Zonivra

Zonivra is a local-first industrial vision intelligence platform designed to make AI-assisted safety monitoring more transparent, traceable, and accountable. The application combines computer vision powered by RF-DETR Nano, configurable monitoring zones, deterministic risk evaluation, and evidence review with human oversight. Instead of relying on AI to make autonomous safety decisions, Zonivra separates visual observations from rule-based risk assessments and preserves the evidence behind each result. For the Meta Global AI Developer Hackathon 2026, we extended our existing MineGuard prototype with Meta SAM 3.1 integration capabilities. Zonivra can import, validate, and review results generated externally by SAM 3.1, including image-linked bounding boxes, structural segmentation information, SHA-256 integrity verification, and explicit provenance records. The platform also demonstrates previously recorded computer vision analysis and reproducible industrial risk scenarios, clearly distinguishing real observations from synthetic simulations. Our vision is to create a trustworthy industrial intelligence platform for mining, construction, warehouses, and other high-risk environments, where AI assists human operators without replacing their judgment. **What makes Zonivra different:** AI observes, deterministic rules evaluate, evidence remains verifiable, and humans retain oversight. Zonivra is currently an experimental decision-support prototype, not a certified safety system. Live in-app Meta inference is not claimed.