Moss is a high-performance retrieval system designed for production AI systems, providing sub-10ms semantic search without needing external vector databases. It enables developers to perform local, low-latency document retrieval that integrates directly into voice agents, copilots, and edge applications.