Skip to content

Latest commit

 

History

History
126 lines (91 loc) · 5.81 KB

install.md

File metadata and controls

126 lines (91 loc) · 5.81 KB

Installation

Русский · Back to the README

Your own server from the app

The easiest way is the Mac or Windows app (download). You don't need a terminal or a domain of your own.

  1. A VPS. Ubuntu or Debian (x86_64 or arm64), 1–2 GB RAM, 3 GB disk, root access. Your hosting provider gives you the IP and root password.
  2. "Your own server" in the app. Enter the IP and root password. For the address, take a free name.dedyn.io from deSEC (a name and an email you confirm) or use your own domain with three A records.
  3. Install. About five minutes, steps shown on screen. The app runs the same installer over SSH, then opens the server setup: name, icon, administrator, invite link.

Password login is switched off after the install, and the app keeps its own key (in the system keychain). If the deSEC email doesn't arrive within 15 minutes, the app lets you try another email or your own domain.

Install by command

For your own domain, or if you prefer a terminal. You need:

  • A VPS on Ubuntu or Debian (x86_64 or arm64), root access. At least 1 GB of memory (2 is better), 3 GB of free disk, ports 80 and 443 free.
  • A domain with three A records pointing at the server IP: the domain itself, app. and a subdomain for calls. The installer picks that name and shows the records to add. On Cloudflare, use the grey cloud (DNS only): calls can't go through its proxy.

On the server:

curl -fsSL https://github.com/atriumchat/atrium/releases/latest/download/install.sh | sh

If GitHub is unavailable, the same release is on npm as the @atriumchat/install package (no Node needed on the server):

curl -fsSL https://cdn.jsdelivr.net/npm/@atriumchat/install@latest/install.sh | sh

The installer fetches the release archive from there too when GitHub doesn't answer, and checks the same SHA-256.

The wizard checks the server, asks for the domain, shows a summary and sets everything up: Docker (get.docker.com), ufw, SSH key login, .env with secrets, nightly backups, the release images. At the end it prints the link https://app.DOMAIN/?setup=CODE. Arrows and Enter choose, Esc goes back a step. Without a terminal or without gum, the same steps run as plain prompts.

Non-interactive, for scripts:

curl -fsSL https://github.com/atriumchat/atrium/releases/latest/download/install.sh | DOMAIN=chat.example.com ATRIUM_YES=1 sh

To update, run it again, or non-interactively:

curl -fsSL https://github.com/atriumchat/atrium/releases/latest/download/install.sh | sh -s update

To print the first-setup link (with the setup code) again, while setup isn't finished:

curl -fsSL https://github.com/atriumchat/atrium/releases/latest/download/install.sh | sh -s link

To restore data from a nightly backup (after a mistake or a broken server; on a new server, first copy the backups into /root/atrium-backups/daily):

curl -fsSL https://github.com/atriumchat/atrium/releases/latest/download/install.sh | sh -s restore

The installer lists the backups with their date and size and asks which one to restore (ATRIUM_YES=1 ATRIUM_RESTORE=2026-10-18 skips the question). It saves the current data as a pre-restore-<time> backup in the same folder, stops the app, puts the data back, starts the app and checks that it responds. To undo a wrong pick, run restore again and choose the pre-restore-… backup. If a step fails (a full disk, say), the data stay as they were and the app is started again.

A deleted channel doesn't need this: it stays in "Server → Deleted channels" for a week, and an admin can restore it from there.

To remove Atrium from the server:

curl -fsSL https://github.com/atriumchat/atrium/releases/latest/download/install.sh | sh -s uninstall

The installer asks for confirmation (ATRIUM_YES=1 skips it), stops the containers and removes their images, the /opt/atrium folder (.env, chats, files), the backups in /root/atrium-backups, the nightly timer, and its own copy and log in /root and /var/log. SSH login goes back to how it was: the 00-atrium-key-only.conf file and the app's key (the line commented atrium-app) are removed. Docker and ufw stay. Nothing removed can be recovered, backups included.

Updates leave .env and the data alone. More installer variables are described at the top of deploy/install.sh: SERVER_NAME, TURN_HOST, ATRIUM_DIR (default /opt/atrium), ATRIUM_NO_GUM=1, ATRIUM_FIREWALL=0, ATRIUM_SSH=0, and options for your own release archive or image registry.

The installer rewrites the cover page on the root domain on every update. To serve your own site, put it in /opt/atrium/landing/site along with an empty .custom file, and the installer will leave it alone. Pages open without .html: /about serves about.html.

Your own Caddy blocks go in *.caddy files in /opt/atrium/deploy/sites; updates leave them alone. For example, www.caddy to redirect from www. (needs a www A record):

www.{$DOMAIN}, http://www.{$DOMAIN} {
	redir https://{$DOMAIN}{uri} permanent
}

Your own routes inside the app (app.DOMAIN) go in *.caddy files in /opt/atrium/deploy/routes/app and take priority over the app's routes; updates leave them alone, and there are none by default. Every page of the app loads /owner.js, which is empty by default; to run your own script, serve it from a route there:

handle /owner.js {
	root * /etc/caddy/routes/app
	file_server
}

The app's CSP allows only its own origin, so anything that script talks to has to go through a route like this too.

To apply changes without dropping calls: cd /opt/atrium/deploy && docker compose exec caddy caddy reload --config /etc/caddy/Caddyfile --adapter caddyfile.