diff --git a/duas_query.py b/duas_query.py
new file mode 100644
index 0000000..20cdfce
--- /dev/null
+++ b/duas_query.py
@@ -0,0 +1,44 @@
+from langchain_openai import OpenAIEmbeddings
+from dotenv import load_dotenv
+from langchain_postgres import PGVector
+
+load_dotenv()
+
+CONNECTION_STRING = 'postgresql+psycopg2://postgres:test@localhost:5433/vector_db'
+COLLECTION_NAME = 'duas_tags_vectors'
+
+# Initialize embeddings (needed for query encoding only)
+embeddings = OpenAIEmbeddings(model="text-embedding-3-small")
+
+# Load existing vector store
+db = PGVector(
+    collection_name=COLLECTION_NAME,
+    connection=CONNECTION_STRING,
+    embeddings=embeddings
+)
+
+def search_duas(query, k=5):
+    results = db.similarity_search_with_score(query, k=k)
+    
+    duas_results = []
+    for doc, score in results:
+        result = {
+            'id': doc.metadata.get('id'),
+            'arabic': doc.metadata.get('arabic'),
+            'transliteration': doc.metadata.get('transliteration'),
+            'translation': doc.metadata.get('translation'),
+            'urdu': doc.metadata.get('urdu'),
+            'romanUrdu': doc.metadata.get('romanUrdu'),
+            'category': doc.metadata.get('category'),
+            'occasion': doc.metadata.get('occasion'),
+            'source': doc.metadata.get('source'),
+            'tags': doc.metadata.get('tags'),
+            'similarity_score': 1 - score
+        }
+        duas_results.append(result)
+    
+    return duas_results
+
+# Now you can search!
+results = search_duas("protection", k=2)
+print(results)
\ No newline at end of file