Hey everyone, running into something frustrating and wondering if others feel the same. My voice agent's quality feels inconsistent. Means: sometimes it handles a scenario perfectly, other times it behaves completely differently on what seems like the same input. And after model updates I notice shifts in behavior I didn't expect. The part that really bothers me: when something goes wrong, I can't reliably reproduce it. I can't just replay the exact conditions! I cant imitate accent, background noise, how the caller phrased it. No chance for me to figure out what broke and why. So I end up either shipping and hoping, or manually calling my own agent over and over trying to catch edge cases. (not really scalable!) Anyone else dealing with this? Or am I just doing something wrong here?