OUTPOST Tactical – Offline AI Field Intelligence

THE IDEA

OUTPOST Tactical – Offline AI Field Intelligence

Type: Software Status: Beta Tech Stack: Python, FastAPI, Uvicorn, Ollama (local LLM/vision inference), LanceDB (RAG), SQLite, rasterio/shapely/pyproj, custom PMTiles reader, MBTiles/Planetiler, Meshtastic (LoRa mesh), pyrtlsdr (SDR), cryptography (AES-256-GCM), React 18, TypeScript, MapLibre GL, Zustand, Tailwind, Vite, Docker, GitHub Actions Problem Statement In civil protection, alpine rescue, expeditions or remote industrial deployments, a reliable network or cloud connection is often unavailable. Common AI assistants, map services and knowledge bases depend on exactly that connection and fail when offline. What is missing is an integrated platform that brings maps and terrain analysis, local AI, sensors, encrypted communication and knowledge together on an ordinary laptop — without network dependency, without telemetry and able to run on modest hardware. Description OUTPOST Tactical turns an ordinary laptop into an offline, AI-supported platform for situational reconnaissance and survival support in the field. A dynamic, manifest-driven module loader provides 19 self-contained modules — terrain analysis (elevation, slope, viewshed, routing), AI image and sound recognition via local Ollama, a RAG knowledge base, AES-256-GCM-encrypted LoRa mesh communication, a sensor mesh, weather forecasting, a radio/SDR reference, medical, resource and group management, and OSINT fusion (RSS, AIS ship and OpenSky flight data; the only module that needs the internet for new data). The async-first FastAPI back end unlocks features adaptively to the hardware via an “OUTPOST score” and five tiers (4–64 GB RAM), and degrades cleanly when models or devices are missing. Authentication works offline (PBKDF2 + HMAC tokens), and content packs (Wikipedia/OSM/SRTM/satellite) are SHA-256-verified and downloaded resumably. The system is operated through a React/MapLibre PWA with three tactical themes and 48 px touch targets. Version 0.1.0 comes with around 456 tests, CI and hardened Docker, but no E2E tests and no external security audit yet. Use Case A network-independent, data-sovereign command post on a laptop — maps, local AI, sensors and encrypted communication keep working where there is no internet and no cloud, for example in civil protection or on expeditions. Link: https://github.com/rawk7000/Outpost (private repo)