EmberCart
Gemini 4 NYC HackathonEmberCart ("PyroBiome-4D + FlashHold-LCC") helps Front Range and wildland-urban interface (WUI) households prepare for wildfire smoke, ash, watershed runoff, and multi-stage recovery without manual guesswork or unapproved purchases. Rather than predicting exact future lightning bolts, our digital twin ranks likely ignition corridors (IGZ-01 through IGZ-05 across Nederland Subalpine Ridge, Sugarloaf/Gordon Gulch, Gold Hill/Fourmile, Magnolia Rd/Boulder Canyon, and Flagstaff/Gregory alongside historical Colorado fire benchmarks) and pairs atmospheric/LCC/smoke timing (49-minute smoke ETA, 252.9 peak PM2.5) with ground, soil hydrophobicity (0.72), watershed ash, and primary-succession-style recovery stages on hydrophobic, ash-impacted soils. Deterministic Python modules (src/plume_watcher.py and src/needs_rules.py) calculate plume timing and size non-medical preparedness kits across 7 categories (air, home, pet, and plant/soil supplies only—never medical advice). Orchestration pairs Gemini/Antigravity with ?replay=1 fallback and live Shop MCP staging (mcp.shop.com, 2026-07-28), keeping continue_url in memory (never logged, all query parameters preserved) until explicit human approval.





