Enhance intent extraction logic to handle property inquiries

- Added a classification check in `extract_and_update_intent` to determine if the user is asking for property listings, allowing the system to skip preference questions when appropriate.
- Updated the state graph to reorder the processing flow, ensuring intent classification occurs before intent extraction, improving the overall logic and efficiency of user interactions.
- These changes aim to refine the user experience by providing more relevant responses based on user intent, while maintaining clarity in the conversation flow.

ai_chat.py

@@ -269,6 +269,14 @@ async def extract_and_update_intent(state):
 
     missing = [f for f in intent_fields if state["intent"].get(f) is None]
     if missing:
+        # Check if user is asking for properties (using AI classification)
+        classification = state.get("classification")
+        if classification == "search_listings":
+            # User is asking for properties, don't interrupt with preference questions
+            print("DEBUG - User asking for properties, skipping preference questions")
+            return {"response": None}
+        
+        # User is setting preferences, ask for missing fields
         questions = {
             "location_preference": "Hi there! Which area or suburb are you interested in?",
             "budget": "Hi there! What is your monthly budget?",
@@ -418,15 +426,15 @@ async def detect_end_chat(state):
 # --- Build LangGraph ---
 graph = StateGraph(ChatState)
 graph.add_node("persona_update", RunnableLambda(extract_and_update_persona))
-graph.add_node("intent_update", RunnableLambda(extract_and_update_intent))
 graph.add_node("classify_intent", RunnableLambda(classify_user_intent))
+graph.add_node("intent_update", RunnableLambda(extract_and_update_intent))
 graph.add_node("property_search", RunnableLambda(extract_and_search_properties))
 graph.add_node("exit_check", RunnableLambda(detect_end_chat))
 graph.add_node("chat", RunnableLambda(chat_with_session_memory))
 graph.set_entry_point("persona_update")
-graph.add_edge("persona_update", "intent_update")
-graph.add_edge("intent_update", "classify_intent")
-graph.add_edge("classify_intent", "property_search")
+graph.add_edge("persona_update", "classify_intent")
+graph.add_edge("classify_intent", "intent_update")
+graph.add_edge("intent_update", "property_search")
 graph.add_edge("property_search", "exit_check")
 graph.add_edge("exit_check", "chat")
 graph.add_edge("chat", END)