Kafka doesn't have fast random-access to a row-tuple in a stream by its primary key, let alone declarative indexing by compound keys / computed expressions.

Kafka with those things would be equivalent to what I'm describing, yes.