Quick Start

Last updated 10/20/2021

Deepgram provides a variety of tools that you can use to configure and deploy our ASR services on-premise. This Quick Start shows only one strategy. If you plan to deploy Deepgram products on-premise using this strategy, follow this Quick Start. Otherwise, deploy Deepgram products on-premise by following the steps in Configure Hardware & Software, Get and Configure Deepgram Products, and Product Deployment.

Supported Deployment Strategy

This Quick Start supports the following:

  • AWS EC2, instance type p2.xlarge with a Tesla K80 GPU and Ubuntu 18.04


To deploy using the supported deployment strategy, read on.


  1. Create a cloud-init.txt file with the following content:

    echo $DOCKER_PASSWORD | docker login --username $DOCKER_USERNAME --password-stdin
    docker pull deepgram/onprem-api:latest
    docker pull deepgram/onprem-engine:latest
    docker logout
    cd /home/ubuntu/config
    sed -i 's/REPLACEWITHKEY/'"$LICENSE_KEY"'/g' api.toml
    sed -i 's/REPLACEWITHKEY/'"$LICENSE_KEY"'/g' engine.toml
    cd /home/ubuntu/
    mkdir engine
    cd /home/ubuntu/engine
    mkdir models
    cd models
    mv general*.dg general.dg
    chmod 644 general.dg
    chown -R ubuntu:ubuntu /home/ubuntu/engine
    cd /home/ubuntu/
  2. Edit cloud-init.txt to enter the appropriate values at the top of the file:


    For GENERAL_MODEL, include the URL of the model provided to you by Deepgram. If you require more than one model or want extra features, like punctuation, contact your sales representative, and we will provide an cloud-init.txt customized to your needs.

  3. Provide the contents of cloud-init.txt as user data.

  4. Start the AWS instance, ssh to it, and start the Deepgram api and engine services by running:

    $ cd /home/ubuntu/config
    $ docker-compose up -d
This uses `docker-compose` as a simple orchestration strategy; you can skip this last step if you plan to use a different orchestration strategy.