@QiaochuYuan well i feel like the current shortcoming IS #4 - the rate limits make it often unavailable, plus downtime, plus changing out models ... open local models could fix this but they aren't as good yet