This project provides automated Docker images of LibreTranslate that include all language models.
- Faster container start-up time: The container with models included starts faster as it avoids downloading approximately 14 GB of data during start-up. This improves responsiveness and prevents liveness probe timeouts, avoiding unnecessary container restarts.
- Reduced bandwidth usage: By including the models in the container, each restart doesn't require re-downloading the language models, resulting in significant bandwidth savings (around 14 GB per restart) in container orchestration environments like Kubernetes or Docker Swarm.
- Offline usability: The self-contained image allows you to use LibreTranslate in private networks without internet connectivity.
- Larger container image size: The resulting container image size is 16 GB (~10GB compressed) due to the inclusion of language models.
check-new-tag.yaml
runs once per hour and checks for new tags in the upstream repo. If upstream has a tag larger than what's available from https://github.com/t-lo/LibreTranslate/pkgs/container/libretranslate a build is triggeredi for that tag. The new build is tagged "latest". This action can be triggered manually.publish-ghcr.yml
builds a LibreTranslate container image from the upstream repo. The image is published to ghcr.io. Also, an email will be sent to notify one or more recipients of the new image. Optionally takes a version to build (tag; defaults to latest main), and a bool whether the image shall be tagged "latest".
Apart from GITHUB_TOKEN
, secrets are needed only for the "new release available" email.
GITHUB_TOKEN
: repo token with write access to the libretranslate ghcr package for image publishingSMTP_SERVER
: SMTP server to use for sending "new image available" email. Leave empty / don't set to disable email report.SMTP_USER
: Username for logging in to SMTP server.SMTP_PASSWORD
: Password for logging in to SMTP server.EMAIL_TO
: Recipient of the Good NewsEMAIL_FROM
: Sender of the email. UsuallySMTP_USER@SMTP_SERVER
.
For more details and usage instructions, please refer to the LibreTranslate repository.