Use RabbitMQ Streams bloom-filter message filtering to let consumers selectively read a subset of a stream

Steps

1. Publish messages with a filter value set via the x-stream-filter-value message annotation (a string chosen by the publisher, e.g. a tenant ID or event type)
2. RabbitMQ aggregates the filter values of all messages in a chunk into a Bloom filter stored in that chunk's header before writing it to disk
3. When attaching a consumer, set the x-stream-filter consumer argument to one string or an array of strings (array values are OR'd together) to request only chunks whose Bloom filter may contain a match
4. Optionally tune the Bloom filter size with x-stream-filter-size-bytes (16-255 bytes, default 16) for higher precision at the cost of chunk header size
5. Implement client-side filtering on top of server-side filtering, since the Bloom filter operates at the whole-chunk level and can yield false positives
6. Consider the newer broker-side SQL filter expressions (introduced 2025) as an alternative/complement for expression-based filtering on message properties