Tauri Plugin Lingua

A Tauri plugin for language detection using lingua-rs.

Features

• Detect language from text with high accuracy
• Support for 75 languages
• Confidence scoring for language detection
• Works with both long and short text fragments
• Thread-safe detector instances
• Works on both desktop and mobile platforms

Installation

Rust

Add to your Cargo.toml:

[dependencies]
tauri-plugin-lingua = "0.2.0"

Compiling a Language Subset

By default all 75 languages are compiled in, which maximizes binary size and compile time. Apps that only need a handful of languages can opt in instead:

[dependencies]
tauri-plugin-lingua = { version = "0.2.0", default-features = false, features = [
  "english",
  "chinese",
  "japanese",
] }

Feature names match the lingua language features (english, french, chinese, ...).

With a subset compiled in, createDetectorForAllLanguages detects exactly the compiled-in languages, and creating a detector for a language that is not compiled in returns an InvalidLanguageCode error instead of crashing.

JavaScript

Install using pnpm:

pnpm add tauri-plugin-lingua-api

Usage

Initialize Plugin

In your Tauri app's main.rs or lib.rs:

fn main() {
    tauri::Builder::default()
        .plugin(tauri_plugin_lingua::init())
        .run(tauri::generate_context!())
        .expect("error while running tauri application");
}

TypeScript API

import {
  createDetectorForAllLanguages,
  createDetectorForLanguages,
  detectLanguage,
  computeLanguageConfidence,
  computeLanguageConfidenceValues,
  type LanguageDetector,
  type LanguageConfidence,
  type CreateDetectorOptions,
} from 'tauri-plugin-lingua-api';

// Create a detector for all 75 supported languages
const detector = await createDetectorForAllLanguages();

// Or create a detector for specific languages (comma-separated ISO 639-1 codes)
const customDetector = await createDetectorForLanguages('en,fr,de,es,zh');

// Create a detector with minimum relative distance option
// This returns null when the top two language confidences differ by less than 0.1 (10%)
// Default is -1 (disabled), which always returns the most likely language
const strictDetector = await createDetectorForAllLanguages({
  minimumRelativeDistance: 0.1
});

// Detect the language of a text
const language = await detectLanguage(detector, "Hello, how are you?");
console.log(language); // "EN"

// Detect language with null handling
const lang = await detectLanguage(detector, "Hi");
if (lang) {
  console.log(`Detected language: ${lang}`);
}

// With minimumRelativeDistance set, ambiguous text returns null
const ambiguous = await detectLanguage(strictDetector, "Hi");
console.log(ambiguous); // null (if confidence difference is less than threshold)

// Get confidence score for a specific language
const confidence = await computeLanguageConfidence(
  detector,
  "Bonjour le monde",
  "FR"
);
console.log(confidence); // e.g., 0.95

// Get confidence values for all languages
const confidences = await computeLanguageConfidenceValues(
  detector,
  "Hola mundo"
);
// Returns: [
//   { language: "ES", confidence: 0.98 },
//   { language: "PT", confidence: 0.45 },
//   ...
// ]

API Reference

Types

LanguageDetector

interface LanguageDetector {
  id: string;
}

LanguageConfidence

interface LanguageConfidence {
  language: string;  // ISO 639-1 code (e.g., "EN", "FR")
  confidence: number; // 0.0 to 1.0
}

CreateDetectorOptions

interface CreateDetectorOptions {
  /** Minimum relative distance between confidence values.
   *  Default: -1 (disabled)
   *  Range: -1 or 0.0-0.99
   *  When set to a value between 0.0-0.99, language detection returns null if
   *  the difference between the top two language confidences is smaller than
   *  this value. Set to -1 to always return the most likely language. */
  minimumRelativeDistance?: number;
}

Functions

createDetectorForAllLanguages(options?)

Creates a language detector for all 75 supported languages.

Parameters:

• options (optional): CreateDetectorOptions object
  • minimumRelativeDistance: Number between -1 and 0.99. Default is -1 (disabled). When set to 0.0-0.99, detectLanguage returns null if the difference between the top two language confidences is smaller than this value. Useful for filtering out ambiguous results.

Returns: Promise<LanguageDetector>

Example:

// Standard detector (minimumRelativeDistance: -1, feature disabled)
const detector = await createDetectorForAllLanguages();

// Detector with minimum relative distance enabled
const strictDetector = await createDetectorForAllLanguages({
  minimumRelativeDistance: 0.1  // 10% minimum difference required
});

createDetectorForLanguages(languages: string, options?)

Creates a language detector for specific languages.

Parameters:

• languages: Comma-separated ISO 639-1 language codes (e.g., "en,fr,de")
• options (optional): CreateDetectorOptions object
  • minimumRelativeDistance: Number between -1 and 0.99. Default is -1 (disabled).

Returns: Promise<LanguageDetector>

Throws: Error if invalid language codes are provided

detectLanguage(detector: LanguageDetector, text: string)

Detects the language of the given text.

Parameters:

• detector: A language detector instance
• text: The text to analyze

Returns: Promise<string | null> - ISO 639-1 code or null if language cannot be detected

computeLanguageConfidence(detector: LanguageDetector, text: string, languageCode: string)

Computes the confidence score for a specific language.

Parameters:

• detector: A language detector instance
• text: The text to analyze
• languageCode: ISO 639-1 code of the language to check

Returns: Promise<number> - Confidence score between 0.0 and 1.0

computeLanguageConfidenceValues(detector: LanguageDetector, text: string)

Computes confidence scores for all languages.

Parameters:

• detector: A language detector instance
• text: The text to analyze

Returns: Promise<LanguageConfidence[]> - Array of language-confidence pairs, sorted by confidence (highest first)

Supported Languages

The plugin supports 75 languages via ISO 639-1 codes:

AF (Afrikaans), SQ (Albanian), AR (Arabic), HY (Armenian), AZ (Azerbaijani), EU (Basque), BE (Belarusian), BN (Bengali), NB (Norwegian Bokmal), BS (Bosnian), BG (Bulgarian), CA (Catalan), ZH (Chinese), HR (Croatian), CS (Czech), DA (Danish), NL (Dutch), EN (English), EO (Esperanto), ET (Estonian), FI (Finnish), FR (French), LG (Ganda), KA (Georgian), DE (German), EL (Greek), GU (Gujarati), HE (Hebrew), HI (Hindi), HU (Hungarian), IS (Icelandic), ID (Indonesian), GA (Irish), IT (Italian), JA (Japanese), KK (Kazakh), KO (Korean), LA (Latin), LV (Latvian), LT (Lithuanian), MK (Macedonian), MS (Malay), MI (Maori), MR (Marathi), MN (Mongolian), NN (Norwegian Nynorsk), FA (Persian), PL (Polish), PT (Portuguese), PA (Punjabi), RO (Romanian), RU (Russian), SR (Serbian), SN (Shona), SK (Slovak), SL (Slovene), SO (Somali), ST (Sotho), ES (Spanish), SW (Swahili), SV (Swedish), TL (Tagalog), TA (Tamil), TE (Telugu), TH (Thai), TS (Tsonga), TN (Tswana), TR (Turkish), UK (Ukrainian), UR (Urdu), VI (Vietnamese), CY (Welsh), XH (Xhosa), YO (Yoruba), ZU (Zulu)

License

Same as lingua-rs - Apache 2.0