I removed the vector database from my AI agent stack (github.com)

Moss has launched a groundbreaking semantic search runtime designed specifically for Conversational AI agents, boasting retrieval times under 10 milliseconds. Unlike traditional AI stacks, which typically rely on remote vector databases that introduce latency due to network calls, Moss integrates retrieval and embedding directly within the local process. This innovation significantly improves response times, making it critical for real-time applications, such as voice bots or customer support agents, where speed and fluidity are paramount. Key features of Moss include hybrid search capabilities combining semantic and keyword querying, built-in embedding models that eliminate dependency on external services, and extensive metadata filtering functionalities. The runtime can be easily embedded into applications via a simple SDK compatible with multiple programming languages including Python, TypeScript, and C. Notably, Moss supports a WebAssembly build that enables client-side queries without server reliance, which further enhances flexibility for developers. Overall, Moss represents a paradigm shift in conversational AI, promising to create more responsive and engaging user experiences.