Officially Supported Languages

Right now, ZBSearch supports 34 languages out of the box in 8 different alphabets.
For every language, ZBSearch provides a default tokenizer, stop-words, and stemmer.

Multilingual mode (zero-config)

If you don't know the language of your documents ahead of time - or a single index mixes several languages - use multilingual instead of a specific language:

import { create } from 'zbsearch'

const db = create({
  schema: { title: 'string', content: 'string' },
  language: 'multilingual'
})

In this mode ZBSearch tokenizes with Intl.Segmenter (Unicode word segmentation, with a Unicode regex fallback on runtimes that lack it), so Latin, Cyrillic, Greek, Arabic, Hebrew, Indic, and CJK text all index correctly with no per-language setup. Tokens are lowercased, and diacritics are folded so that café matches cafe, ёлка matches елка, and آلاف matches الاف.

Trade-offs compared to a per-language configuration:

• No stemming and no stop-words by default. Queries for inflected forms (running vs. run) won't match unless you pass a custom stemmer function. Per-language installs with @zbsearch/stemmers remain the quality ceiling for single-language content: in the quality benchmark that ships with the repo (benchmarks/, npm run benchmark:multilingual-quality), the multilingual mode reaches ~0.70 recall@10 vs ~0.98 for tuned per-language installs, with most of the gap on inflection queries - while non-Latin scripts go from 0 (English default) to fully searchable.
• CJK segmentation is script-based, not dictionary-based. It works well with prefix search, but for production Japanese or Chinese search prefer the dedicated tokenizers.
• Sorting falls back to the runtime's default collation, since there is no single locale.

For multilingual sites (one locale per document), the recommended pattern is to keep a single multilingual index and store the locale as an enum, then filter at query time:

const db = create({
  schema: { title: 'string', content: 'string', locale: 'enum' },
  language: 'multilingual'
})

await search(db, { term: 'getting started', where: { locale: { eq: 'it' } } })

Latin Alphabet

(*) = also uses a few additional diacritic marks
(**) = uses both Cyrillic and Latin scripts

Cyrillic Alphabet

(*) = uses both Cyrillic and Latin scripts

Greek Alphabet

Devanagari Script

Arabic Script

Armenian Alphabet

Tamil Script

Chinese Characters (Logographic Script)