ParentFull threadakvadrako·Can't you use fine-tuning for this?A other option is to ask GPT to compress your tokens into a shorter prompt for itself.View on HN