OpenAI Provider

Translify uses the official openai npm package to translate your i18n keys.

Setup

// translify.config.ts
export default {
  ai_translation: {
    enabled: true,
    provider: 'openai',
    openai_api_key: process.env.OPENAI_API_KEY,
    model: 'gpt-5.6-luna', // recommended for cost-sensitive workloads
    temperature: 0, // deterministic (best for translations)
    batch_size: 50, // keys per API call
    verify: false,
    values_only: false,
  },
};

Set your API key:

export OPENAI_API_KEY=sk-...

Models

Model	Best fit
gpt-5.6-luna	Cost-sensitive, high-volume translation
gpt-5.6-terra	Higher quality while balancing cost
gpt-5.6-sol	Quality-first translation and difficult cases

We recommend gpt-5.6-luna as the direct OpenAI starting point. OpenAI describes it as the GPT-5.6 model for cost-sensitive, high-volume workloads. Model pricing changes; follow the live OpenAI model page and run a representative translation sample before choosing a production default.

Interpolation variables

The provider automatically preserves interpolation variables:

{ "greeting": "Hello, {name}!" }

Translates to French as:

{ "greeting": "Bonjour, {name} !" }

Rate limits and batching

batch_size controls how many keys are sent per API call (default: 50). Reduce this if you hit rate limits on your OpenAI plan.

OpenAI reports token usage, so translify translate prints prompt, completion, and total tokens when the API returns them. Cost reporting is provider-specific and is currently available with OpenRouter.