Survival LLM runs an emergency library on your laptop

Someone built Survival LLM, an offline AI assistant designed for emergencies. It is a local language model bundled with survival manuals, first aid guides, water purification instructions, and radio operation handbooks. The entire package fits on a laptop with no internet required. The interesting bit is not the technical stack. Ollama for the model, SQLite for search, a web UI. Standard pieces. What caught my attention is the use case. Most people building local LLMs optimise for privacy or cost. This optimises for grid failure. The bundled documents include the US Army Survival Manual, WHO medical guidelines, and amateur radio licensing materials. Practical text that might actually matter if infrastructure collapses. The model answers questions by retrieving chunks of these documents, not by hallucinating survival advice from training data. I have been thinking about offline-first systems since working on network monitoring at AT SecurDI. We logged incidents locally before syncing to cloud storage because connectivity dropped during the incidents we were meant to detect. Same principle here. The assistant you need during a disaster is the one that does not depend on AWS being reachable. The repo includes instructions for adding your own documents. You could drop in local building codes, grid maps, contact lists. It becomes a queryable backup of information you might actually need when you cannot Google it. This feels less like a prepper novelty and more like a forcing function. If you design a system assuming the internet will disappear, you end up with something that works reliably when the internet is just slow. The offline constraint makes better software. I would add network diagnostics guides and a local mirror of network protocol RFCs. If the scenario is grid-down, you are probably troubleshooting why your mesh network will not route.

---

Source: Survival LLM: AI assistant and emergency library that runs on your own computer