This library is proper fast. Don't get distracted comparing this to other libraries just on the basis of similar string template syntax; it doesn't have a dependency on pg because it implements its own low level client.
I've built systems on this loading multiple 10k records at time and it crushes.
Can load 10-20k records(ymmv) in 20ms or under where it would otherwise take 150ms. That can be a game changer based on use case by itself, but if you are used to thinking about what's viable in Ruby or etc it's even more inspiring.
It also supports logical decode which opens a lot of doors for processing transitional events.
I've built systems on this loading multiple 10k records at time and it crushes.