As long as you get a public IP address, a domain name and TLS certificate, you can federate with other servers.

I run mine in a small Hetzner box. Two gigabytes of RAM, two AMD cores, it’s almost like doing nothing even I’m federating a ton of messages all the time regarding the logs.

When I worked for Prisma, you can check our rust setup from the public flake:

https://github.com/prisma/prisma-engines/blob/main/flake.nix

CD to the project and nix-direnv loads the flake. Get to work.

Now when I’m working in Grafbase, our flake is a bit different:

https://github.com/grafbase/grafbase/blob/main/flake.nix

Instead of the Rust overlay, we use rustup and rust-toolchain.toml. This makes it easier to enforce the same Rust version for nix and non-nix users.

Both ways work really well. The deal is to define the rust env per project instead of defining it globally. Use direnv to make it working seamlessly.

I run my own SearXNG instance too. I set it up to a Hetzner box, then blocked all ports from the firewall except on the Tailscale network. This means the machine which wants to use the search needs to be connected first to the same Tailscale network. It allows me to prevent being blocked by the search providers for too much traffic and is been working great. I just open http://SearXNG from my browser and start searching.

NixOS unstable in my workstation and my laptop. Using sway on Wayland on top of all-AMD hardware. I play games with this setup and I write Rust and TypeScript for living. I love the customizability and the reproducibility of NixOS: I just clone my config and I have exactly the desktop I’ve always had, every little tool and customization included. If my hard drive fails, I just plug a new one and I am productive in about 15 minutes.

My sway desktop has been looking and working similarly for years, and before that I used i3 on Xorg for almost a decade. I like how the UI doesn’t really change that much.