OpenRouter Provider

Translify uses the official @openrouter/sdk package to translate your i18n keys through OpenRouter.

Setup

// translify.config.ts
export default {
  ai_translation: {
    enabled: true,
    provider: 'openrouter',
    openrouter_api_key: process.env.OPENROUTER_API_KEY,
    model: 'deepseek/deepseek-v4-flash',
    temperature: 0,
    batch_size: 50,
    verify: false,
    values_only: false,
  },
};

Set your API key:

export OPENROUTER_API_KEY=sk-or-...

Models

For translation, we recommend starting with deepseek/deepseek-v4-flash: it is optimized for efficient, high-throughput workloads and is currently one of the lowest-cost capable options in the OpenRouter catalogue. Prices and routed providers can change, so check the live model page and validate quality against your own catalogue.

Use any model slug from the OpenRouter catalogue:

model: 'deepseek/deepseek-v4-flash'; // recommended
model: 'openai/gpt-5.6-luna';
model: 'anthropic/claude-sonnet-4';
model: 'google/gemini-2.5-flash';

Usage And Cost

OpenRouter reports token usage and cost when available. After translify translate, Translify prints prompt tokens, completion tokens, total tokens, and total USD cost.

Verification

Enable a second LLM pass when you want the translated batch checked and corrected before writing:

ai_translation: {
  provider: 'openrouter',
  verify: true,
  verify_model: 'deepseek/deepseek-v4-flash',
}

If verify_model is omitted, Translify uses the same model for translation and verification.

Values Only Mode

By default, Translify sends key-value pairs to the provider. If you want to send only source text values, enable:

ai_translation: {
  values_only: true,
}

The provider must return the same number of translated values in the same order; Translify then maps them back to the original keys.