Getting Started with Devbox for Data Science¶

1. What is Devbox and Why Use It?¶

Devbox is a command-line tool that allows you to create isolated, reproducible, and portable development environments without relying on Docker or virtual machines.
It is powered by Nix, but without the complexity of Nix language.
For a data scientist, this means:
- Easy sharing of dependencies (Python, ML libraries, etc.) in a single devbox.json file.
- Consistency across machines: every teammate works with the same versions.
- Ability to experiment with tools without polluting your global system.

2. Installing Devbox¶

Open your terminal (Linux, macOS, or Windows with WSL2).

Run the following command (non-root user):

curl -fsSL https://get.jetify.com/devbox | bash

If Nix is not installed, Devbox will install it automatically.

3. Creating Your First Devbox Environment¶

Initialize Devbox in your project directory:
```
devbox init
```
This generates a devbox.json file.
Search for packages:
```
devbox search uv
```
Add uv (a fast Python package installer and environment manager) and data science libraries:
```
devbox add uv pandas
```
Start your environment:
```
devbox shell
```

Create and manage a Python environment with uv:

uv venv .venv
source .venv/bin/activate
uv pip install numpy scikit-learn

Exit the environment:
```
exit
```

4. Automating with Scripts and Hooks¶

You can customize workflows in devbox.json:

{
  "packages": [
    "uv",
    "pandas"
  ],
  "shell": {
    "init_hook": [
      "uv pip install -r requirements.txt"
    ],
    "scripts": {
      "run-notebook": "uv run jupyter notebook"
    }
  }
}

init_hook: runs commands when entering devbox shell.
scripts: custom commands launched with:
```
devbox run run-notebook
```

Example: automatically install dependencies or start Jupyter Notebook when entering the environment.

5. Managing Global Packages¶

Some tools are useful across all projects (like git or ripgrep). Install them globally:

devbox global add ripgrep git

To make them available outside of devbox shell, add this to your .bashrc or .zshrc:

eval "$(devbox global shellenv --init-hook)"

6. Best Practices for Data Scientists¶

| Step | Recommendation | | - | | | 1 | Always version-control devbox.json and devbox.lock. | | 2 | Use uv to manage Python environments and package installations. | | 3 | Use init_hooks to automate dependency installation with uv pip. | | 4 | Use scripts for tasks like launching Jupyter or running tests. | | 5 | Leverage devbox global for tools you need everywhere. | | 6 | Explore Devbox templates (like Jupyter setups) to kickstart projects quickly. |

Summary¶

Using Devbox with uv provides a modern, fast, and reproducible way to manage Python environments for data science. This workflow combines the reliability of Devbox with the speed of uv, ensuring smooth and consistent setups across projects.