vLLM’s reference system for K8S-native cluster-wide deployment with community-driven performance optimization - View it on GitHub