reflection

Very cool project, I think you could even improve the quality by taking every output and then saying to GPT 3.5 Turbo “refine the quality, consistency, style, depth, helpfulness, …. Of the given text / hotel reservation chat / …”

From my experience with playing with GPT giving the model a second chance to improve the output improve the quality, basically letting it to reflect upon itself.