In T395044 we introduced user agents in the QS data analytics to be able to recognize more confidently whether this is an original query that a user intended to execute (it came from QS UI, or from another web-based UI, or another client with identifiable signature) vs. a subquery (received from one of our Wikibase QSs or another endpoint with an identifiable signature).

There will of course still be a number of queries that could not be confidently placed into either of these buckets - the unknowns.

The goal of this ticket is to do this clustering into 3 buckets and estimate our level of confidence (based on the amount of 'unknowns'). This will allow us to arrive to more meaningful metrics for query federation.

Since the user-agent change was only merged today (July 10, 2025), we should wait at least a month before we can start using that data.