diff --git a/mcp-servers/knowledge-base/database.py b/mcp-servers/knowledge-base/database.py
index 41c2da5..551037d 100644
--- a/mcp-servers/knowledge-base/database.py
+++ b/mcp-servers/knowledge-base/database.py
@@ -57,6 +57,9 @@ class DatabaseManager:
     async def initialize(self) -> None:
         """Initialize connection pool and create schema."""
         try:
+            # First, create pgvector extension (required before register_vector in pool init)
+            await self._ensure_pgvector_extension()
+
             self._pool = await asyncpg.create_pool(
                 self._settings.database_url,
                 min_size=2,
@@ -66,7 +69,7 @@ class DatabaseManager:
             )
             logger.info("Database pool created successfully")
 
-            # Create schema
+            # Create schema (tables and indexes)
             await self._create_schema()
             logger.info("Database schema initialized")
 
@@ -77,6 +80,19 @@ class DatabaseManager:
                 cause=e,
             )
 
+    async def _ensure_pgvector_extension(self) -> None:
+        """Ensure pgvector extension exists before pool creation.
+
+        This must run before creating the connection pool because
+        register_vector() in _init_connection requires the extension to exist.
+        """
+        conn = await asyncpg.connect(self._settings.database_url)
+        try:
+            await conn.execute("CREATE EXTENSION IF NOT EXISTS vector")
+            logger.info("pgvector extension ensured")
+        finally:
+            await conn.close()
+
     async def _init_connection(self, conn: asyncpg.Connection) -> None:  # type: ignore[type-arg]
         """Initialize a connection with pgvector support."""
         await register_vector(conn)
@@ -84,8 +100,7 @@ class DatabaseManager:
     async def _create_schema(self) -> None:
         """Create database schema if not exists."""
         async with self.pool.acquire() as conn:
-            # Enable pgvector extension
-            await conn.execute("CREATE EXTENSION IF NOT EXISTS vector")
+            # Note: pgvector extension is created in _ensure_pgvector_extension()
 
             # Create main embeddings table
             await conn.execute("""
@@ -286,7 +301,14 @@ class DatabaseManager:
         try:
             async with self.acquire() as conn, conn.transaction():
                 # Wrap in transaction for all-or-nothing batch semantics
-                for project_id, collection, content, embedding, chunk_type, metadata in embeddings:
+                for (
+                    project_id,
+                    collection,
+                    content,
+                    embedding,
+                    chunk_type,
+                    metadata,
+                ) in embeddings:
                     content_hash = self.compute_content_hash(content)
                     source_path = metadata.get("source_path")
                     start_line = metadata.get("start_line")
@@ -397,7 +419,9 @@ class DatabaseManager:
                         source_path=row["source_path"],
                         start_line=row["start_line"],
                         end_line=row["end_line"],
-                        file_type=FileType(row["file_type"]) if row["file_type"] else None,
+                        file_type=FileType(row["file_type"])
+                        if row["file_type"]
+                        else None,
                         metadata=row["metadata"] or {},
                         content_hash=row["content_hash"],
                         created_at=row["created_at"],
@@ -476,7 +500,9 @@ class DatabaseManager:
                         source_path=row["source_path"],
                         start_line=row["start_line"],
                         end_line=row["end_line"],
-                        file_type=FileType(row["file_type"]) if row["file_type"] else None,
+                        file_type=FileType(row["file_type"])
+                        if row["file_type"]
+                        else None,
                         metadata=row["metadata"] or {},
                         content_hash=row["content_hash"],
                         created_at=row["created_at"],