A CFAR-style prediction market for Slack.
This code needs to run on a server that Slack can forward commands to, and it needs to have a Postgres database available. Within Wave this is deployed on Heroku. This is convenient, because Heroku manages both the web server and the database server for you, but it's also more expensive. A cheaper option would be to either install this on a server you're already running, or on a small VPS, but then you need to do more of the ops work yourself.
See runtime.txt for the python version this depends on, and requirements.txt for the dependencies.
To make /predict
work in Slack, you need to tell Slack to forward commands to
your server:
- Visit https://api.slack.com/apps?new_app=1
- Enter "Prediction Market" for the name
- On the next screen, under "Basic Information" / "Building Apps for Slack" / "Add features and functionality" choose "Slash Commands"
- Click "Create New Command"
- Enter:
- Command: "/predict"
- Request URL: the url your server is listening on
- Short Description: "a prediction market"
- Usage Hint: "/predict help"
- Click "Save"
- Click "Basic Information" on the left nav under "Settings"
- Scroll down to "Verification Token" under "App Credentials"
- Copy that token, and set it as the environment variable
SLACK_TOKEN
on your server. This is how your server knows to only accept requests that are actually from your Slack - Click "Install App" on the left nav under "Settings"
- Click "Install App to Team"
- Click "Authorize"
- Try out
/predict help
in a channel.
To set up a new local clone for development, do this once:
git clone git@github.com:waveremit/predictions
cd predictions
source virtualenv.sh
Then, workon predictions
will take you into the proper environment. If there
have been changes to the requirements file, run pip install -r requirements .txt
.
Within Wave, PRs are auto-deployed to Heroku after merging to master if they pass CI.
If you modified the models, there isn't any automated migration system. You have to run the sql commands manually to make the current DB match your model.
pytest .
In one terminal:
createdb predictionslocal
SLACK_TOKEN=1 python app.py
In another:
curl -d 'token=1&user_name=test&text=COMMAND' localhost:5000