Update app.py

app.py

@@ -40,6 +40,7 @@ if user_input:
     # Get response from GGUF LLM
     response = generate_llm_response(user_input)
     end_time = time.time()
+    inference_time = end_time - start_time
     # Display response
     st.chat_message("assistant").write(response)
     st.session_state["messages"].append({"role": "assistant", "content": response})