Search the caller’s bubbles by keyword, returning matching bubble summaries ranked by relevance. Matches fuzzily (prefix/typo-tolerant) against each bubble’s title, its author’s name, and its AI-generated meeting notes/summary — so a keyword that appears only inside the notes still finds the bubble; title matches rank highest, then author, then notes. Requires current Home, visit, or visible-Space discovery in addition to read access — a bubble the caller cannot read is never returned (a notes-only match on an unreadable bubble never surfaces). Each result row includes the bubble’s AI meeting-notes/summary text (summary, capped at 1000 characters; null when there are no notes) so an agent can read why it matched inline. q is required (non-empty, ≤256 chars); an empty or overlong q yields 400 invalid_request. Supports the same team_id/space_id/type/created_after/created_before filters as the list endpoint. Search pagination covers authorized matches within the first 1,000 fuzzy-ranked candidates; has_more and next_cursor refer to that window. Pagination is by opaque cursor (relevance-ordered); rows exclude transcripts and comments (fetch a bubble by id for those).