docs/UX: search limit hard-capped at 100 — over-limit reads as empty results, not an error

Summary

The search limit parameter is hard-capped at 100, and exceeding it fails as a Pydantic validation error rather than clamping — from the client's side the call simply returns nothing useful, which reads as "no results," not "bad request."

How it bit us

A recall path over-fetched a candidate pool (limit = k * 10) for client-side dedup. It worked for k <= 10 and silently returned [] for k >= 11 — shipped that way, caught only when a later consumer used a bigger k (fixed in Smart-AI-Memory/attune-ai#666 by clamping to 100 client-side).

Ask

• Document the cap on limit (client + REST docs).
• Consider either clamping server-side or surfacing the validation error in a way the client can't mistake for an empty result set.

Server + client 0.14.0. Happy to PR a docs note.

Start with the client and REST documentation for the search limit parameter, then trace the server-side validation behavior for values above 100. Document the 100-item cap and the observed over-limit response, and verify the documentation against a request using limit > 100; clamping or improved error handling is an optional follow-up.