diff --git a/surfsense_backend/alembic/versions/73_add_user_memories_table.py b/surfsense_backend/alembic/versions/73_add_user_memories_table.py
index 40ecfd91b..c0a1fb97e 100644
--- a/surfsense_backend/alembic/versions/73_add_user_memories_table.py
+++ b/surfsense_backend/alembic/versions/73_add_user_memories_table.py
@@ -12,7 +12,6 @@ about users across conversations.
 from collections.abc import Sequence
 
 from alembic import op
-
 from app.config import config
 
 # revision identifiers, used by Alembic.
diff --git a/surfsense_backend/app/agents/new_chat/tools/user_memory.py b/surfsense_backend/app/agents/new_chat/tools/user_memory.py
index 540189a6f..30ca4503e 100644
--- a/surfsense_backend/app/agents/new_chat/tools/user_memory.py
+++ b/surfsense_backend/app/agents/new_chat/tools/user_memory.py
@@ -10,7 +10,6 @@ Features:
 """
 
 import logging
-from datetime import UTC, datetime
 from typing import Any
 from uuid import UUID
 
@@ -179,44 +178,23 @@ def create_save_memory_tool(
             # Generate embedding for the memory
             embedding = config.embedding_model_instance.embed(content)
 
-            # Convert numpy array to list of Python floats for PostgreSQL
-            import numpy as np
-            if isinstance(embedding, np.ndarray):
-                embedding_list = embedding.tolist()
-            else:
-                embedding_list = list(embedding)
-
-            # Create new memory using ORM with proper enum handling
-            # Use the enum's value attribute directly
-            from sqlalchemy import text as sql_text
-            
-            now = datetime.now(UTC)
-            
-            # Use raw SQL with proper parameter binding for asyncpg
-            insert_sql = sql_text("""
-                INSERT INTO user_memories (user_id, search_space_id, memory_text, category, embedding, updated_at, created_at)
-                VALUES (:user_id, :search_space_id, :memory_text, CAST(:category AS memorycategory), :embedding, :updated_at, :created_at)
-                RETURNING id
-            """)
-            
-            result = await db_session.execute(
-                insert_sql,
-                {
-                    "user_id": uuid_user_id,
-                    "search_space_id": search_space_id,
-                    "memory_text": content,
-                    "category": category,  # Already lowercase string
-                    "embedding": str(embedding_list),  # Convert to string format for pgvector
-                    "updated_at": now,
-                    "created_at": now,
-                }
+            # Create new memory using ORM
+            # The pgvector Vector column type handles embedding conversion automatically
+            new_memory = UserMemory(
+                user_id=uuid_user_id,
+                search_space_id=search_space_id,
+                memory_text=content,
+                category=MemoryCategory(category),  # Convert string to enum
+                embedding=embedding,  # Pass embedding directly (list or numpy array)
             )
-            new_memory_id = result.scalar_one()
+            
+            db_session.add(new_memory)
             await db_session.commit()
+            await db_session.refresh(new_memory)
 
             return {
                 "status": "saved",
-                "memory_id": new_memory_id,
+                "memory_id": new_memory.id,
                 "memory_text": content,
                 "category": category,
                 "message": f"I'll remember: {content}",