Lazy load heavy LLM libraries for faster startup

src/backend/chatbot.py

@@ -1,26 +1,35 @@
 import streamlit as st
 import pandas as pd
 import os
-from langchain.chat_models import ChatOpenAI
-from langchain.prompts.chat import (
-    ChatPromptTemplate,
-    SystemMessagePromptTemplate,
-    AIMessagePromptTemplate,
-    HumanMessagePromptTemplate,
-)
-from llama_index import (
-    SimpleDirectoryReader,
-    VectorStoreIndex,
-    ServiceContext,
-)
-from llama_index.llms import LlamaCPP
-from llama_index.llms.llama_utils import (
-    messages_to_prompt,
-    completion_to_prompt,
-)
 import subprocess
 import time
 
+# Lazy imports - only load when actually needed (saves 5-10 seconds on startup)
+def _lazy_import_llm_libs():
+    """Import heavy LLM libraries only when needed"""
+    global ChatOpenAI, ChatPromptTemplate, SystemMessagePromptTemplate
+    global AIMessagePromptTemplate, HumanMessagePromptTemplate
+    global SimpleDirectoryReader, VectorStoreIndex, ServiceContext
+    global LlamaCPP, messages_to_prompt, completion_to_prompt
+    
+    from langchain.chat_models import ChatOpenAI
+    from langchain.prompts.chat import (
+        ChatPromptTemplate,
+        SystemMessagePromptTemplate,
+        AIMessagePromptTemplate,
+        HumanMessagePromptTemplate,
+    )
+    from llama_index import (
+        SimpleDirectoryReader,
+        VectorStoreIndex,
+        ServiceContext,
+    )
+    from llama_index.llms import LlamaCPP
+    from llama_index.llms.llama_utils import (
+        messages_to_prompt,
+        completion_to_prompt,
+    )
+
 # set version
 # st.session_state.demo_lite = False
 
@@ -87,6 +96,9 @@ def init_llm(model, demo_lite):
     if demo_lite == False:
         print("BP 5 : running full demo")
         
+        # Load heavy LLM libraries now (lazy import)
+        _lazy_import_llm_libs()
+        
         # Detect GPU and environment
         env_config = detect_gpu_and_environment()
         n_gpu_layers = env_config["n_gpu_layers"]