Dot Lexicon

🌐 A microservice for quick and local translation using A.I.

This service starts a local webserver used for neural machine translation.

🚀 Features

For inference, all models are ran on the CPU. Every model utilized in this service are 8-bit quantized, which results in decreased latency and storage costs.

📦️ Setup

Note: after setup.py is executed, 5GB will be used.

Docker will take ~16GB in total.

# Clone the repository:
git clone https://github.com/dothq/lexicon.git
cd lexicon/

# Build Dockerfile
sudo docker build -t translate .

Run the container with sudo docker run -d -p 3000:3000 translate

The default authorization key is universal

Continue without docker:

Install required dependencies:

# Install requirements
pip3 install -r requirements.txt
cd lexicon/

# Download all the available models:
python3 setup.py

# Launch the server
python3 application.py

Your API is then available for use via POST to /translate (port 3000) using JSON:
Header: Authorization: Bearer TOKEN (default key universal)

{
  "from": "en",
  "to": "es",
  "input": "This is a test translation using Dot Lexicon!"
}

Response will be in JSON.

🔧 Contributing

We accept all positive contributions that affects this repository and service as a whole; we accept trained .argosmodels files via pull request.

Our models: (These will automatically be installed on setup via latest.json)

❤️ Acknowledgements

Argos Translate, which is built on OpenNMT, is widely used in this repository for translation.

📜 Licenses

Dot Lexicon is licensed under the MIT license.