LMCache releases Coordinator control plane for fleet KV cache

LMCache has released a coordinator that acts as a control plane for fleet-wide KV cache in LLM inference workloads. Operating off the request path, the opt-in process aggregates cache events from standalone multi-process servers to provide fleet-wide visibility, quotas, evictions, pins, and prefetching. It features a pluggable controller architecture that lets developers customize policies without modifying inference engines or data paths. Open article →