Reference architecture patterns and mental models for working with Large Language Models (LLM’s)