Effort is a dial, not a plan
WonderSearch asks a small question each time you search: how much room should this query have to explore? You choose Small, Medium, or Large per query, so one product can serve a quick lookup and a harder question side by side.
The price follows the choice. Search compute is $0.001 per query for Small, $0.0025 for Medium, and $0.01 for Large. Storage is billed separately at $0.33 per GB per month, whatever effort you use.
Because the decision is made per query, you can start simple and adjust where it matters, instead of paying for the hardest case on every request.
Match the effort to the question
Small suits everyday lookups: a policy name, a product detail, a question with a clear answer somewhere in the collection. Medium gives a question with more context a closer look. Large gives the hardest questions more room to explore.
A practical pattern is to default to the smallest size that answers your common questions well, and reserve larger sizes for the places where a better result changes what the user does next: a support escalation, a research task, or context for an AI assistant that will act on what it reads.
If an assistant runs several searches to answer one message, decide the effort for each of those searches. The cost of an answer is the sum of the searches behind it.
Measure before you settle
The right size depends on your collection and your users, so test it. Take a set of representative questions, run them at each size, and read the passages that come back. Note where a larger size finds something a smaller one missed, and where it makes no difference.
Our benchmarks page shows how Small, Medium, and Large compare on public and internal retrieval datasets. Use it to orient your decision, then let your own questions make it.