off-heap
Zero-infrastructure off-heap cache for AI workloads. One annotation in Java, one decorator in Python. No Redis, no Docker.