Configuration

lin is designed to work with minimal configuration by auto-detecting your setup. However, you can customize every aspect of its behavior.

Note

All properties in the config file can also be used as CLI flags. For example, with: 'def' in your config is equivalent to running a command with --with def.

Config File

lin uses unconfig to find and load your configuration files. You can use one of the following:

• lin.config.ts (or .js, .mjs, etc.)
• .linrc (without extension, or .json)
• lin property in package.json

i18n Config

If your framework is not auto-detected, you can put your i18n config inside your lin config, or create a separate i18n.config.ts file.

lin.config.ts

Add an i18n object to your main lin.config.ts file.

import { defineConfig } from '@rttnd/lin'

export default defineConfig({
  i18n: {
    locales: ['en-US', 'es-ES'],
    defaultLocale: 'en-US',
    directory: 'locales',
  },
  // ... other lin config
})

i18n.config.ts

If you don’t plan to use other lin config, just create a i18n.config.ts file.

import { defineI18nConfig } from '@rttnd/lin'

export default defineI18nConfig({
  locales: ['en-US', 'es-ES'],
  defaultLocale: 'en-US',
  directory: 'locales',
})

LLM Config

You need to specify the model and the provider in your configuration or via the --model (-m) and --provider (-p) CLI flags.

Make sure the corresponding API key is set in your env variables (e.g., OPENAI_API_KEY).

Example lin.config.ts with LLM options:

import { defineConfig } from '@rttnd/lin'

export default defineConfig({
  options: {
    provider: 'openai',
    model: 'gpt-4.1-mini',
    temperature: 0,
  }
})

Use lin models to see all available models.

Available providers: openai, anthropic, google, xai, mistral, groq, cerebras, azure.

Note that lin can work with models not listed here, just pass the model name as a string.

To add a new provider, consider opening a PR. It’s easy to do!

All properties under options are passed to the Vercel AI SDK.

Presets

To save LLM options, you can define and name different model configurations in your lin.config.ts file.

lin.config.ts

export default defineConfig({
  options: {
    provider: 'openai',
    model: 'gpt-4.1-mini',
    temperature: 0.2,
  },
  presets: {
    'creative-claude': {
      provider: 'anthropic',
      model: 'claude-sonnet-4-0',
      temperature: 0.6,
      context: 'Use a playful tone.'
    },
    'fast-deepseek': {
      provider: 'groq',
      model: 'deepseek-r1-distill-llama-70b',
    },
  }
})

You can then activate a preset using the --model flag. Any other CLI flags will override the preset’s values.

# Use the 'creative-claude' preset
lin sync -m creative-claude

# Use the 'fast-deepseek' preset, but override the temperature
lin add ui.new.feature A new feature -m fast-deepseek -t 0

Context

This simple string is directly added to the system prompt. Use it to provide extra information about your project.

lin.config.ts

export default defineConfig({
  context: 'Translate everything in a formal and polite tone.',
})

Batching

lin provides several options to control how translation requests are batched, which is crucial for handling large projects and managing LLM context limits.

Locale Batching (batchSize)

The batchSize option controls how many locales are processed together in a single translate or sync operation. This is useful when you have a large number of languages. The default is 10.

You can set this in your lin.config.ts or use the --batchSize (or -b) flag.

Key-based Batching (keyBatchSize & charLimit)

Within each locale, lin further batches the keys sent to the LLM. A new batch is created when either of these limits is reached:

• keyBatchSize: The maximum number of keys in a batch. (Default: 50)
• charLimit: The maximum number of characters of all values in a batch. This acts as a proxy for token limits. (Default: 4000)

These can be configured in lin.config.ts or with the --kbs and --cl flags.

with (Context Profiles)

The with option allows you to control which locale files are included in the LLM’s context window. This can significantly improve translation quality by providing the model with more context about your project’s wording and style.

• none (default): Only the keys to be translated are sent to the LLM.
• def: Includes the entire default locale JSON file.
• tgt: Includes the full JSON of each locale currently being translated.
• both: Includes both the default locale file and the target locale files.
• all: Includes every locale JSON file in the context. This may be expensive.
• <locale>: You can also provide one or more specific locale codes (e.g., es-ES, fr).

Config

lin.config.ts

export default defineConfig({
  with: 'tgt',
})

CLI

# Override config and use 'both' profile
lin sync -w both

# Provide specific locales for context
lin add ui.new.key New Key -w es-ES -w fr

# Force no additional context
lin sync -w none

Parser Config

You can configure the key parser to look for keys in different files or use custom lexers.

This is used by the check and translate commands.

lin.config.ts

export default defineConfig({
  parser: {
    input: ['src/**/*.{js,jsx,ts,tsx,vue,svelte,astro}'],
    // ... other i18next-parser options
  }
})