Semantic and intent-based memory for AI SDK. Supports one-shot semantic caching and multi-turn intent-aware conversations. Reduces token costs by reusing similar responses.