DEV Community

정상록
정상록

Posted on

Project NOMAD: Building an Offline Survival Computer with Docker, Local AI, and 99.6GB of Wikipedia

#ai #opensource #showdev #docker

TL;DR

Project NOMAD is a Docker-based offline survival computer that bundles local AI (Ollama), full Wikipedia, Khan Academy, OpenStreetMap, and security tools. Free, open-source (Apache 2.0), runs on any Linux PC. Hit #1 on GitHub trending with 11,300+ stars.

What is Project NOMAD?

Have you ever wondered what happens when the internet goes down? Not for an hour, but permanently?

Project NOMAD (Node for Offline Media, Archives, and Data) answers that question with code. It's a self-contained, Docker Compose-based system that runs an entire knowledge infrastructure offline.

Created by Chris Sherwood of Crosstalk Solutions (475K YouTube subscribers, 20+ years in networking), NOMAD hit #1 on GitHub trending on March 21, 2026 — gaining 2,054 stars in a single day.

The Tech Stack

Here's what's running under the hood:

# Core services in docker-compose.yml
services:
  ollama:        # Local LLM runtime
  open-webui:    # AI chat interface
  qdrant:        # Vector DB for RAG
  kiwix:         # Offline Wikipedia engine
  kolibri:       # Education platform
  cyberchef:     # Security/encoding tools
  flatnotes:     # Markdown notes
  dozzle:        # Container monitoring
  command-center: # Management dashboard
Enter fullscreen mode Exit fullscreen mode

Each service runs in its own Docker container. You can enable/disable services based on your hardware and needs.

Key Features

1. Local AI with GPU Acceleration 

# Pull a model
docker exec -it ollama ollama pull llama3.2:1b

# For more capable AI (needs 16GB+ RAM)
docker exec -it ollama ollama pull llama3.2:8b
Enter fullscreen mode Exit fullscreen mode

The AI stack includes:

  • Ollama for running LLMs locally
  • Open WebUI for a ChatGPT-like interface
  • Qdrant vector database for RAG (upload documents and query them)
  • NVIDIA Container Toolkit support for GPU acceleration

No API calls. No telemetry. Complete data sovereignty.

2. Offline Knowledge Library

Kiwix serves the entire Wikipedia (99.6GB) as ZIM files, plus:

  • 75,000 books from Project Gutenberg
  • Medical references
  • Survival manuals

3. Education Platform

Kolibri provides Khan Academy courses with:

  • Progress tracking
  • Multi-user support
  • Offline content delivery

4. Maps and Security

  • OpenStreetMap: Offline maps with routing
  • CyberChef: 200+ data transformation operations in the browser

Installation 

# Clone and run
git clone https://github.com/Crosstalk-Solutions/project-nomad.git
cd project-nomad
docker compose up -d

# Access at http://localhost:8080
Enter fullscreen mode Exit fullscreen mode

That's it. After the initial content download, no internet required.

Hardware Requirements

Spec Minimum Recommended
CPU Dual-core Ryzen 7 / i7+
RAM 4GB 32GB
Storage 5GB 1TB SSD
GPU None RTX 3060 12GB+
OS Ubuntu 22.04+ Ubuntu 22.04+

An old PC you have lying around can run the base features. GPU is only needed for AI acceleration.

How It Compares

Feature NOMAD PrepperDisk ($199-699) Doom Box ($199-699)
Hardware Any PC RPi only RPi only
AI GPU-accelerated LLM None None
License Apache 2.0 Partial Proprietary
Wikipedia Full (99.6GB) Full Limited
Education Khan Academy Khan Academy None

NOMAD is free, runs on any hardware, and includes AI. The paid alternatives can't compete on features.

Community Growth

  • 11,300+ GitHub stars
  • 1,000+ forks
  • Hacker News front page
  • Community forks: MacOS version, Homelab edition, RPi adaptation

Why It Matters

This isn't just a prepper tool. It's a statement about digital sovereignty and offline-first computing. In an era where everything depends on cloud services, NOMAD proves that a fully functional knowledge infrastructure can run on a single machine.

GitHub: Crosstalk-Solutions/project-nomad

Have you set up NOMAD or a similar offline system? I'd love to hear about your experience in the comments.

Top comments (0)