argosopentech/argos-translate

Open-source offline translation library written in Python

Argos Translate – Offline machine‑translation library

What it is – A Python package that lets you run neural‑machine‑translation models locally, without needing an internet connection. It wraps the OpenNMT /CTranslate2 inference engine and provides a simple API, a command‑line tool, and (via a separate repo) a desktop GUI.

Key capabilities

  • Plug‑in language models – Translation models are distributed as ".argosmodel" zip files. The argospm-index repository lists dozens of pre‑trained models (Arabic, Chinese, French, Swahili, …) and the package manager (argospm) can download and install them.
  • Pivot translation – If a direct model for a language pair is missing, Argos Translate can automatically route the text through an intermediate language (e.g., es → en → fr) using the models you have installed.
  • GPU support – Setting the environment variable ARGOS_DEVICE_TYPE=cuda (or auto) passes the device choice to the underlying CTranslate2 engine, enabling CUDA acceleration.
  • Multiple front‑ends –
    • Python library (argostranslate.translate.translate)
    • CLI (argos-translate and argospm commands)
    • Desktop GUI (in the separate argos-translate-gui repo)
    • Web/API – the LibreTranslate service is built on top of Argos Translate and exposes a REST endpoint.
  • Extensions – Helper libraries for translating HTML (translate-html) and whole files (argos-translate-files).

Installation

# Core library
pip install argostranslate

# Optional GUI
pip install argostranslategui

You can also clone the repo and install in a virtual environment with pip install -e ..

Quick example (Python)

import argostranslate.package, argostranslate.translate

from_code, to_code = "en", "es"
# fetch latest model list and install the needed pair
argostranslate.package.update_package_index()
pkg = next(p for p in argostranslate.package.get_available_packages()
           if p.from_code == from_code and p.to_code == to_code)
argostranslate.package.install_from_path(pkg.download())

print(argostranslate.translate.translate("Hello World", from_code, to_code))
# → ¡Hola Mundo!

Command‑line usage

argospm update                     # refresh model index
argospm install translate-en_de    # install a specific model
argos-translate --from en --to de "Hello World!"
# → Hallo Welt!

Ecosystem & related projects

  • LibreTranslate – a public web API and demo site that runs Argos Translate under the hood.
  • argos‑train – scripts for training your own models compatible with Argos Translate.
  • MetalTranslate – a C++ wrapper around the same translation engine.
  • Language‑model bindings for many languages (Python, Rust, Go, Java) are listed in the README.

License – Dual‑licensed under MIT or CC0, allowing both permissive commercial use and public‑domain dedication.

Who might use it – Developers needing on‑device translation (privacy‑sensitive apps, edge devices, offline tools), researchers experimenting with OpenNMT models, or anyone who wants a free, self‑hosted alternative to cloud translation services.

Related

  • Project
    openai/gpt-oss

    OpenAI’s gpt‑oss repo delivers reference code and tooling for the open‑weight 120 B and 20 B LLMs, including a special Harmony chat format, MXFP4 quantization, and back‑ends for PyTorch, Triton, Metal, vLLM, Ollama, and more. Licensed Apache‑2.0, it’s aimed at researchers and developers who want to run, fine‑tune, or build tool‑augmented agents with these models.

  • Project
    knights-analytics/hugot

    A Go library for running and fine-tuning ONNX transformer pipelines, enabling Hugging Face compatible inference and training natively in Golang.

  • Project
    istupakov/onnx-asr

    A lightweight Python package for Automatic Speech Recognition using ONNX models, eliminating the need for PyTorch or Transformers for fast deployment on edge and server hardware.

  • Project
    alexzhang13/rlm

    An inference engine and training environment for Recursive Language Models (RLMs) that enable LLMs to handle near-infinite context by programmatically decomposing tasks via recursive sub-calls in a REPL environment.

  • Project
    onnx/onnx

    ONNX is an open, framework‑agnostic model format and reference implementation that lets you move AI models between tools and runtimes. Install via `pip install onnx`, use the Python API to build or load a graph, then run inference with any ONNX‑compatible engine.