viq / autotune-docker

Docker image for running oref0 autotune

Autotune Docker Image

Docker image for running oref0-autotune

Information

This image is designed to allow you to run oref0-autotune (docs) without needing to have a full copy of oref0 installed to your system. I made this image because oref0 can be quite heavy in terms of filesystem and dependency requirements for someone who wants to use the autotune portion of the system. In addition, the setup process can be quite tricky and time consuming if you're trying to use the autotune feature. By using a docker image for this we can make it simple to run as well as clean up afterwards.

It installs the dependencies, and then copies oref0 that you need to check out into this directory.

Building the image

You can build this image by navigating into the directory that includes the Dockerfile and running

docker build -t autotune .

Recommended setup

This will run autotune on the Nightscout server https://mynightscout.azurewebsites.net starting from June 1, 2018 and ending on June 5, 2018. This should give a good example for most people to start off with.

docker run \
    --rm -it \
    --name=autotune \
    -e "START_DATE=2018-06-01" \
    -e "AUTOTUNE_PREFS=--end-date=2018-06-05" \
    -e "NS_HOST=https://mynightscout.azurewebsites.net" \
    -v $PWD/autotune:/data \
    autotune

Getting recommendations

Assuming you created a volume for /data the recommendations will be output to autotune_recommendations.log in the volume.

Volumes

Providing a volume at /data in the container will automatically link and use these files and directories. The currently supported files are:

• autotune_recommendations.log - the output of autotune (will be overwritten automatically)
• profile.json - file telling autotune what settings your pump uses currently. Instructions for how to make this file can be found here, scrolling down to step 3

docker run \
    --name=autotune \
    -v /host/path:/data \
    autotune

Environment variables

This image works mainly through environment variables. These can be passed through docker run by adding -e followed by the variable declaration. As an example, adding this to the docker run command will set the Nightscout API secret to hunter2: -e "API_SECRET=hunter2" The currently supported variables are:

• NS_HOST - Required: URL of your Nightscout website to run autotune on. Do not leave a leading slash, it should end with the TLD and nothing more
• START_DATE - Required: date of where to run autotune from in YYYY-MM-DD format
• API_SECRET - API secret of your Nightscout website
• AUTOTUNE_PREFS - Extra command line options to pass along to autotune. More information can be found below

Additional autotune preferences

Autotune permits some additional command line preferences that are not handled by this image. You can specify them under the AUTOTUNE_PREFS environment variable

docker run \
    --name=autotune \
    -e "AUTOTUNE_PREFS=--end-date=2018-06-05" \
    autotune

Troubleshooting

First check the troubleshooting page provided by OpenAPS. Please don't bother the OpenAPS contributors about setup issues since this docker image this is not supported or created by them.