RT @guohao_li: This is promising for solving one of the most annoying problems in LLM - getting structured outputs. As one of the earliest…