Why not just use ChatGPT roleplay?
ChatGPT produces conversations. vLeader produces consequences — identical for every student, auditable, and usable in Assurance-of-Learning reporting. A language model can give different, sometimes fabricated feedback to every student; deterministic rules can't.