This can be a good fit for something like SimpleNote. Looking at the data structure since it’s not persisted I am a little skeptical on memory usage; one good solution could be abstract out the interface to write adapters. This can allow storages like IndexDB, or NodeJS with LevelDB etc. Would be really nice if you would have show some memory usage with benchmarks.
The memory usage can definitely go higher as the number of documents increase. It is pretty good at compressing prefixes, but suffixes don’t get the same compression. The actual API allows you to store and load the index as needed, but I might look into database adapters.