putt @st .cache_resource on conversational_retriever

app.py

@@ -102,6 +102,23 @@ def load_llm_model():
     return llm
 
 
+@st.cache_resource
+def load_conversational_qa_memory_retriever(llm_model):
+    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')
+    
+    
+    
+    conversational_qa_memory_retriever = ConversationalRetrievalChain(
+        retriever=vector_database.as_retriever(),
+        question_generator=question_generator,
+        combine_docs_chain=doc_chain,
+        return_source_documents=True,
+        memory = memory,
+        get_chat_history=lambda h :h)
+    return conversational_qa_memory_retriever
+
 def load_retriever(llm, db):
     qa_retriever = RetrievalQA.from_chain_type(llm=llm, chain_type="stuff",
                             retriever=db.as_retriever(),
@@ -213,24 +230,13 @@ 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(llm_model)
 print("all load done")
 
 #Addional things for Conversation flows
-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')
 
 
 
-conversational_qa_memory_retriever = ConversationalRetrievalChain(
-    retriever=vector_database.as_retriever(),
-    question_generator=question_generator,
-    combine_docs_chain=doc_chain,
-    return_source_documents=True,
-    memory = memory,
-    get_chat_history=lambda h :h)