Skip to content

A plug-and-play Debian Live ISO with Ollama preinstalled — bootable from USB, air-gapped by design, and ready to run large language models offline.

License

Notifications You must be signed in to change notification settings

vedmakk/ollama-rucksack

Repository files navigation

Ollama Rucksack Logo

🎒 Ollama Rucksack

A plug-and-play Debian Live ISO with Ollama preinstalled — bootable from USB, air-gapped by design, and ready to run large language models offline.

Latest Release ISO Bootable Build License

🎯 What Is This?

Ollama Rucksack is a self-contained, bootable Debian Live ISO preloaded with Ollama and open-source language models.

  • No internet required.
  • No installation needed.
  • Just boot and run.

It's ideal for:

  • Exploring LLMs without touching your main OS
  • Educational or lab use
  • Developers who want a clean, minimal LLM sandbox
  • Bringing AI to places where internet is limited or unavailable

Whether you're on a train, in the mountains, or just dodging flaky Wi-Fi, your models go where you go.

🌍 Why It Matters

Access to AI shouldn't depend on access to the cloud.

With Ollama Rucksack, anyone with a basic computer can boot into a ready-to-use LLM environment — no installation, no internet, no setup. This makes it possible to bring powerful language models to places where connectivity is poor and infrastructure is limited.

  • In rural schools without stable internet
  • In disaster zones or field research setups
  • For learners and developers with older machines
  • For anyone who wants to explore AI offline and on their terms

By packaging open models into a portable, air-gapped ISO, Ollama Rucksack helps democratize access to human knowledge — using just a USB stick.

📦 What's Inside?

  • 🧠 Ollama – Preinstalled and ready to run
  • 🗂️ Preloaded models – Choose your own or bundle them into the ISO
  • 🖥️ XFCE Desktop – Lightweight and fast
  • 📴 Air-gapped by default – Network drivers are blacklisted for maximum isolation (but easily reversible if you want)

This isn't a secure communication platform or privacy tool — just a portable LLM lab you can boot from anywhere.

🚀 Quick Start

Head over to the Releases page. There you'll find:

  • The latest ISO files hosted on archive.org.
  • A sha256sums.txt file to verify the download.

Step-by-step:

  1. Download the ISO(s) you want.
  2. Download the SHA256 checksum file into the same directory.
  3. Verify your ISO:
sha256sum -c sha256sums.txt

Then burn it to your USB stick like it's 2009:

sudo dd if=/path/to/iso of=/dev/sdX bs=4M status=progress oflag=sync

Replace /path/to/iso and /dev/sdX with your actual values. ⚠️ Triple-check the device or you'll be reformatting your cat pictures.

Or, build your own ISO. Instructions below.

🛠️ Build your own ISO

Prefer to customize or verify everything yourself? You can build the ISO locally using the provided scripts.

Requirements

  • Debian-based host system (e.g., Debian, Ubuntu)
  • Root privileges
  • Docker installed
  • make installed (sudo apt-get install make)
  • Internet connection (for building only)
  • Pre-downloaded Ollama models

Build Instructions

Clone the repository and run the setup script:

make setup

Then build the ISO:

sudo make build

# or

sudo make build -- [OPTIONS]

Use --help for all available options.

Example

sudo make build -- \
  --models-path ~/.ollama/models \
  --no-usb

Note: The -- is required before any options.

Output

The generated ISO image containing all tools and configurations: ./dist/ollama-rucksack-<version>-<arch>.iso

🧩 Ollama Models

To bundle models into your custom ISO:

  1. Install Ollama on your host.
  2. Run ollama pull <model> for each model you want.
  3. Point the build script at the model directory:
sudo make build -- --models-path /home/user/.ollama/models

Models will be copied into the ISO.

Note

The default path for Ollama to save pulled models when you installed it using the official install script for linux is /usr/share/ollama/.ollama/models. If you installed Ollama using a different method, you need to specify the correct path. Typically ~/.ollama/models.

🔧 Customizing the Live System

Customize the live system by editing the build.sh script file.

You can tweak:

  • Installed packages in config/package-lists/custom.list.chroot
  • Kernel module blacklist in config/includes.chroot/etc/modprobe.d/blacklist-network.conf
  • Desktop README (--readme path)

You can even disable the network lockdown entirely — just modify the config before build.

🤝 Contributing

Ideas, improvements, PRs — all welcome. If you want to help make this ISO better, faster, or more flexible, open an issue or submit a pull request.

📜 License

MIT — use freely, modify openly, and share widely. See the LICENSE file for details.

About

A plug-and-play Debian Live ISO with Ollama preinstalled — bootable from USB, air-gapped by design, and ready to run large language models offline.

Resources

License

Stars

Watchers

Forks