Hey guys, I'm working on a Chinese writing project attempting to divide a chapter (approximately 2000 Chinese characters) into different scenes (much like how an act in a movie has multiple scenes). Each scene is scripted by the author (human) with the simplest scene description. Then, I'll have an LLM expand and polish each scene, and finally string them together into one chapter. Since I don't have a fine-tuned LLM, I want a bot to process each script separately based on the user-specified imitation style (a short example) and preset prompts for expansion and polishing. To save tokens, I plan to store different style examples in a knowledge base or database for the bot to call directly based on user instructions. I'm pondering whether this is a good approach. Or does anyone have better ideas?