Build a private offline knowledge library for emergencies, travel, field work, classrooms, and unreliable networks.
Offline Survival is a desktop app that downloads useful public knowledge packs, keeps them on your computer, indexes them for search, and can optionally run a local AI assistant over the material. It is designed so a non-technical person can choose a library size, press Download, and later search or share that library without needing the internet.
- Project Overview
- What It Does
- Main Functionality
- Download The App
- First Run
- Choosing A Library Size
- Search And Local AI
- Sharing An Offline Library
- What Gets Downloaded
- Privacy And Safety
- Contributing
- For Developers
- Releases
Offline Survival is a desktop tool for preparing a useful offline knowledge archive before you need it. It is meant for emergencies, travel, field work, classrooms, labs, clinics, workshops, homelabs, and any place where internet access may be slow, censored, unreliable, expensive, or gone.
The project is not a single static archive and it is not a cloud service. It is a library builder:
- you choose a library size/profile
- the app downloads public knowledge sources into a folder you control
- the app keeps the original files and prepares/indexes what it can
- you search and open that material locally
- you can optionally run Local AI over indexed local sources
- you can package a prepared library for another computer or USB drive
The goal is practical continuity: a person should be able to prepare a disk with survival, medical, repair, agriculture, engineering, reference, and education material and still use it when there is no network.
Offline Survival helps you create a local library with material such as:
- survival manuals and first-aid references
- Simple English Wikipedia and selected Wikipedia collections
- medicine, repair, farming, engineering, cooking, maps, textbooks, and practical how-to archives
- your own extra PDFs, EPUBs, Markdown, HTML, CSV, JSON, text files, and ZIM files
The app keeps the original files, prepares readable copies when needed, builds a local search index, and can package the result for another computer.
| Functionality | What it means |
|---|---|
| Profile-based library building | Choose from small emergency profiles through large civilization-recovery profiles. Profiles control which sources are downloaded and help keep disk usage predictable. |
| Download manager | Downloads selected public sources, records progress/state, supports retries/resume paths, and keeps files inside the chosen library folder. |
| Source catalog | Sources are defined in manifests with URL, license, attribution, expected size, category, tags, runtime behavior, and profile membership. |
| Local readers/openers | ZIM files can be served through Kiwix, PDFs open with the system viewer, and supported archives can be extracted and opened locally. |
| Indexing and search | Supported files are normalized into searchable text and indexed locally so the library can be searched without internet access. |
| Extra local knowledge | You can import your own PDFs, EPUBs, Markdown, HTML, CSV, JSON, text files, and ZIM files into the local library. |
| Optional Local AI | The app can install and run an app-managed Ollama runtime and recommended models. AI answers use indexed local context when available. |
| RAM safety guard | Local AI only starts when the selected installed chat model fits the currently available RAM and swap pressure is acceptable. |
| Share packages | Build a package containing selected downloaded sources, search data, and available app files for Windows, macOS, and Linux. |
| Cross-platform releases | GitHub Actions publishes Windows, macOS, and Linux builds for x64 and arm64, plus an all-platforms bundle. |
| Recovery and maintenance | The app tracks state in the library folder, marks interrupted downloads, can retry/clean partial downloads, and keeps generated data separate from source code. |
| Privacy by default | Libraries, indexes, imported files, logs, and optional model data stay local. Internal services bind to localhost by default. |
Go to the latest release:
https://github.com/carlospolop/offline-survival/releases/latest
Download the file for your computer:
| Your computer | Download |
|---|---|
| Windows | Offline-Survival-windows-x64 or Offline-Survival-windows-arm64 |
| macOS Intel | Offline-Survival-macos-x64 |
| macOS Apple Silicon | Offline-Survival-macos-arm64 |
| Linux Intel/AMD | Offline-Survival-linux-x64 |
| Linux ARM | Offline-Survival-linux-arm64 |
Each platform ZIP also contains the all-platforms app bundle, so the same download can be reused when preparing a USB drive for Windows, macOS, and Linux machines. The separate Offline-Survival-all-platforms archive is available if you only want the app files without choosing a primary platform.
