return question_generator from the function to check history chat

app.py

@@ -104,7 +104,7 @@ def load_llm_model():
 
 @st.cache_resource
 def load_conversational_qa_memory_retriever():
-    global question_generator
+
     question_generator = LLMChain(llm=llm_model, prompt=CONDENSE_QUESTION_PROMPT)
     doc_chain = load_qa_chain(llm_model, chain_type="stuff", prompt = PROMPT)
     memory = ConversationBufferWindowMemory(k = 3,  memory_key="chat_history", return_messages=True,  output_key='answer')
@@ -118,7 +118,7 @@ def load_conversational_qa_memory_retriever():
         return_source_documents=True,
         memory = memory,
         get_chat_history=lambda h :h)
-    return conversational_qa_memory_retriever
+    return conversational_qa_memory_retriever, question_generator
 
 def load_retriever(llm, db):
     qa_retriever = RetrievalQA.from_chain_type(llm=llm, chain_type="stuff",
@@ -232,7 +232,7 @@ embedding_model = load_embedding_model()
 vector_database = load_faiss_index()
 llm_model = load_llm_model()
 qa_retriever = load_retriever(llm= llm_model, db= vector_database)
-conversational_qa_memory_retriever = load_conversational_qa_memory_retriever()
+conversational_qa_memory_retriever, question_generator = load_conversational_qa_memory_retriever()
 print("all load done")
 
 #Addional things for Conversation flows