A custom operator that scales pods based on real-time websocket latency metrics rather than just CPU usage.