It might be necessary to not only search on what the client side has, but also do some sort of backend search by using some caching service that agregates content from majority of public relay's data. Something along the lines of what npub16c0nh3dnadzqp… and npub12vkcxr0luzwp8… are intending on doing. For backend side, elasticsearch is a proven and great tool which uses lucene underneath. AFAIW very few relays implement searching, probably with time more will and more and more effectively. But the interesting thing is do they just search among the content users posted to them or do they aggregate content from other relays content as well and do a more widespread searching....?
