gpt-translate-json v0.2.0

GPT Translate JSON

Command to automatically translate your app JSON files using OpenAI GPT Chat Completions API

npm install gpt-translate-json --save-dev

Requirements

• Node.js >= 16
• OpenAI Api key

Usage

Add the command in package.json, and provide required parameters:

"scripts": {
  "gpt-translate-json": "gpt-translate-json --apiKey=openai_api_key --model=gpt-3.5-turbo --maxTokens=3000 --langs=en-US,it-IT --originalLang=en-US"
}

Available options:

• apiKey OpenAI API key. Required
• model OpenAI Chat Completion model. Required
• maxTokens OpenAI model max tokens per request. Required
• langs All supported languages. Required
• originalLang Original language. Required
• basePath The base path. Default to './'
• assetsPath Path to translation files: [basePath]/[assetsPath]/[lang]/*.json. Default to 'i18n'
• rules Prompt rules. Defaults:
  • 'do not translate proper names'
  • 'do not translate texts enclosed in double braces {{}}'
  • 'do not translate the html tags''
  • 'do not translate URLs'

Depending on the model used, requests can use up to maxTokens shared between prompt and completion. Keep the number of maxTokens lower than the maximum allowed by the model: in fact, the command splits the files into multiple requests to respect the maximum number of tokens in each request based on the English language, but depending on the target language, the number of tokens used can vary significantly

Having a file structure like this:

i18n/
│   
└───en-US/
        app.json

i18n/en-US/app.json

{
  "app": {
    "title": "<h1>Library to translate JSON using GPT</h1>"
  }
}

the command:

npm run gpt-translate-json

will generate:

i18n/
│   
└───en-US/
│       app.json
└───it-IT/
│       app.json
└───.metadata/
        translated.json
        translated-langs.json

i18n/it-IT/app.json

{
  "app": {
    "title": "<h1>Libreria per tradurre JSON usando GPT</h1>>"
  }
}

Add translations

The file .metadata/translated.json contains the paths of translated values, so if you add new translations:

i18n/en-US/app.json

{
  "app": {
    "title": "<h1>Library to translate JSON using GPT</h1>"
  },
  "about": "About us"
}

the command will request only the new translations, reducing tokens usage, and the files will be updated:

i18n/it-IT/app.json

{
  "app": {
    "title": "<h1>Libreria per tradurre JSON usando GPT</h1>>"
  },
  "about": "Chi siamo"
}

Add languages

The file .metadata/translated-langs.json contains the langs already translated, so if you add a new lang:

"scripts": {
  "gpt-translate-json": "gpt-translate-json --apiKey=openai_api_key --model=gpt-3.5-turbo --maxTokens=3000 --langs=en-US,it-IT,es-ES --originalLang=en-US"
}

will generate:

i18n/
│   
└───en-US/
│       app.json
└───it-IT/
│       app.json
└───es-ES/
│       app.json
└───.metadata/
        translated.json
        translated-langs.json

i18n/es-ES/app.json

{
  "app": {
    "title": "<h1>Biblioteca para traducir JSON usando GPT</h1>"
  },
  "about": "Sobre nosotros"
}

Using it programmatically

Rather than using the command, you can invoke gptTranslateJson function:

import { gptTranslateJson } from 'gpt-translate-json';

await gptTranslateJson({
  apiKey: 'openai_api_key',
  model: 'gpt-3.5-turbo',
  maxTokens: 3000,
  langs: ['en-US', 'it-IT'],
  originalLang: 'en-US',
  rules: [
    // your custom rules
  ]
});

Development Builds

Build

npm install
npm run build

npm test

License

MIT