Have you ever wished you could effortlessly recreate your entire computer setup on any machine, regardless of its operating system? Imagine a world where sharing your exact development environment with colleagues is as simple as sharing a text file. Enter Nix and NixOS, revolutionary tools that are changing the game in system management and software development. Whether you’re a seasoned programmer or just starting out, this article will introduce you to a powerful approach that can save you countless hours of setup time, eliminate “it works on my machine” problems, and give you unprecedented control over your computing environment. Get ready to discover how Nix can transform the way you think about software installation, system configuration, and project reproducibility.
Introduction to Nix
Nix is a funny thing. It’s 3 things, really. It’s a software packaging system, it’s an operating system, and it’s a language for describing Linux systems, or parts of them. And that makes it a rather powerful way to express systems, scratching an itch I’ve had for years. How do you make just about everything generic and portable, de-coupled from hardware?
I’ve been down the virtual machine and container path. You can check out Levinux, an old endeavor of mine that’s still a fun curiosity. But the itch it was scratching is addressed by Nix. Linux system reproducibility. And as an extra bonus, you’ve got it on Macs and Windows, not so much as reproducing a Mac or Windows system on other machines, but reproducing a Linux machine as if in the folder on Mac or Windows.
The Power of Nix
I know that’s a bit abstract. But it’s great for trading systems. Not everything needs to be a Web app or something in an app store. You can still just create and trade software between computers, not much worrying what kind of computer, so long as it’s somewhat modern. Actually, old or barely just Internet capable is fine too. Nix makes trading systems easy because of high reproducibility.
The barriers are mostly just wrapping your mind around it all. There’s techie reasons it all works so well, having to do with something called a /nix/store/
which can appear just like that as a path even on Macs and Windows, and from there, everything is packed flat, hierarchy-wise. You don’t have to think about it much when you’re editing text files that define systems.
Understanding NixOS
If your sytem is fully Nix-built, it’s likely NixOS. NixOS is a distribution of Linux that installs from self-booting media like USB and CDs, just like any other Linux. But after that, you can take some fairly serious control of your system from a file /etc/nixos/configuration.nix
. You could in fact build your entire system out from there, including all your development environments and home directory stuff. But generally, NixOS users break it out to a few more files for organization.
With NixOS, especially if you’re using git and working out of folders that are software repositories, it’s easy to allow yourself to think of installing software into the repo. If the same software is used in folder after folder and repo after repo, it’s still only stored once, because of /nix/store/
. And therefore, when you’re creating your main /nixos/configuration.nix
it feels right leaving most things out that you won’t need until later.
My NixOS Configuration
Some of the exceptions I make are getting a few global services running, which in my mind now includes a running Ollama LLM. It also includes the specialized software for accelerated Nvidia CUDA hardware performance. From there, any repos that require a global LLM server can easily connect to it, and it’s receiving the GPU speed-boost, as could any GPU-acceleratable components like Pandas. And for this reason a relatively simple /nixos/configuration.nix
could look like this…
{ config, pkgs, lib, ... }:
{
# Import hardware-specific configuration
imports = [
./hardware-configuration.nix
];
# Set the system state version
system.stateVersion = "unstable";
# Boot loader configuration
boot.loader.systemd-boot.enable = true;
boot.loader.efi.canTouchEfiVariables = true;
# Enable experimental features for Nix
nix.settings.experimental-features = [ "nix-command" "flakes" ];
# Networking configuration
networking.hostName = "nixos";
networking.networkmanager.enable = true;
# Set time zone and locale
time.timeZone = "America/New_York";
i18n.defaultLocale = "en_US.UTF-8";
# X Server and Desktop Environment configuration
services.xserver = {
enable = true;
displayManager.gdm.enable = true;
desktopManager.gnome.enable = true;
xkb.layout = "us"; # Keyboard layout
videoDrivers = ["nvidia"];
};
# Graphics configuration (updated from OpenGL)
hardware.graphics = {
enable = true; # Renamed from hardware.opengl.enable
enable32Bit = true; # Renamed from hardware.opengl.driSupport32Bit
};
# NVIDIA configuration
hardware.nvidia = {
modesetting.enable = true;
powerManagement.enable = true;
powerManagement.finegrained = false;
open = false;
nvidiaSettings = true;
package = config.boot.kernelPackages.nvidiaPackages.stable;
};
# Allow unfree packages and enable CUDA
nixpkgs.config.allowUnfree = true;
nixpkgs.config.cuda.enable = true;
# Audio configuration
# sound.enable = true;
hardware.pulseaudio.enable = false;
security.rtkit.enable = true;
services.pipewire = {
enable = true;
alsa.enable = true;
alsa.support32Bit = true;
pulse.enable = true;
};
# User configuration
users.users.mike = {
isNormalUser = true;
description = "Mike";
extraGroups = [ "networkmanager" "wheel" ];
};
# Auto-login configuration
services.displayManager.autoLogin = {
enable = true;
user = "mike";
};
# Enable Firefox
programs.firefox.enable = true;
# System packages
environment.systemPackages = with pkgs; [
code-cursor
coreutils
figlet
curl
git
gnome-terminal # Updated from gnome.gnome-terminal
google-chrome
libreoffice
msmtp
neovim
obs-studio
screen
vim
wget
xclip
feh
(python311.withPackages (ps: with ps; [
simplenote
]))
cudaPackages.cudatoolkit
ollama
];
# Environment variables
environment.variables = {
CUDA_HOME = "${pkgs.cudaPackages.cudatoolkit}";
};
# Power management configuration
services.logind = {
lidSwitch = "ignore";
extraConfig = ''
HandleSuspendKey=ignore
HandleLidSwitch=ignore
HandleLidSwitchExternalPower=ignore
IdleAction=ignore
'';
};
# Ollama service configuration
services.ollama = {
enable = true;
acceleration = "cuda";
};
# Git daemon configuration
services.gitDaemon = {
enable = true;
basePath = "/srv/git";
exportAll = true;
user = "mike";
group = "users";
};
# Create Git repository directory
systemd.tmpfiles.rules = [
"d /srv/git 0755 mike users -"
];
# Open firewall port for Git daemon
networking.firewall.allowedTCPPorts = [ 9418 ];
# Create symlinks for Vim and Neovim configurations
system.activationScripts.createVimNvimSymlinks = {
text = ''
mkdir -p /root/.vim
ln -sf /home/mike/.config/nvim/init.vim /root/.vim/vimrc
mkdir -p /root/.config/nvim
ln -sf /home/mike/.config/nvim/init.vim /root/.config/nvim/init.vim
mkdir -p /home/mike/.vim
ln -sf /home/mike/.config/nvim/init.vim /home/mike/.vim/vimrc
'';
};
# Shell aliases
environment.shellAliases = {
vim = "nvim";
xc = "xclip -selection clipboard <";
xv = "xclip -selection clipboard -o >";
};
}
Creating a Data Science Environment
From here, I have a template which creates a data science repo from which you can run JupyterLab and such. Interestingly, even though huge amounts of the Python PyPI pip installable repository is in the Nix packaging system too and could be installed “native” on Nix in one of these flake.nix
files, I still tend to set up Python virtual environments for a few reasons.
