diff --git a/engine/preprocessing.py b/engine/preprocessing.py
index 4234beb..19d4d4b 100644
--- a/engine/preprocessing.py
+++ b/engine/preprocessing.py
@@ -1,20 +1,11 @@
-import re
-
-def normalize_query(query: str) -> str:
+def preprocess_query(query: str) -> str:
     """
-    Normalize user query before retrieval:
-    - Lowercase conversion
-    - Remove unnecessary punctuation
-    - Remove extra whitespace
+    Basic query preprocessing:
+    - Lowercases text
+    - Strips extra whitespace
     """
+    if not query:
+        return ""
 
-    # Convert to lowercase
-    query = query.lower()
-
-    # Remove punctuation (keep alphanumeric and spaces)
-    query = re.sub(r"[^\w\s]", "", query)
-
-    # Remove extra whitespace
-    query = re.sub(r"\s+", " ", query).strip()
-
+    query = query.lower().strip()
     return query
\ No newline at end of file
diff --git a/engine/rag_engine.py b/engine/rag_engine.py
index 2e6caf5..dbd099e 100644
--- a/engine/rag_engine.py
+++ b/engine/rag_engine.py
@@ -14,6 +14,7 @@
 import streamlit as st
 import numpy as np
 import logging
+from engine.preprocessing import preprocess_query
 logger = logging.getLogger(__name__)
 
 try:
@@ -202,9 +203,6 @@ def _emb_search(query: str, top_k: int = 3):
     try:
         model = load_embedding_model()
         
-        # --- USE CACHED MODEL HERE ---
-        model = load_embedding_model()
-        
         qvec = model.encode([query], convert_to_numpy=True)[0]
 
         scores = []
@@ -262,6 +260,7 @@ def _keyword_search(query: str, top_k: int = 3):
     return results[:top_k]
 
 def search_pdfs(query: str, top_k: int = 3):
+    query = preprocess_query(query)
     """
     Hybrid Retrieval:
     - Runs vector search (if available)