I'm trying to use a conversational bot which can assist me to use LLMs to find certain metrics which user has to manually enter from a big form on my platform. For example, the user has to manually select between 20 metrics from Facebook, 20 metrics from TikTok and so on. Idea is now the user can just type: "Tell me more about my ad spent over the last month". I have a prompt generated so GPT narrows down the important metrics in a json format. My question is can I just train a LLM model on botpress with all these metrics so I don't have to pass it every time in the prompt? There are more than 100 metrics, so I don't want to keep passing them again and again.

Why not use codeblocks for each set of metrics? For example - Option 1 is "Ads Metric Set" Clicking on it would pass metric parameters 1,4, and 5 to GPT. Make loads of these.