High-speed semantic vector cache for multi-agent swarms. Redis-backed similarity lookup with in-memory fallback before expensive embedding pipelines.