Neat! I appreciate your approach to preventing hallucinations. I've used something similar in a different context. People make a big deal about hallucinations but I've found that validation is one of the easier aspects of AI architecture.