✨ Just In Case

Resilience when the internet goes dark. An offline emergency-knowledge search app that keeps preparedness, austere medicine, engineering, and survival references available with zero network connection.

Install & availability

Overview

Just In Case is a locally-hosted offline emergency-knowledge search appliance. It bundles a verified manifest of 115 curated preparedness sources — military field manuals, austere medicine guides, food preservation techniques, water/power engineering handbooks, emergency communications protocols, and civil resilience documents — and makes them searchable in natural language with source-grounded citations.

It is powered by llama.cpp for in-process LLM inference and embeddings, MuPDF for native document text extraction, and SQLite with sqlite-vec and FTS5 for Reciprocal Rank Fusion (RRF) hybrid search.

Curated sources

115

Verified preparedness PDFs and TXTs in the core profile.

from knowledge areas

Knowledge domains

9

Survival through Transport, numbered 100 to 900.

from knowledge areas

Core profile

~350MB

The default download; ZIM and video profiles are optional extras.

from tiered profiles

Default GGUF models

~4.6GB

Gemma 4 E4B Mobile (about 4.3 GB) plus EmbeddingGemma (about 265 MB), fetched once.

from quickstart

Content Library & Tiered Profiles

Content is acquired using the built-in content fetcher (fetch-source-data.sh) supporting tiered download profiles:

The default profile is a small fraction of the optional archives, which is why the larger ones come with BitTorrent distribution:

BitTorrent & Magnet Distribution

For multi-gigabyte corpora, JIC includes a zero-dependency BitTorrent creator (helper-scripts/create-torrent.py) that can package arbitrary local folders into .torrent files and magnet URIs with public tracker fallbacks.

Categorization & Knowledge Areas

The 115 curated documents cover 9 core domains:

1. 100_Survival: FEMA disaster checklists, US Army Survival FM 21-76, Map Reading FM 3-25.26, Kearney Nuclear War Survival Skills, FM 5-103 Survivability & Fortifications, FEMA CERT manual.
2. 200_Medical: Hesperian Where There Is No Doctor / Dentist / Women, Emergency War Surgery (5th Ed), Austere Medicine, Special Operations Forces Medical Handbook (ST 31-91B), Emergency Childbirth.
3. 300_Food: USDA Complete Guide to Home Canning, Peace Corps ICE agricultural series (Field Crops, Soils, Grain Storage, Animal Traction, Fish Pond Culture, Beekeeping).
4. 400_Engineering: NREL Solar Resource, VITA Village Technology, Gravity-Flow Water Systems, SODIS, Army TMs (Hand Tools, Welding, Machine Tools, Hydraulics, Electrical, Plumbing FM 5-420, Roads FM 5-430, Generators).
5. 500_Comms: ARRL Emergency Comms, CISA NIFOG v2.02, CISA AUXFOG, FCC Part 97 Amateur Radio, FM 24-18 Tactical Radio, FM 24-19 Radio Operator’s Handbook.
6. 600_Education: OpenStax Elementary Algebra, Anatomy & Physiology 2e, College Physics 2e, Chemistry 2e.
7. 700_Social: Universal Declaration of Human Rights, US Constitution, Federalist Papers, King James Bible.
8. 800_Software: Think Python 2e, Think OS, Eloquent JavaScript 3e, Pro Git 2e.
9. 900_Transport: TM 9-8000 Automotive Vehicles, FM 21-305 Wheeled Vehicle Driver, FM 20-22 Vehicle Recovery, NAVEDTRA 14050 Construction Mechanic, USCG Seamanship & Navigation.

Architecture

flowchart LR
    B["Browser / Mobile Client"] -->|":8080"| S["jic-server
UI · /query · /status · /api/library"]
    S <--> DB[("jic-data volume
SQLite: vec0 + FTS5")]
    S --> V[("jic-sources volume
PDF · TXT · ZIM library")]
    I["jic-ingestion
MuPDF → chunk → embed"] --> V
    I --> DB
    F["content-fetch
(profile selector)"] -->|"sources.yaml & Torrents"| V
    M[("./gguf_models
read-only")] -.-> S & I

Install from the Hub App Store

The Hub app runs the same container as the Docker Compose route below. Its definition gives the container two folders from the app’s data directory: data, which holds the search index, and gguf_models, which holds the language and embedding models. The app does not download models and does not include the content fetcher, so a fresh install has no models and no documents.

Install

Search for Just In Case in the Hub App Store and install. The install form has no options.

Add the two model files

Place gemma-4-E4B-it-Q4_0.gguf (about 4.3 GB) and embeddinggemma-300M-qat-Q4_0.gguf (about 265 MB) in the app’s gguf_models folder, which is inside the app’s folder under app-data (see Folder structure). The download addresses are in helper-scripts/fetch-models.sh in the JustInCase repository . Then restart the app.

Open the app and check its status

Open Just In Case from its card on the Hub dashboard. Until both model files are present, the page opens but shows LLM missing and the message Language model not loaded — place the GGUF files in gguf_models/ and restart. The library stays browsable, and answers are unavailable.

Load documents

Load the curated library with the Docker Compose route below, or add your own PDFs and text files to a category folder under the app’s public/sources directory. The Hub app does not currently keep that directory outside the container, so documents you add there can be lost when the app is recreated.

Setup & Quickstart

This route runs from a clone of the JustInCase repository  on any machine with Docker and Docker Compose, and needs about 4 GB of disk for models plus room for your library.

./helper-scripts/fetch-models.sh

docker compose up --build -d

docker compose --profile fetch run --rm content-fetch

curl -s -X POST http://localhost:8080/query \
  -H "Content-Type: application/json" \
  -d '{"query": "How do I purify surface water with chlorine bleach?"}'

Links

• JustInCase repository  — public
• Folder structure — where Hub keeps each app’s data
• Spellbook
• Kiwix Upstream Mirrors