added memory

backend.py

@@ -13,8 +13,10 @@ from llama_cpp import Llama
 import spaces
 from huggingface_hub import login
 from llama_index.core.memory import ChatMemoryBuffer
-from typing import Iterator
+from typing import Iterator, List
 from llama_index.core.chat_engine import CondensePlusContextChatEngine
+from llama_index.core.llms import ChatMessage, MessageRole
+
 
 
 
@@ -66,39 +68,16 @@ def handle_query(query_str: str,
           
         memory = ChatMemoryBuffer.from_defaults(token_limit=3900)
         
-        
-        """template = (
-        "Sei un assistente Q&A italiano di nome Odi, che risponde solo alle domande o richieste pertinenti in modo preciso. \n"
-        "---------------------\n"
-        "{context_str}"
-        "\n---------------------\n"
-        "Usa la cronologia delle chat precedenti, o il contesto sopra, per interagire e aiutare l'utente a rispondere alla domanda: {query_str}\n"
-        )
-        system_prompt_template = PromptTemplate(template)
-        
-        
-        system_message_content = system_prompt_template.format(context_str="Il vino è più buono a 30 gradi", query_str=query_str) #
-        
-        chat_engine = CondensePlusContextChatEngine.from_defaults(
-        index.as_retriever(),
-        memory=memory,
-        #llm=GemmaLLMInterface(),
-        system_prompt=(
-            system_message_content
-            ),
-        verbose=True,
-        )"""
-        
-        
-        conversation = []
+        conversation: List[ChatMessage] = []
         for user, assistant in chat_history:
             conversation.extend(
                 [
-                    {"role": "user", "content": user},
-                    {"role": "assistant", "content": assistant},
+                    ChatMessage(role=MessageRole.USER, content=user),
+                    
+                    ChatMessage(role=MessageRole.ASSISTANT, content=assistant),
                 ]
             )
-        conversation.append({"role": "user", "content": query_str})
+        conversation.append( ChatMessage(role=MessageRole.USER, content=query_str))
         
         chat_engine = index.as_chat_engine(
         chat_mode="condense_plus_context",
@@ -119,10 +98,8 @@ def handle_query(query_str: str,
         )
         
         
-        
-        
         outputs = []
-        response = chat_engine.stream_chat(conversation)
+        response = chat_engine.stream_chat(query_str, conversation)
         #response = chat_engine.chat(query_str)
         for token in response.response_gen:
           #if not token.startswith("system:") and not token.startswith("user:"):