From the [2026 Q3 roadmap](https://github.com/AMD-AGI/Infera/issues/9). **Today:** Routed on load only (no cache locality). **Goal:** MM-aware routing; encode / prefill / decode (EPD) separation. - MM-aware routing (image-affinity) landed in **#35** (Rust + Python routers). Remaining: encode/prefill/decode (EPD) separation for multimodal serving.
From the 2026 Q3 roadmap.
Today: Routed on load only (no cache locality).
Goal: MM-aware routing; encode / prefill / decode (EPD) separation.