Project NOMAD packs offline Wikipedia and local AI on a PC

Project NOMAD is aimed at the gap that opens when internet access is unreliable: the moment you most need reference material is often the moment you cannot reach it. The package bundles an "Information Library" for Wikipedia and other reference content, an Offline Maps feature based on OpenStreetMap, and an "Education Platform" built around the Kolibri learning software with math and science materials. The trade-off is coverage: the reviewer noted that no Japanese-language teaching materials are included, so usefulness depends on which content sets a user downloads.

The AI side is deliberately local. AI Assistant connects to locally run models through Ollama or LM Studio, and the Knowledge Base adds a RAG layer so the model can draw on uploaded PDFs and similar documents. In testing, the assistant eventually surfaced the code "BLUE-123" from a test document, though answer quality and the number of exchanges needed varied with the model used. That fits the project's stated design of running entirely on hardware the user owns.

Deployment is not turnkey. The reviewer set it up under WSL2 with Ubuntu on a Windows 11 machine with an NVIDIA GPU, fetched an installer script, applied a change for Docker Desktop, and ran it with sudo. Two rough edges stood out: Offline Maps showed nothing after the roughly 2.5GB Japan map download, and a benchmark had not completed after more than 30 minutes, so no result was available. Because the dashboard lacks built-in login authentication, the reviewer advised against exposing it directly beyond the local network.