LMCache

Open-source KV-cache management layer for scalable LLM inference that persists, reuses, transfers, and observes KV cache across inference engines and storage tiers.