Routing diffusion inference traffic across three providers

TL;DR: We route a mix of diffusion and LLM traffic across three providers from a single Go-based...