Slang.Gpt.Cli

Slang.Net - GPT

Use GPT to automatically translate your app at compile time.

Currently, only the OpenAI API is supported.

Getting Started

Install slang-gpt-cli:

dotnet tool install -g slang.gpt.cli --prerelease

Then add the following configuration in your slang.json:

{
  "base_culture": "ru",
  "gpt": {
    "model": "gpt-4o-mini",
    "description": "Showcase for Slang.Net.Gpt"
  }
}

Let's run this:

slang-gpt <csproj> --target=ru --api-key=<api-key>

Configuration

Command line arguments

Models

1k tokens = 750 words (English)

GPT context length

Each model has a different context length. Try to avoid exceeding it as the model starts to "forget".

Luckily, slang_gpt supports splitting the input into multiple requests.

The max_input_length is optional and defaults to some heuristic.

If you work with less common languages and the model starts to forget, try to reduce the max_input_length.

Alternatively, you can also use a model with a larger context length.

Partial translation

By default, slang_gpt will only translate missing keys to reduce costs.

You may add the --full flag to translate all keys.

slang-gpt <csproj> --target=ru --api-key=<api-key> --full

To avoid a specific subset of keys from being translated, you may add the ignoreGpt modifier to the key:

{
  "key1": "This will be translated",
  "key2(ignoreGpt)": {
    "key3": "This will be ignored"
  }
}

Target language

By default, slang_gpt will translate to all existing locales.

You may add the --target flag to translate to a specific locale. This may be useful if you want to translate to a new locale.

Additionally, you may also use predefined language sets (keep in mind that English must be the base locale):

By GDP (Gross Domestic Product):

By region and population: