So initially I started writing a service mesh (for fun), then everyone started talking about AI workloads. So here we are, a transparent gateway that can be applied to running workloads (no restart) that allows control between agents/workloads and an LLM.
kube-gateway.io