There are plans, but right now there are not enough resources to push this task forward.
This should be as simple as storing messages indexed by thread-id in the additional table (column family) in Cassandra, e.g.:
thread-id-1 : { message-uuid-1, ..., message-uuid-n },
thread-id-2 : { ... },
...
}
Messages can be added to thread indexes as they arrive and you can retrieve list of conversations through new API endpoint. The main problem I see here is logic for grouping threads - it's not straightforward as one may think.
-
Rustam.