Deploy on premises

In case you need larger capacity than supplied by business plan, we also supply on-premises deployment option that has no daily query limit.

In order to use our NER model locally you should:

  • register as user (if not yet registered)
  • upgrade to enterprise plan in User area -> Upgrade
  • execute a payment with invoice number as a purpose of payment
  • once a license key is available at User area -> Credentials page you can start installing
  • deployment status is available at User area -> Deployments

It is recommended to use Docker images for on-premises deployment. Deployment works fine with either Docker, Docker Compose, Docker Swarm, or Microk8s. Example setup package includes Dockerfile, and bash start script using microk8s.


Hardware requirements for on-premises deployment:
Space on disk required by built Docker image: < 660MB
Space on disk required by NER model:< 30MB
Peak memory usage by Docker container:< 450MB

Example package contents:

You can download example setup package here: download (~2KB). Contents of ZIP file:

DockerfileBuilds docker image for NER model service.
Microk8s deployment configurationSets up single container, single pod deployment running locally on 8005 port.
Microk8s service configurationSets up http service available on private cluster IP on 8005 port.
Bash start script Builds and deploys docker image using Microk8s configuration on local machine.
This file requires modifications, such as setting License Key or Internal API Key.

Upstream communication:

API server deployed on premises does very little communication with upstream server. Currently following requests are being sent:

  • Instance up status (on server start)
  • Instance ping status (once an hour)
  • Latest model download request (once an hour)
  • Instance down status (on server shutdown)
Each request contains as little information as possible. Currently only license key and instance ID are being sent to server. License key should be set by system administrator when installing a server, and instance ID (random UUID) is generated on start of server, which can be retrieved by calling http://your-server-ip:8005/ for locally running server. No processed data queries, usage statistics, or other data is collected or sent to upstream server by on-premises deployments.