One of the reasons for not using Quickwit for user facing search is the latency: for example, you pay 70ms of latency when you make a request on AWS S3... and g

One of the reasons for not using Quickwit for user facing search is the latency: for example, you pay 70ms of latency when you make a request on AWS S3... and generally you expect latency below that figure. Decoupling compute and storage while keeping a very low latency may be then impossible unless ending up by caching all your data on disk :). You can have a look at lnx (https://lnx.rs/ https://lnx.rs/) that is based on tantivy and is performing quite well. It's not yet distributed but the author Chillfish8 has some thoughts about how to do it.