Screenshare Redux

Another LIL screenshare project. This one is based on Django Channels.

This project serves two functions: first, to allow a team to post pictures to a Slack channel and have them appear on a monitor in a common workspace; second, as a context for experimentation with web technologies.

Deployment

This system should be deployed on a server accessible from the public web, so that Slack can reach it. It subscribes to certain Slack events by exposing the endpoint /slack_event; create an app, enter the URL in Event Subscriptions (something like https://api.slack.com/apps/<app ID>/event-subscriptions?), and subscribe to the bot events message.channels, reaction_added, and reaction_removed. In config/.env, you'll need to configure SLACK_SIGNING_SECRET (Signing Secret in the App Credentials section of Settings/Basic Information) and SLACK_BOT_ACCESS_TOKEN ("Bot User OAuth Token" in the OAuth & Permissions section of the app's config); see config/.env.example.

In production, this system is running on a Dokku instance. The complete (?) sequence for setting it up is

dokku apps:create screenshare
dokku redis:create screenshare-store
dokku redis:link screenshare-store screenshare
dokku domains:set screenshare <your domain>
dokku letsencrypt:enable screenshare
dokku config:set --no-restart screenshare "SECRET_KEY=..."
dokku config:set --no-restart screenshare "SLACK_SIGNING_SECRET=..."
dokku config:set --no-restart screenshare "SLACK_BOT_ACCESS_TOKEN=..."
dokku config:set --no-restart screenshare "ALLOWED_HOSTS=<your domain>"
dokku config:set --no-restart screenshare "POST_CHANNEL=..."
dokku config:set --no-restart screenshare "ASCII_FIRE_URL=..."
sudo dokku plugin:install https://github.com/bensteinberg/dokku-primitive
dokku primitive:create prim
dokku primitive:link prim screenshare

Finally, having set up your git remote with something like git remote add dokku dokku@<your dokku instance>:screenshare, deploy with git push dokku develop.

(Screenshare used to be set up on a VM running Debian, with the application served by daphne via systemd, and exposed with nginx. Important dependencies include redis-server.)

Depending on your setup, you may want to use a firewall and/or nginx to restrict access to the / endpoint.

Prior to a production deployment, you may want to try this out locally, using ngrok to expose the service to Slack. The complete sequence for doing this is as follows:

• copy config/.env.example to config/.env
• install redis, probably with brew install redis
• in one terminal, run redis-server
• install ngrok, probably with brew install ngrok/ngrok/ngrok
• in another terminal, run ngrok http 8000
• put the ngrok endpoint in ALLOWED_HOSTS in config/.env
• set up a Slack app as described above, and put SLACK_SIGNING_SECRET and SLACK_BOT_ACCESS_TOKEN in config/.env
• add the app to whatever channel you like, and set the channel name in config/.env like this: POST_CHANNEL='#bottest'
• additionally set ASCII_FIRE_URL in config/.env
• install Poetry, probably with curl -sSL https://install.python-poetry.org | python3 -
• in yet another terminal, run the primitive API; probably the simplest way is docker run -p 8081:8000 registry.lil.tools/harvardlil/primitive:0.03, but you can also clone https://github.com/bensteinberg/primitive-api and run docker build -t primitive-api . && docker run -it -p 8081:8000 primitive-api. Set PRIMITIVE_URL=http://localhost:8081/ in config/.env.
• in yet another terminal, in this directory, run poetry install
• in the same terminal, generate a secret key with poetry run python -c "from django.core.management.utils import get_random_secret_key; print(get_random_secret_key())" and set the SECRET_KEY in config/.env
• in the same terminal, in this directory, run poetry run ./manage.py collectstatic
• in the same terminal, in this directory, run poetry run daphne config.asgi:application --port 8000 --bind 0.0.0.0 -v2
• in the Slack app config page, go to Event Subscriptions and set and verify the Request URL, which will be the ngrok endpoint with /slack_event appended

You should now be able to open http://127.0.0.1:8000/ (or the ngrok endpoint), post an image to the channel you added the app to, and see it appear in your browser.

(Next steps might be to script this, and/or embody it in a docker compose setup.)

Development

For development, use Poetry, but make sure to export the conventional requirements file if you make changes to poetry.lock:

poetry export -o requirements.txt

Note that daphne, when run as shown in the ngrok example above, does not auto-reload on code changes. This issue suggests switching to uvicorn for an ASGI server.