Take It With You — Free Offline Copy

This whole guide is yours to keep. Download a package, put it on a USB drive or your phone, and it works forever with no internet. Free to download, free to share. If it's useful and you can spare it, a donation keeps it online and growing — but it's never required.

Choose a package

Lite~1 MB
The guide app only
  • All 27 sections, 14 tools, 36 skills
  • Full offline search
  • Bunker Bot (add your own model)
  • No reference PDFs, maps, or AI index
⬇ Download Lite

Both extract to a last-light-survival-guide/ folder. Open index.html in any browser, or serve it with python3 -m http.server for the offline AI, maps, and PWA install. Verify your copy with sh verify.sh.

+ Add the offline AI (optional)

Bunker Bot can run a real AI model fully offline. The easiest way is a llamafile — a single file that is the model and the server, no install, works on Windows/Mac/Linux. It's a separate download (1–3 GB) so the packages above stay small.

1
Download a llamafile — start with Llama-3.2-3B-Instruct (~2.3 GB; or the 1B, ~0.9 GB, for older hardware) from Mozilla's llamafile project or Hugging Face.
2
Make it runnable and start it:chmod +x *.llamafile && ./your-model.llamafile --server --nobrowser
3
Open the guide, click Bunker Bot, choose the Llamafile mode — it connects automatically. Ask away, fully offline.

⚠️ Small models can be wrong, especially on medical questions — Bunker Bot leans on the bundled books to stay grounded, but always verify anything critical.

+ Add Wikipedia, Gutenberg & world maps

Too big to bundle, but one click away: the Expansion Library links every offline knowledge base — all of Wikipedia and Project Gutenberg via Kiwix, and offline maps of anywhere on Earth.

This project is free and always will be. Hosting, maps, and the reference library cost a little to keep online — if it's helped you, chipping in keeps it alive and ad-free for the next person.
100% optional. The guide is yours either way.

Want to mirror or re-share it? Please do — it's licensed Creative Commons (CC BY-SA 4.0). Full source is on GitHub; for censorship-resistant copies, see the torrent / IPFS options in the project README.