Harnessing LLM's For Hard Gameplay Dynamics

I would like to demo the progress I’ve made in constructing a RAG and game side logic that translates player’s inputs into use-able, & saveable variables and states. The game I’m making is a detective game where you can say what you want to witnesses, and subjects to get the information you need to solve the case. In the demo I will demonstrate questioning a suspect. The suspect will lie to the player until the player exposes his lie by introducing proof of his lie. Then the NPC will then change their state and begin telling the truth about the subject originally asked.