Pass user input to extract_city_from_rag to increase result accuracy

extra/rag_agent.txt

@@ -142,16 +142,19 @@ def search_vector_db(query):
     return snippets
 
 # --- See if we can pull a city from the RAG results ---
-def extract_city_from_rag(snippets):
-    """Try to extract city names from RAG results"""
+def extract_city_from_rag(snippets, user_input):
+    """Try to extract city names from RAG results.  Uses conversation history like an LLM for match validation"""
     KNOWN_CITIES = ["New York", "San Francisco", "Chicago", "Austin", "Boston",
                     "London", "Toronto", "Tokyo", "Sydney", "Berlin"]
 
+    known_cities_lower = {city.lower(): city for city in KNOWN_CITIES}
+    user_input_lower = user_input.lower()
+
     for snippet in snippets:
-        for city in KNOWN_CITIES:
-            if city.lower() in snippet.lower():
-                print(f"{GREEN}RAG detected city: {city}{RESET}")
-                return city
+        if user_input_lower in known_cities_lower and user_input_lower in snippet.lower():
+            city = known_cities_lower[user_input_lower]
+            print(f"{GREEN}RAG detected city: {city}{RESET}")
+            return city
     return None
 
 # --- As a fallback, try to pull a city via the LLM ---
@@ -236,7 +239,7 @@ if __name__ == "__main__":
         rag_snippets = search_vector_db(user_input)
 
         # City Detection Workflow
-        detected_city = extract_city_from_rag(rag_snippets)
+        detected_city = extract_city_from_rag(rag_snippets, user_input)
         if not detected_city:
             detected_city = fallback_detect_city_with_llm(user_input)