Skip to main content

Node deployment

In order to run an deploy a serving node in self hosted enviroment please follow these instructions:

Run with docker:​


$ docker run --rm   \
                  -p 8080:80 \
                  -e NODE_NAME=[node_name] \
                  -e LABELATORIO_API_TOKEN={your_api_token} \
                  --platform linux/amd64 \
                  blipsolutions/labelatorio-serving-node:latest

Configuration options​

  • NODE_NAME [required] - Name of the node (used to access configuration)
  • LABELATORIO_API_TOKEN [required] - API token of user that is used to access to labelator.io
  • MAX_CONNECTIONS_COUNT [optional] - uvicorn max connections settings
  • MIN_CONNECTIONS_COUNT [optional] - uvicorn max connections settings
  • ROOT_PATH [optional] - Root path for serving behing reverse proxy, when a subpath is added to URL
  • ALLOWED_HOSTS [optional] - CORS settings
  • SERVICE_ACCESS_TOKEN [optional] - for internal use, token that allows you force refresh configuration
  • LABELATORIO_URL [optional]

Volumes​

Mounting volumes is not nescesary since no data are beeing permanently saved in the container, however mounting app/models_cache folder might be usefull, since you optimize startup time on recreation (models that have been downloaded in the past wont be downloaded again)

Also mounting /queues would preserve data yet to be send to Labelator.io for review (might be usefull if unexpected shutdown occured in the middle of the operation)