stas-server

stas-server is an alternative standalone translation server for Sugoi Translator.

stas-server name is an abbreviation of Sugoi Translator Alternative Standalone Server.

Not affiliated with Sugoi Translator.

The project is in active development.

Todo List to v1.0.0 Release

• Async server (bottle.py with tornado)
• Multiline support
• Newlines support
• Batch support
• Only translate Japanese
• Access-Control-Allow-Private-Network header
• XUAT Placeholder support
• CUDA support (cannot test because I don't have Nvidia GPU)
• Detailed documentation
• Cache support when running (Reset on Server Shutdown)
• New models directory support

User Guide

Requirements

• Sugoi Japanese Toolkit from MingShiba. Need to apply CudaInstallForToolKit from Sugoi Translator Discord if using version lower than V7 and below. Starting from V8, ct2Model available by default. Make sure to take note of the path to models folder.

For V8, the usual path is [EXTRACTEDFOLDER]/Code/backendServer/Program-Backend/Sugoi-Japanese-Translator/offlineTranslation/models.

For V7 and below using CudaInstallForToolKit (you need to run this first), the usual path is [EXTRACTEDFOLDER]/Code/backendServer/Program-Backend/Sugoi-Japanese-Translator/offlineTranslation/ct2. Inside ct2 folder, ct2_models need to be renamed to ct2Model.

• Make sure to use Python 3.9 and above. Python 3.12 is recommended.

• Node is also required. Make sure to use the latest major LTS version or higher. Only used during installation.

Installation

This project is not intended as a library, so installation through pipx or uv is recommended.

In the future, the package will be published to pypi for easier installation. No guarantee through.

Download the wheel in the Releases and install the package.

Using pipx

pipx install ./stas_server-x.x.x-py3-none-any.whl

Using uv

uv tool install ./stas_server-x.x.x-py3-none-any.whl

Upgrade Server

Using pipx

Because this package is not installed from pypi, to upgrade this package, your need to force pipx to install the new version.

pipx install --force ./stas_server-x.x.x-py3-none-any.whl

Using uv

Same as installation.

uv tool install ./stas_server-x.x.x-py3-none-any.whl

Running Server

Usage: stas-server [OPTIONS] [PORT]

  Run stas-server, an alternative standalone translation server for Sugoi
  Translator.

Options:
  -v, --version           Show the version and exit.
  --cuda                  Enable CUDA.
  --no-cache              Disable cache.
  --models_dir DIRECTORY  Path to models folder.  [required]
  -h, --help              Show this message and exit.

Options

Name	Default	Description
port	14366	Port of the server.
cuda	false	Enable CUDA support.
models_dir	./models	Path to models directory. By default, the program will use the folder named models in the current working directory.

Development

This project is developed using the latest Python and managed by Rye.

To start developing for this project, make sure to install Rye. It will automatically download Rye-managed Python. It will not clash with your system Python because it is only used for this project.

Refer to Rye official docs for the installation instructions.

Node also needed during installation because this project is using, PythonMonkey. The latest Node LTS is used during development.

Initialize venv and install dependencies.

rye sync

Launch and test the server.

rye run stas-server

Build the project wheel.

rye build -c --wheel

License

This project is licensed under the MIT license.

Acknowledgement

• Thanks to MingShiba for creating the Sugoi Japanese Toolkit and making high-quality (still machine translation) available to enjoy many untranslated Japanese works.
• Thanks to Tenerezza and bimbmsm on Sugoi Toolkit Discord for CudaInstallForToolKit script for adding CT2 support. Some of the code is based on the included CT2 multiline server script.
• Thanks to Vin-meido for Sugoi Translator XUAT support. Some of the code is based on the included server script.