The size of the library is not the number of readers
A collection of documents can be valuable even when nobody is searching it. A support team needs old policies available. An internal assistant needs reference material that may be useful only a few times a month. Keeping that knowledge around should be a deliberate storage decision.
Search activity has a different rhythm. A launch can bring a burst of questions. A quiet weekend can bring almost none. Thinking about storage and search compute separately makes it easier to understand what changes when your product grows.
With WonderSearch, search compute is charged per query. Storage is billed separately at $0.33 per GB per month. When the queries stop, the per-query compute charge stops accumulating. Keeping data stored still has a cost.
Make the arithmetic easy to inspect
Small searches cost $0.001 per query, Medium searches cost $0.0025, and Large searches cost $0.01. At 100 GB and 1,000 Small searches in a month, the usage estimate is $34: $33 for storage and $1 for search compute.
Keep the collection the same and run 10,000 Small searches, and that usage estimate becomes $43. Keep the collection but run no searches, and the storage portion remains $33 while search compute is $0. These examples assume the data is stored for the full month and exclude taxes, subscription fees, credits, and discounts.
The separation is the useful part. You can see the cost of retaining knowledge and the cost of asking questions without folding the two into a single mysterious number.
Budget around the experience
A search budget should reflect the way your product behaves. Estimate monthly queries, choose a search size for the task, and account for the data you expect to retain. If an assistant makes several retrieval requests to answer one user question, budget for those requests rather than counting only the message the user typed.
Then try a few scenarios. What happens after a successful launch? What happens in a quiet month? How much of the estimate comes from storage? The calculator is there to make those assumptions visible, so you can change them before they become a bill.
For larger requirements, the next step is a conversation. Bring the shape of your workload and the experience you want to build. A useful pricing discussion starts with those details, not with a guess about how many servers your customers might eventually need.