Skip to main content

Installation

info

RepX is currently in Active Development. APIs may change between minor versions.

Prerequisites​

To use RepX effectively, you need:

  1. Nix: The foundation of RepX's reproducibility.
    • Requirement: Flakes must be enabled.
  2. Rust (Optional): Only if building the CLI from source.
  3. Python 3.10+ (Optional): For analyzing results with repx-py.

Installing the CLI​

Nix is the recommended way to install RepX. It guarantees that your environment matches the experiment definition exactly.

To install the repx CLI globally on your system using Nix profiles:

nix profile install github:repx-org/repx

Verify the installation:

repx --version

Method 2: NixOS / Home Manager​

You can add RepX to your system configuration declaratively.

flake.nix:

{
  inputs.repx.url = "github:repx-org/repx";
  # ...
}

configuration.nix or home.nix:

environment.systemPackages = [
  inputs.repx.packages.${pkgs.system}.default
];

Method 3: Cargo (Rust)​

If you prefer using Rust's package manager, you can install the CLI directly from source.

# Install from Git
cargo install --git https://github.com/repx-org/repx.git repx-cli

Setting up a Project​

To use RepX in a project, you need to add it as an input to your flake.nix. This allows you to use the RepX Library (repx.lib) to define your experiments.

flake.nix:

{
  description = "My RepX Experiment";

  inputs = {
    nixpkgs.url = "github:NixOS/nixpkgs/nixos-unstable";
    repx.url = "github:repx-org/repx";
  };

  outputs = { self, nixpkgs, repx }: {
    # 1. Define your Lab (Experiments)
    # See 'User Guide > Defining Experiments' for details.
    
    # 2. (Optional) Create a development shell with the CLI and Python client
    devShells.${nixpkgs.system}.default = nixpkgs.legacyPackages.${nixpkgs.system}.mkShell {
      buildInputs = [
        repx.packages.${nixpkgs.system}.default  # The CLI
        repx.packages.${nixpkgs.system}.repx-py  # The Python Client
      ];
    };
  };
}

Installing the Python Client (repx-py)​

The Python client allows you to query experiment results and load data into Pandas.

Inside a Nix Shell (Best Practice)​

The best way to use repx-py is to include it in your experiment's devShell (as shown above). This ensures the library version matches your experiment and CLI.

Manual Install (pip)​

If you are working outside of Nix (e.g., in a global Conda environment or Jupyter Notebook), you can install repx-py directly from GitHub:

pip install "git+https://github.com/repx-org/repx.git#subdirectory=python/src"

Upgrading​

To upgrade RepX to the latest version:

Nix Flake:

nix flake update repx

Nix Profile:

nix profile upgrade repx