An admin considers adding a second individual retriever to make an existing retriever into an ensemble for Ridgeline's insurance-FAQ agent. What is a reasonable trade-off to weigh before doing so?
Select an answer to reveal the explanation.
Short Explanation
More retrievers searching more sources can mean better coverage, but it isn't free — it's worth adding an ensemble where it actually improves answers, not sprinkling it on everything by habit.
Full Explanation
Combining retrievers into an ensemble can genuinely broaden coverage and improve answer completeness, but querying additional sources is not free — it can add retrieval overhead, so a thoughtful admin evaluates whether the accuracy gain for a specific use case, like Ridgeline's insurance-FAQ agent, is worth that cost rather than adding ensemble retrieval by default everywhere regardless of need. Claiming there is literally no trade-off at all ignores that querying more sources is inherently more work than querying one, even if the platform handles it efficiently. There's no rule banning ensemble retrievers for FAQ-style use cases — in fact, an FAQ agent spanning multiple related document sets is a fairly natural candidate for ensemble retrieval when a single source doesn't reliably cover every question. Adding a second retriever to form an ensemble also doesn't disable the first — an ensemble is additive, combining multiple retrievers' outputs together, not swapping one out for another. For the exam, understand that configuration decisions like this involve weighing accuracy benefits against reasonable overhead, not treating every option as either free or forbidden.