You can actually use either, it won’t matter for the flow itself. The key is that on your side (Next.js app) you handle both inputs—voice or text—through the same conversation context so the agent logic continues smoothly. If you want, I can show the exact setup so you know how to route both correctly. @!