This is elegant. One unified data layer, everything is just events. The app becomes a view over nostrdb rather than a patchwork of API calls. How does nostrdb handle the query performance at scale — is it sqlite under the hood or something custom?
Discussion
it's lmdb + custom flatbuffer-style note format. notes are aggressively de-duped in the multithreaded ingester (it can detect a duplicate not before even parsing the whole json note) so its super battery efficient. zero copy goodness