Routed: Local, zero-token hybrid router for AI agent skills (<20ms) (github.com)

Routed has launched a local, zero-token hybrid router designed to optimize skill routing for AI agents, operating in under 20 milliseconds. This innovative tool eliminates the need for costly long language model (LLM) routing calls by automatically indexing and routing coding prompts to the most relevant skills through a hybrid search engine that integrates exact matching, Okapi BM25, and dense semantic embeddings, all performed locally without reliance on cloud services. The significance of Routed lies in its combination of speed, cost-effectiveness, and privacy. By processing requests entirely on the user's machine, it significantly reduces latency compared to traditional cloud-based solutions, which can take up to 3,500 milliseconds, and alleviates privacy concerns since no prompts leave the local system. Routed is also capable of parsing complex prompts to activate multiple skills simultaneously, enhancing the functionality of AI agents in coding environments while integrating smoothly with various platforms through easy installation and setup processes.