This repository provides Docker images for psiTurk, via DockerHub.
According to the documentation, psiTurk is:
designed to help you run fully-customized and dynamic web-experiments on [Amazon Mechanical Turk]. Specifically, it allows you to:
- Run a web server for your experiment
- Test your experiment
- Interact with [Amazon Mechanical Turk] to recruit, post [Human Intelligence Task]s, filter, and pay participants ([Amazon Mechanical Turk] workers)
- Manage databases and export data
psiTurk also includes a powerful interactive command interface that lets you manage most of your [Amazon Mechanical Turk] activity.
The images are tagged using psiTurk version numbers, as well as the tag
latest for the most recent version of psiTurk available
on PyPI and dev for the most recent version of psiTurk
available on the master branch of
the psiTurk repository. The current tags are:
devlatest2.2.32.2.22.2.12.2.02.1.72.1.62.1.4It's likely that you'll want to use either the latest or dev
tag. The dev tag is often quite a bit further along than the latest
release on PyPI. In particular, at the time of writing
(2018-07-28), if you want to use a recent version of MySQL for your
database, you'll need to use the dev tag, since the latest tag for
the version of psiTurk that is on PyPI uses a very outdated version of
SQLAlchemy, which only supports MySQL versions 5.7.x and older. If you
want to use an 8.x.y version of MySQL, you'll need to use the dev tag.
Also note that there is not a tag for 2.1.5 because there is not a
version for psiTurk 2.1.5 on PyPI. All other older versions of
the 2.x.y series are not available because they try to install
gnureadline,
which
requires installing libncurses5-dev via apt-get,
and I'd rather not increase the image size just for some older psiTurk
versions that folks are unlikely to want to ever use. None of the
1.x.y versions are available because none of these support
the Ad Server.
It's worth noting that us_only and approve_requirement did not work
prior to version 2.2.0 (see, e.g., here). Use
older versions at your own risk.
These Docker images are intended to be the base for your own (very simple!) Docker images. (For help using psiTurk and Docker together, as well as integrating a database, check out this series of blog posts that I wrote on this very topic.)
To get started, first install Docker.
Next, create a directory for your experiment:
mkdir stroop-task
cd stroop-task
Then, create a Dockerfile (replace latest with whatever version of
psiTurk that you want to use; see the available tags above):
echo "FROM adamliter/psiturk:latest" > Dockerfile
You have two options about how to set up your Docker image. You can not include the experiment files in the image and instead keep the experiment files on the host machine, mounting the directory containing the experiment files to the Docker container.
Alternatively, you can include the experiment files in the Docker image itself.
With the first option, you will be able to edit the files on your local host machine, and these changes will be reflected in the experiment that is being served by the Docker container without having to rebuild the image. I'd suggest this option for developing and tweaking an expeirment.
With the second option, the files are copied into the Docker image, and so the Docker image has to be rebuilt each time you make a change to the experiment files. I'd suggest this option for deployment of the final version of your experiment and/or distributing it to other researchers.
In either case, let's use the example experiment generated by running
the command psiturk-setup-example. So, run the command:
psiturk-setup-example
This will create a directory called psiturk-example with the example
experiment files. Unless you already have a ~/.psiturkconfig file, it
most likely also created this file, too.
The Docker image expects the .psiturkconfig file to be inside the
same directory as the experiment files, so run the following command:
mv ~/.psiturkconfig ./psiturk-example/
Now, to set up your Dockerfile for the first option, run the following command:
echo 'VOLUME ["/psiturk"]' >> Dockerfile
Then, build your Docker image:
docker build -t <YOUR_USERNAME>/stroop-task .
And, finally, run a container from the built image:
docker run -it --rm --name stroop-task -p 22362:22362 -v `pwd`/psiturk-example:/psiturk <YOUR_USERNAME>/stroop-task
This will pop you into a bash session where you can start the psiTurk server:
psiturk
server on
To check it out, navigate to http://localhost:22362 in your browser.
To detach from the Docker container without killing it, hit CTRL+p then CTRL+q.
Now you can edit your experiment files, and these changes should be reflected in your browser.
To set up your Dockerfile for the second option, run the following command:
echo "COPY ./psiturk-example /psiturk" >> Dockerfile
Then build your Docker image:
docker build -t <YOUR_USERNAME>/stroop-task .
And, finally, run a container from the built image:
docker run -it --rm --name stroop-task -p 22362:22362 <YOUR_USERNAME>/stroop-task
This will pop you into a bash session where you can start the psiTurk server:
psiturk
server on
To check it out, navigate to http://localhost:22362 in your browser.
If you want to make any changes, you will have to rebuild the image because the experiment files are inclued in the image.
If you're having trouble getting things to work, try changing the
default host setting in config.txt from localhost to 0.0.0.0.
For some reason, the localhost option has never worked for me, but
0.0.0.0 does.
Content type
Image
Digest
Size
393.7 MB
Last updated
over 5 years ago
docker pull cpllab/psiturk