Integrate lightweight LLM for extraction tasks in ai_chat.py

- Added a new lightweight LLM instance (llm_light) in config.py for simpler extraction tasks, optimizing performance with lower temperature and token limits.
- Updated the extract_and_update_persona function to utilize llm_light for persona extraction, enhancing efficiency in processing user inputs.

ai_chat.py

@@ -1,7 +1,7 @@
 from langgraph.graph import StateGraph, END
 from langchain_core.runnables import RunnableLambda
 from typing import TypedDict
-from config import llm, OPENAI_API_KEY
+from config import llm, llm_light, OPENAI_API_KEY
 from database import get_session_messages, save_message, update_user_persona, update_user_intent, search_properties, end_session
 import re
 
@@ -304,7 +304,7 @@ async def extract_and_update_persona(state):
         """
 
         # c. Call the LLM
-        response = await llm.ainvoke([{"role":"user","content":extraction_prompt}])
+        response = await llm_light.ainvoke([{"role":"user","content":extraction_prompt}])
         import json
         import re
         extracted = {}

config.py

@@ -25,6 +25,14 @@ else:
 # --- OpenAI Client Setup ---
 llm = ChatOpenAI(openai_api_key=OPENAI_API_KEY, model="gpt-4o-mini")
 
+# Lightweight LLM for simple extraction tasks (persona, classification, etc.)
+llm_light = ChatOpenAI(
+    openai_api_key=OPENAI_API_KEY, 
+    model="gpt-4.1-nano",
+    temperature=0.1,  # Lower temperature for more consistent extraction
+    max_tokens=200    # Lower token limit for simple tasks
+)
+
 # Initialize LLM only if API key is available
 if not OPENAI_API_KEY:
     print("Warning: OPENAI_API_KEY not set. AI functionality will be limited.")