For our CI, we use GitHub's built-in GitHub Actions as defined by this config file. All this does it check for front-end build errors. In the future this is where things like unit and integration tests should be added. In an ideal world, master should always have a check (meaning no build errors) and PRs should not be merged until there are passing.
Both the front-end and back-end are run on the same Heroku dyno (think: very small and cheap server) which runs $ node server.js. Heroku settings can be viewed at dashboard.heroku.com/apps/ai-global-portal. Here one can also find production's environment variables, DNS settings, server logs, and error metrics. HTTPS is handled completely by Heroku.
A single CNAME record in namesilo directs portal.responsible.ai to Heroku.
Heroku is current set to automatically deploy anything that's pushed to master. This means it's very important to verify PRs work before merging into master. In the event that a bad PR breaks production, one can easily (and only temporarily) roll back to an old version of the code in the Heroku settings.
Our MongoDB database is hosted on Atlas for free (up to 512MB). Atlas settings can be view at cloud.mongodb.com/v2/5f9f71470cc0b741eab50a50. The "collections" tab can be very useful for editing database objects manually. Note: MONGODB_URL is effectively the password for this, NEVER commit or share it.
File storage is handled by AWS s3. The storage bucket can be accessed and edited at s3.console.aws.amazon.com/s3/buckets/ai-portal-files. Note 1: Everything in the ai-portal-files bucket is public, NEVER store confidential information there. Note 2: The current CORS settings only allow access from the portal (although this is easily updatable).