- Open Offline Survival.
- Choose a Library path. This is where downloaded knowledge files will live.
- Press Easy Install if you want the recommended setup for your computer.
- Leave the app open while it downloads, verifies, prepares, and indexes files.
The app does not upload your library. Downloaded files stay in the folder you choose.
The profiles are ordered from small to large:
| Profile | Best for |
|---|---|
| Survival Essential | Small emergency library for limited disks. |
| Survival Plus | More reference material without going huge. |
| Civilization Core | A serious practical library for long offline periods. |
| Civilization Rebuild | Repair, engineering, science, and broader rebuilding material. |
| Civilization Max | The largest catalog option for big disks. |
The sidebar recommends the largest profile that fits a conservative disk budget for the current machine.
Search works after sources are indexed. Local AI is optional.
The Local AI panel can install a small chat model and an embedding model. Startup is protected by a RAM guard:
- the app checks available RAM before starting Ollama
- the check depends on the installed chat model
- if the machine is under memory pressure, Local AI stays blocked instead of freezing the computer
- normal search still works without AI
Local AI answers are grounded in indexed local sources. For medical, electrical, structural, chemical, or other high-risk topics, treat the answer as a pointer back to the cited source, not as professional advice.
The Share tab builds a package with the selected downloaded sources, search data, and available app files for Windows, macOS, and Linux.
Use this for:
- copying a prepared library to a USB drive
- moving the app to a computer with no internet
- preparing a classroom, clinic, workshop, or field laptop
When creating a share package, choose the primary target OS. The package still includes every available OS app folder, plus generated start files:
- Windows:
Run-Offline-Survival-Windows.bat - macOS:
Run-Offline-Survival-macOS.command - Linux:
Run-Offline-Survival-Linux.sh
Linux launchers can point directly at the included library. Windows and macOS installers may need one first-run step after installation: choose the included OfflineSurvival-Library folder as the library path.
For the smoothest mixed-OS package, extract a release ZIP first and, in the Share tab, choose its Offline-Survival-all-platforms folder as the app bundle folder.
The catalog is defined in manifests/sources/catalog.yaml. It stores source URLs, sizes, licenses, categories, and profiles. The repository stores the catalog and app code only.
The repository must not contain downloaded ZIMs, PDFs, archives, model files, local SQLite state, or generated release bundles. Those are ignored by .gitignore.
- Your downloaded library is local.
- Search indexes are local.
- Extra knowledge files are copied into your selected library path before indexing.
- Local AI runs through a local Ollama runtime when installed.
- The app binds its internal services to localhost by default.
- RAM checks prevent Local AI from starting when the selected model is unsafe for the current machine.
Want to add a feature, improve the app, or propose a new source for one of the library profiles? Read CONTRIBUTING.md.
The short version: add code/docs as normal, add downloadable sources through manifests/sources/catalog.yaml, choose profiles carefully, record license and attribution, and never commit downloaded PDFs, ZIMs, archives, local databases, model files, or generated builds.
Requirements:
- Node.js 22+
- npm
- Rust and Cargo
- platform dependencies for Tauri
Install and validate:
npm ci
npm run validateRun the web UI and backend during development:
npm run api
npm run devBuild the desktop app locally:
npm run build
npm run sidecars:prepare
npm run tauri:buildThe sidecar script prepares platform-specific backend binaries under app/src-tauri/binaries/. Those files are generated and ignored.
Every push to main that changes the app, manifests, packaging tools, package metadata, or release workflow runs the desktop release workflow. When the builds pass, GitHub Actions publishes a new release tagged like main-<run>-<sha> and marks it as the latest release, so the website download links under /releases/latest/download/... point at the newest available app packages.
Versioned releases can still be created from tags like:
git tag v0.1.0
git push origin v0.1.0The release workflow builds Windows, macOS, and Linux packages for x64 and arm64 where GitHub-hosted runners are available, uploads each platform artifact, and creates an all-platforms archive.
The GitHub Pages site is published from site/ and is available at: