better caching

app.py

@@ -95,6 +95,7 @@ logging.getLogger().addHandler(logging.StreamHandler(stream=sys.stdout))
 llama_debug = LlamaDebugHandler(print_trace_on_end=True)
 callback_manager = CallbackManager([llama_debug])
 
+@st.cache_resource
 #One doc embedding
 def load_emb_uploaded_document(filename):
  # You may want to add a check to prevent execution during initialization.
@@ -164,7 +165,6 @@ if 'emb_model' not in st.session_state:
 # Use the models from session state
 query_engine = st.session_state.emb_model
 
-
 # ------------------------------------layout----------------------------------------
 
 with st.sidebar:
@@ -235,8 +235,10 @@ with tab2:
 with tab3:
  st.title("📝 One single document Q&A with Llama Index using local open llms")
  if st.button('Reinitialize Query Engine', key='reinit_engine'):
- query_engine = st.session_state.emb_model
+ del st.session_state["emb_model_upload_doc"]
+ st.session_state.emb_model_upload_doc = ""
  st.write("Query engine reinitialized.")
+
  uploaded_file = st.file_uploader("Upload an File", type=("txt", "csv", "md","pdf"))
  question = st.text_input(
  "Ask something about the files",
@@ -251,20 +253,23 @@ with tab3:
  if not os.path.exists("draft_docs"):
  st.error("draft_docs directory does not exist. Please download and copy paste a model in folder models.")
  os.makedirs("draft_docs")
-
  with open("draft_docs/"+uploaded_file.name, "wb") as f:
  text = uploaded_file.read()
  f.write(text)
  text = uploaded_file.read()
+ # Embedding Model Loading
+ if 'emb_model_upload_doc' not in st.session_state:
+ st.session_state.emb_model_upload_doc = load_emb_uploaded_document("draft_docs/"+uploaded_file.name)
+ # Use the models from session state
+ query_engine_upload_doc = st.session_state.emb_model_upload_doc
  # if load_emb_uploaded_document:
  # load_emb_uploaded_document.clear()
  #load_emb_uploaded_document.clear()
- query_engine = load_emb_uploaded_document("draft_docs/"+uploaded_file.name)
  st.write("File ",uploaded_file.name, "was loaded successfully")
 
  if uploaded_file and question and api_server_info:
  contextual_prompt = st.session_state.memory + "\n" + question
- response = query_engine.query(contextual_prompt)
+ response = query_engine_upload_doc.query(contextual_prompt)
  text_response = response.response
  st.write("### Answer")
  st.markdown(text_response)