Update final_recommender.py

final_recommender.py

@@ -1,66 +1,77 @@
-"""
-FINAL SMART RECOMMENDER SYSTEM
-Embeddings + Intent + Filtering + Reranking + Keyword Boosting
-"""
-
-from filtered_search_engine import SmartRecommender
-from reranker import Reranker
-from keyword_boosting_layer import apply_keyword_boost  # <-- booster imported
-import pandas as pd
-
-class FinalSalahkar:
-
-    def __init__(self):
-        print("✨ Initializing Final AI Recommender...")
-        self.engine = SmartRecommender()
-        self.reranker = Reranker()
-        self.df = pd.read_csv("salahkar_enhanced.csv")
-        print("🚀 System Ready!")
-
-    def ask(self, query, k=7):
-        print("\n==============================================")
-        print(f"🧠 INPUT QUERY → {query}")
-        print("==============================================")
-
-        # STEP 1 → BASE SEARCH (filtered FAISS)
-        # Now returns (results_list, detected_intent)
-        results, intent = self.engine.recommend(query, k=k)
-
-        # STEP 2 → Prepare embedding text for reranking
-        prepared = []
-        for item in results:
-            name = item["name"]
-            # Find the full row for this item
-            row = self.df[self.df["name"] == name].iloc[0]
-            
-            prepared.append({
-                "name": name,
-                "domain": item["domain"],
-                "category": item["category"],
-                "region": item["region"],
-                "embedding_score": item["score"],
-                "text": row["search_embedding_text"],
-                "intent": intent # Pass intent for boosting
-            })
-
-        # STEP 3 → Cross-Encoder Reranking
-        ranked = self.reranker.rerank(query, prepared)
-
-        # STEP 4 → Keyword Boost (final scoring)
-        boosted = apply_keyword_boost(query, ranked)
-
-        # STEP 5 → Display final sorted results
-        print("\n🏆 FINAL SMART RANKED RESULTS:")
-        for i, item in enumerate(boosted[:k]):
-            print(f"{i+1}. {item['name']}  |  Final Score: {round(item['final_score'], 3)}")
-
-        return boosted
-
-
-if __name__ == "__main__":
-    bot = FinalSalahkar()
-
-    bot.ask("romantic historical place india")
-    bot.ask("spiritual peaceful temple")
-    bot.ask("best south indian spicy breakfast")
-    bot.ask("sweet festival food india")
+"""
+FINAL SMART RECOMMENDER SYSTEM
+Embeddings + Intent + Filtering + Reranking + Keyword Boosting
+"""
+
+from filtered_search_engine import SmartRecommender
+from reranker import Reranker
+from keyword_boosting_layer import apply_keyword_boost
+import pandas as pd
+
+class FinalSalahkar:
+
+    def __init__(self):
+        print("✨ Initializing Final AI Recommender...")
+        self.engine = SmartRecommender()
+        self.reranker = Reranker()
+        self.df = pd.read_csv("salahkar_enhanced.csv")
+        
+        # Optimization: Pre-compute name-to-row mapping for O(1) lookup
+        # This is much faster than running df[df["name"] == name] for every result
+        self.name_map = self.df.set_index("name").to_dict('index')
+        
+        print("🚀 System Ready!")
+
+    def ask(self, query, k=7):
+        print("\n==============================================")
+        print(f"🧠 INPUT QUERY → {query}")
+        print("==============================================")
+
+        # STEP 1 → BASE SEARCH (filtered FAISS)
+        # Now returns (results_list, detected_intent)
+        results, intent = self.engine.recommend(query, k=k)
+
+        # STEP 2 → Prepare embedding text for reranking
+        prepared = []
+        for item in results:
+            name = item["name"]
+            
+            # Optimization: FAST LOOKUP
+            row = self.name_map.get(name)
+            
+            if row:
+                prepared.append({
+                    "name": name,
+                    "domain": item["domain"],
+                    "category": item["category"],
+                    "region": item["region"],
+                    "embedding_score": item["score"],
+                    "text": row["search_embedding_text"], # Use index text for reranking
+                    "intent": intent 
+                })
+
+        # STEP 3 → Cross-Encoder Reranking
+        ranked = self.reranker.rerank(query, prepared)
+
+        # STEP 4 → Keyword Boost (final scoring)
+        # This applies the location/food/heritage boosts we configured
+        boosted = apply_keyword_boost(query, ranked)
+
+        # STEP 5 → Display final sorted results
+        print("\n🏆 FINAL SMART RANKED RESULTS:")
+        for i, item in enumerate(boosted[:k]):
+            score = item.get('final_score', item.get('rerank_score', 0))
+            print(f"{i+1}. {item['name']}  |  Final Score: {round(score, 3)}")
+
+        return boosted
+
+
+if __name__ == "__main__":
+    bot = FinalSalahkar()
+
+    # Test the complex new queries to verify accuracy
+    bot.ask("romantic historical place india")
+    bot.ask("spiritual peaceful temple")
+    bot.ask("best south indian spicy breakfast")
+    bot.ask("sweet festival food india")
+    bot.ask("famous beaches near ongole") # Test new data point