Skip to content

Repository files navigation

IronMiner

React TypeScript Vite Python FastAPI Supabase LiveKit Android Docker

About

IronMiner is a spatial intelligence platform for construction site management built for UMD x IronSite 2026. Three AI agents — powered by the Claude API and GPU processing via Vast.ai — analyze job site video footage to detect congestion, safety violations, and productivity bottlenecks. The React/TypeScript frontend and FastAPI backend are complemented by an Android headset client for live worker streaming over LiveKit, with persistent data stored in Supabase.

Features

  • Video Agent ingests site footage, extracts frames, and runs spatial analysis via GPU (Vast.ai)
  • Safety Agent detects PPE non-compliance, zone violations, and blocked corridors
  • Productivity Agent scores congestion per zone, identifies trade overlap, and suggests resource reallocation
  • Review Mode: upload footage, receive AI briefings, browse zone congestion maps and alert feeds
  • Live Mode: monitor real-time camera feeds, communicate with workers, and receive instant alerts via WebSockets
  • Android headset client streams live video from workers using LiveKit audio/video
  • BIM-style zone overlay visualizes spatial layout directly in the dashboard
  • Teams board tracks worker assignments and pool allocation across active sites

Technology Stack

  • Frontend: React, TypeScript, Vite
  • Backend: Python, FastAPI, WebSockets
  • AI: Claude API (vision + text), Vast.ai (GPU inference)
  • Database: Supabase (PostgreSQL)
  • Live Comms: LiveKit
  • Mobile: Android (Kotlin)
  • Containerization: Docker

About

UMD x IronSite 2026 | Construction Site Intelligence

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages