Full-text search

Mark a text column .searchable(). When you create the tables, forge builds the right full-text index for each database:

Dialect	What forge emits
Postgres	CREATE INDEX … USING gin(to_tsvector('simple', col))
MySQL	ALTER TABLE … ADD FULLTEXT(col) (plus WITH PARSER ngram / mecab if requested)
SQLite	CREATE VIRTUAL TABLE <table>_fts USING fts5(...) + insert/update/delete triggers
DuckDB	PRAGMA create_fts_index(...) (fts extension)
Mongo	createIndex({ col: 'text' })
MSSQL	Manual — requires a full-text catalog (out-of-band). Schema field is kept.

Then query it with the search operator. The operator is portable; the ranking and tokenisation behind it is whatever the dialect provides.

const Post = model('posts', {
  id:   f.id(),
  body: f.text().searchable(),
});

await db.post.findMany({ where: { body: { search: 'database wrapper' } } });

On SQLite, queries through the shadow FTS5 table are joined automatically — you don't have to know it exists. On Postgres, the index is over to_tsvector('simple', col); you can pass a custom configuration via a raw expression: index for language-specific stemming.

See more — docs/FTS.md for every dialect's FTS engine, ranking score retrieval, multi-column composition, languages and analyzers, hybrid BM25 + vector (RRF), highlighting, and six worked patterns.

---