FTS5 tokenizer mode. unicode61 (default) tokenizes on Unicode word boundaries with diacritic folding — good fit for natural-language markdown. trigram indexes every 3-char substring — slower to build but better recall on CJK / agglutinative scripts.
unicode61
trigram
FTS5 tokenizer mode.
unicode61(default) tokenizes on Unicode word boundaries with diacritic folding — good fit for natural-language markdown.trigramindexes every 3-char substring — slower to build but better recall on CJK / agglutinative scripts.