Basic Installation

The easiest way to get started with Articulate is to use our included docker-compose file. Articulate relies on several services besides it’s own UI and API. These all get started when you use the provided compose.

  Warning / Be aware / Notice / Look out / etc…
Memory Requirement Loading and Training these language models dictates a certain amount of RAM. For experimenting with smaller agents 3Gb may be sufficient, but for larger agents much more RAM may be required. If training takes more than a few minutes for small agents or more than half an hour for large agents, it is a good indicator that you are running low on memory.
Windows Shared Drives On Windows, make sure you have shared the drive with Docker so that it can mount directories inside of the container.
Docker Toolbox The below instructions work best on Linux and when using Docker for Windows or Docker for Mac. If you are using Docker toolbox follow the remote installation instructions to specify the docker machine IP via the API_URL environment variable.


  1. Installing Prerequisites - Make sure you have Docker, Docker Compose, and Git installed:

    • Docker (If faced with any choices, get the Community Edition (“CE”))
    • Docker Compose (should be auto-installed as part of Docker on mac and windows)
    • Git (installed by default on mac and most linux varieties. To install on windows get it here)
  2. Downloading - Download the current release from this page:

  3. Extracting - Unzip and get a terminal or command prompt in the new directory

Once you’ve got everything extracted you’re ready to start Articulate. The below instructions cover running locally and in the cloud:

Running Locally

This is the simplest setup and for most people wont required anything more than:

docker-compose up

A few gotchas:

After Docker downloads the images and starts all of them Articulate should be available at http://localhost:3000 and the API documentation should be available at http://localhost:7500/documentation

Running Remotely

The only extra requirement when running this way is an environment variable which informs the UI where to look for the API (the public url).

API_URL=http://xxx.xx.xx.xx:7500 docker-compose up