Last released Nov 20, 2026
A full-duplex voice framework whose turn-taking is a set of plain-language decisions judged by Jev, not a trained model.