The files the build (note: Build) writes under search/ so that a page searches (note: Search) the whole corpus (note: Corpus) without a server: an entity table (5 passages, no note) and one shard per two-character (3 passages, no note) prefix (17 passages, no note), holding the words of every entity, its title, aliases (11 passages, no note), summary (24 passages, no note), note text, type (note: Type), application, domain (3 passages, no note), status (8 passages, no note) and source, each with a weight. The page loads the table (40 passages, no note) when the search field (37 passages, no note) takes focus (8 passages, no note) and one shard (6 passages, no note) per word typed (37 passages, no note), so that a query (note: Query) matches (note: Occurrence) by prefix as the reader types (4 passages, no note), over file:// as behind a server.
Not to be confused with
The alphabetical index, the page that lists every word by initial letter (17 passages, no note); the search index is a set of files no reader (note: Reader) opens.