Open-source real-time voice AI agent framework by the TEN Framework — build sub-second multimodal voice agents with tool use, vision, and avatar support.