This is the companion translation REST service for the OmegaT machine translation available at https://github.com/pikatech/omegat-opus-mt-plugin
It provides an Opus MT English-to-German model that can be queried via GET requests. The only parameter that should be set is the 'text' parameter. While it also takes 'source' and 'target' parameters, those are set to "en" and "de", respectively, by default and should be left alone. They are only needed to check if the query posed by OmegaT is valid. Otherwise, the service will reply: "This model translates from English to German."
After installing the dependencies via the usual
pip install -r requirements.txt
and picking your preferred version of PyTorch, you can start the service with
uvicorn opus_mt:app
Alternatively, you can build a Docker image with
docker build .
and run the service in a container via
docker run -p 8000:80 <IMAGE ID>
Please change the PyTorch installation settings in the Dockerfile to your needs. By default, it is set up to use the CPU.
curl -X 'GET' \
'http://localhost:8000/translate?text=I%27m%20an%20example%20text.&source=en&target=de' \
-H 'accept: application/json'
Response:
{
"translation": "Ich bin ein Beispieltext."
}