modelrelay

Central HTTP LLM proxy that routes inference requests to authenticated remote workers over WebSocket — queueing, streaming, and cancellation included.