Thanks @Hari Prathap Balamurugan how do you uncover edge cases in this way? Maybe I have an irrational fear that it will malfunction or start saying something it shouldn't because I haven't actually deployed one before. My use case is an AI receptionist for a pest company that answers FAQs and scheudles clients into the businesses GHL calendar and transfers the call to the owner if it cant resolve the query. In your experience what has the greatest affect on performance? Is it the quality of the system prompt?