{
inputs = {
nixpkgs.url = "github:nixos/nixpkgs/nixos-24.05";
flake-utils.url = "github:numtide/flake-utils";
};
outputs = inputs @ { self, nixpkgs, flake-utils, ... }:
flake-utils.lib.eachDefaultSystem (system: let
pkgs = import nixpkgs {
inherit system;
config.allowUnfree = true;
};
localConfig = if builtins.pathExists ./local.nix then import ./local.nix else {};
cudaSupport = if localConfig ? cudaSupport then localConfig.cudaSupport else false;
cudaPackages = pkgs.lib.optionals (cudaSupport && system == "x86_64-linux") (with pkgs; [
cudatoolkit
cudnn
(ollama.override { acceleration = "cuda"; })
]);
envWithScript = script:
(pkgs.buildFHSUserEnv {
name = "python-env";
targetPkgs = pkgs: (with pkgs; [
python311
python311.pkgs.pip
python311.pkgs.virtualenv
pythonManylinuxPackages.manylinux2014Package
cmake
ninja
gcc
git
stdenv.cc.cc.lib
] ++ cudaPackages);
runScript = "${pkgs.writeShellScriptBin "runScript" (''
set -e
export NIXPKGS_ALLOW_UNFREE=1
export LD_LIBRARY_PATH=${pkgs.stdenv.cc.cc.lib}/lib:$LD_LIBRARY_PATH
echo "Welcome to the Pipulate development environment on ${system}!"
${if cudaSupport then "echo 'CUDA support enabled.'" else ""}
test -d .venv || ${pkgs.python311.interpreter} -m venv .venv
source .venv/bin/activate
pip install --upgrade pip --quiet
pip install -r requirements.txt --quiet
${script}
'')}/bin/runScript";
})
.env;
in {
devShell = envWithScript "bash";
});
}
Managing Python Dependencies
Such a method requires that a standard pip requirements.txt
file is sitting there next to it. Here’s what I often start out with for basic AI-enabled Juptyer Notebooks and a cool new Pythonic web development package…
requests
numpy
pandas
sqlitedict
matplotlib
python-fasthtml
jupyterlab
jupyter_ai
nbdev
By taking this approach, you get the expected ability to pip install
as a Python user without running up against the Nix frustrations of immutable OS where you normally can’t just install things. This approach also lets you get the very latest out of PyPI (pip) that may not have reached the Nix packaging system yet, such as FastHTML and JupyterAI.
What you loose when you use normal pip is some of the determinism (almost guaranteed to run) of Nix. But the trade-off, especially when you’re doing fast-and-furious Python cobbling, is worth it.
Simplifying the Flake
The above flake is still a bit more complicated than it needs to be, because of looking for GPU stuff. This can be simplified even a little more if you don’t have to sort out accelerated hardware…
{
inputs = {
nixpkgs.url = "github:nixos/nixpkgs/nixpkgs-unstable";
flake-utils.url = "github:numtide/flake-utils";
};
outputs = inputs @ {
self,
nixpkgs,
flake-utils,
...
}:
flake-utils.lib.eachDefaultSystem (system: let
pkgs = import nixpkgs {inherit system;};
envWithScript = script:
(pkgs.buildFHSUserEnv {
name = "python-env";
targetPkgs = pkgs: (with pkgs; [
python3
python3Packages.pip
python3Packages.virtualenv
# Support binary wheels from PyPI
pythonManylinuxPackages.manylinux2014Package
# Enable building from sdists
cmake
ninja
gcc
pre-commit
]);
runScript = "${pkgs.writeShellScriptBin "runScript" (''
set -e
test -d .nix-venv || ${pkgs.python3.interpreter} -m venv .nix-venv
source .nix-venv/bin/activate
set +e
pip install --upgrade pip --quiet
pip install -r requirements.txt --quiet
''
+ script)}/bin/runScript";
})
.env;
in {
devShell = envWithScript "bash";
});
}
Conclusion
And there it is. That’s a system and flake template. All you need is one /etc/nixos/configuration.nix
to define your system, and then you can sprinkle flake.nix
files throughout your system and share them with others.