Boilerplate code for generating the foundation of your own online psychology experiment. v1.0, requires Python 3
Contains setup scripts which a) build the boilerplate of a PsiTurk jsPsych online experiment and b) upload your code to Heroku easily and link it to said PsiTurk experiment. The experiment is setup for jsPsych + PsiTurk MTurk experiments, but the structure PsiTurk uses can easily be repurposed if you don't need MTurk. I'm working on a more concrete non-MTurk pipeline, but that will be helped much more with an actual example of a non-MTurk experiment.
Requires Python 3. You may also need to replace python and pip in this tutorial with python3 and pip3 depending on your system.
- Clone this repo somewhere on your computer. It doesn't matter where, as the generation script builds the boilerplate experiment folder externally.
git clone https://github.com/LevyDecisionNeuroLab/levylab-online-experiments.git
- Install the requirements for the generation script. (Only configparser at the moment).
cd /path/to/levylab-online-experiment
pip install -r requirements.txt
- Run the generation walkthrough script!
python generation.py
You can now navigate with cd into wherever you generated the experiment boilerplate and begin development! Run the command psiturk to start the PsiTurk CLI (which will manage our server stuff) and then type server on. This will run the local dev server on http://0.0.0.0:22362 through which you can test your experiment. To do so, you need to generate a debug URL as if you are an MTurk worker––this is done by running the command debug in the psiturk CLI.
After you're mostly happy with your experiment and want to run it publicly, you're ready to begin with the next step and upload to Heroku.
- Navigate into your experiment directory (where
config.txtis, among the other files)
cd /path/to/my-experiment
- Install the setup script requirements
pip install -r internals/setup-requirements.txt
- Run the setup walkthrough script! You'll need to use your Heroku account's API token. You can either create your own account and join the Levy Lab team or you can ask Ruby for the credentials for the Levy Lab account.
python internals/setup.py
Once you finish with the walkthrough script, you'll be given a URL with which you can access your experiment much like you did locally. This time, however, the codebase it is pulling from is a git remote heroku/master you can push to to update production. You can now create publicly accessible MTurk HITs in the local PsiTurk CLI. Just run psiturk and then hit create (note you do not need to run the local server, HITs are stored on the Amazon MTurk account itself which is common between the local and production/Heroku servers). Refer to the PsiTurk docs for more info about what you can do and the structure of a PsiTurk experiment.
Email [email protected] if you have any questions or suggestions at all!