Find your next open-source project

Search by name, purpose, or keyword — or paste a GitHub repository URL

Immortal-Protocols/lastbook

Productivity & Utilities

Offline desktop library with manuals, videos, maps, guides and optional local AI Q&A for disaster-prepared households.

285
35
Python
MIT
4 days old

What it is

Lastbook is a desktop application that assembles a searchable offline reference library: PDF manuals, videos, ZIM encyclopedias, a world/US base map with one selected state's detailed map, and short curated guides organized by urgent need. Chat answers questions from the library and shows the pages it cites, and an optional local model can generate answers on the machine itself. It runs on macOS (Apple Silicon), Windows 64-bit and Ubuntu 24.04 x86_64.

The application does not expose a listening port. The UI and the engine communicate through an in-memory bridge, and network access is limited to two person-initiated actions: downloading library packs, maps and the AI model during setup, and the manual update check. Downloads go through HTTPS to allow-listed hosts, are validated against SHA-256 manifest checksums, and new data is unpacked and swapped in beside the existing library.

Quick start

On Ubuntu 24.04, download the Ubuntu package and install it with apt so dependencies are resolved:

sudo apt install ./Lastbook_1.1.9_amd64.deb

Then open Lastbook from the application menu. On first run, setup asks you to choose a library size (Small, Medium or All), a US state map, and optionally the local AI, then downloads and verifies the selected data. You can instead point it at a library folder or pack zip you already have. macOS and Windows builds are distributed from lastbook.org.

To run the desktop app from source, you need Python 3.11+ and an existing data folder (a library-data/catalog.json plus a derived/ folder):

python3 -m venv .venv
.venv/bin/pip install -r packaging/requirements_macos.txt   # Windows: packaging/requirements_app.txt
.venv/bin/python app/main.py --data /path/to/LastbookData

Without --data, the app checks LASTBOOK_DATA and then its own data folder; with no library it opens setup. There is also a library-server mode that runs the same engine in a browser on port 8890 for devices on the same Wi-Fi; see the library server notes.

Constraints and limits

  • The Linux package targets Ubuntu 24.04 LTS x86_64 only. It was tested on Ubuntu 24.04.5 under WSL2, including installation, playback, maps, address search, PDFs and CPU-only AI; native Linux GPU performance and other distributions have not been validated.
  • The local AI is optional and large: a pinned Gemma 4 26B A4B GGUF of about 17 GB, offered only when the machine has roughly 18 GB of memory for model and context plus a reserve. It is downloaded from Hugging Face, not bundled with the app.
  • Meaning-based search uses bge-small-en-v1.5 on Windows only; the Mac build uses keyword search.
  • The detailed map covers one US state chosen during setup. Lastbook does not detect location.
  • The library itself (hundreds of gigabytes of third-party manuals, videos and encyclopedias) is not in the repository and ships as packs with their own source and permission listings. The original code and documentation are MIT-licensed, while bundled components such as libzim and third-party content keep their own licenses.
  • Removing the Linux package leaves downloaded libraries and settings in your home folder.

HotRepo's take

Lastbook entered the HotRepo new-project list on 2026-11-16 with 223 stars and 108 stars gained over a roughly 1.5-hour window, alongside 25 forks. That is a fast early burst with no earlier ranking history to compare against, so it says little about sustained interest. The more notable engineering signal is the documented offline-first design: a portless bridge, pinned TLS roots, checksum-verified packs and atomic library swaps, plus a signed Windows 1.1.8 release whose 90 installed binaries all carry valid signatures. Those claims are worth checking against the code and release artifacts before relying on them.

Written by AI from the project's README, updated . The project's own docs are the source of truth.

You might also like

  1. anonnerdfrenzy/survival-llmAsk an offline AI assistant for emergency and medical guidance drawn from a locally stored survival library, with visible sources.Python694
  2. Crosstalk-Solutions/project-nomadPackage offline Wikipedia, courses, maps, notes and optional local AI into a self-hosted Docker stack on hardware you own.TypeScript41,318
  3. Nutlope/byggaTurns an IKEA product name or article number into an animated, narrated 3D assembly guide.TypeScript348