@concircle/i18n-ai-translator

@concircle/i18n-ai-translator

Translates UI5 i18n.properties files with OpenAI, preserves placeholders, supports glossary terms, and writes language-specific .properties files.

Features

• Preserves placeholders like {0}, {name}, %s, ${variable}, and %1$s
• Supports missing and overwrite translation modes
• Applies shared and language-specific glossary terms
• Writes UI5-style language files such as i18n_de.properties
• Supports optional Unicode escaping with per-language overrides
• Uses a local cache to avoid repeated translation requests

Installation

npm install --save-dev @concircle/i18n-ai-translator

CLI

Example package.json script:

{
  "scripts": {
    "i18n:translate": "i18n-ai-translator --input ./webapp/i18n/i18n.properties --languages de,fr --mode missing"
  }
}

Show CLI help:

npx i18n-ai-translator --help

Enable Unicode escaping:

{
  "scripts": {
    "i18n:translate": "i18n-ai-translator --input ./webapp/i18n/i18n.properties --languages de,fr --mode missing --encode-unicode"
  }
}

Configuration

Example i18n-ai.config.json:

{
  "sourceLanguage": "en",
  "targetLanguages": ["de", "uk"],
  "translationMode": "missing",
  "encodeUnicode": false,
  "languageOptions": {
    "de": { "encodeUnicode": true },
    "uk": { "encodeUnicode": false }
  },
  "provider": "openai",
  "providerOptions": {
    "apiKey": "sk-...",
    "model": "gpt-4.1-mini",
    "temperature": 0,
    "maxOutputTokens": 4000
  },
  "files": {
    "input": "./webapp/i18n/i18n.properties"
  },
  "cache": {
    "enabled": true,
    "ttlMs": 604800000
  },
  "batchSize": 20,
  "verbose": false
}

Supported top-level config fields:

interface TranslatorConfig {
  sourceLanguage?: string;
  targetLanguages: string[];
  translationMode?: 'missing' | 'overwrite';
  encodeUnicode?: boolean;
  languageOptions?: Record<string, { encodeUnicode?: boolean }>;
  provider?: 'openai';
  providerOptions?: {
    apiKey?: string;
    model?: string;
    baseURL?: string;
    organization?: string;
    temperature?: number;
    maxOutputTokens?: number;
  };
  files?: {
    input?: string;
    outputDir?: string;
    languageFilePattern?: string;
  };
  glossary?: GlossaryConfig | string;
  cache?: {
    enabled?: boolean;
    ttlMs?: number;
    dir?: string;
  };
  rules?: string[];
  batchSize?: number;
  verbose?: boolean;
}

Glossary

Example glossary file:

{
  "shared": [
    {
      "source": "Order",
      "target": "Auftrag",
      "context": "SAP sales document"
    }
  ],
  "languages": {
    "de": [
      {
        "source": "Plant",
        "target": "Werk"
      }
    ],
    "uk": [
      {
        "source": "AI",
        "target": "ШI"
      }
    ]
  }
}

Programmatic Usage

Translate a file directly:

import { Translator } from '@concircle/i18n-ai-translator';

const translator = new Translator({
  sourceLanguage: 'en',
  targetLanguages: ['de'],
  translationMode: 'missing',
  provider: 'openai',
  providerOptions: {
    apiKey: process.env.OPENAI_API_KEY
  },
  files: {
    input: './webapp/i18n/i18n.properties'
  }
});

const result = await translator.translateFile();
console.log(result);

Load config from file:

import { Translator } from '@concircle/i18n-ai-translator';

const translator = await Translator.fromConfig({
  configPath: './i18n-ai.config.json'
});

const result = await translator.translateProject();
console.log(result);

Use the top-level helper:

import { translateProject } from '@concircle/i18n-ai-translator';

const result = await translateProject({
  inputPath: './webapp/i18n/i18n.properties',
  languages: ['de'],
  dryRun: true
});

console.log(result);

Translation Modes

• missing: only translates keys that are missing or empty in the target file
• overwrite: retranslates all keys from the source file

Placeholder Handling

The translator masks placeholders before sending text to the model and validates that they are still present afterwards.

Supported placeholder styles:

• {0}, {1}
• {name}, {email}
• %s, %d
• %1$s, %2$d
• ${variable}

Output Files

By default, UI5-style language files are generated next to the source file:

• i18n.properties
• i18n_de.properties
• i18n_fr.properties

You can override the output directory or file naming pattern in files.outputDir and files.languageFilePattern.

Cache

Caching is enabled by default. Cache keys include:

• provider
• model
• source language
• target language
• source value
• glossary terms
• translation rules

Default cache location:

~/.i18n-ai-translator-cache/

Environment Variables

Supported environment variables:

export OPENAI_API_KEY=sk-...
export OPENAI_MODEL=gpt-4.1-mini

Examples

See examples/ for small runnable examples:

• basic.mjs
• with-glossary.mjs
• advanced.mjs

Development

npm run build
npm run lint
npm test

License

MIT © 2026 Herbert Kaintz - Concircle

Contributing

See CONTRIBUTING.md.

Security

See SECURITY.md.