This is an awkward feature. Are those typed keys added client or server side? Can you get two aggregations with the same name and different type? Either way giving the option make sense, want to try to implement it @fs-chris?

Long story is that I'd prefer a non-ambiguous response where the key is always the same, and that the body of the aggregation response contains either the separate name of the aggregation and the type, or vice-versa.

I suppose (without having looked into the code) that this is added by the client. The client prefixes the aggregation's name in the request which then occurs identically in the response, since the server does not touch the aggregation's name.
So, this simply preserves the type information beyond the request/response round trip.

In the response it is then used for determining the target type before parsing the aggregation's result. I.e., the aggregation's key contains the target type that is used to decide how to parse the actual result body.

Hi, is there a way to disable this mechanism in opensearch-java? I got similar problem that type name breaks client applications when migrating from Elasticsearch library to opensearch-java library.

@chanon-onman I am currently investigating this