     llm = Llama(
         model_path="models/mistral-7b-instruct-v0.2.Q6_K.gguf",
@@ -39,8 +40,18 @@ def respond(
     settings = provider.get_provider_default_settings()
     settings.max_tokens = 2000
     settings.stream = True
+
+    messages = BasicChatHistory()
+    print(history)
+
+    for msn in history:
+        dic = {
+            'role': msn[0]
+            'content': msn[1]
+        }
+        messages.add_message(dic)
     
-    stream = agent.get_chat_response(message, llm_sampling_settings=settings, returns_streaming_generator=True)
+    stream = agent.get_chat_response(message, llm_sampling_settings=settings, chat_history=messages, returns_streaming_generator=True)
     
     outputs = ""
     for output in stream: