Okyanus/pomona-greenhouse-demo

Pomona -- Guarded Greenhouse & Hydroponic Risk Demo

Enter a tomato sensor reading from a substrate/soil greenhouse or a hydroponic system and get back risk labels, safe next checks, and any blocked actions -- exactly what the Pomona platform computes today at POST /v1/reasoners/tomato-risk.

Advisory only

This demo never controls irrigation, dosing, or any other hardware. It never issues a pesticide or fertigation command and never gives a definitive disease diagnosis -- those are always blocked actions that require a human to review and decide. A deterministic safety checker is the final authority in the full platform; this demo shows that same guarded logic in isolation.

Local LoRA inference is not yet wired into the Pomona platform runtime, so this demo currently runs the same deterministic rules path the platform runs in production today -- not a simplified stand-in.

Try three preset scenarios

• —Routine reading (substrate greenhouse) -- normal ranges, no risk
• —Hydroponic nutrient / pH risk -- high pH and EC in a recirculating system, triggers a blocked fertigation action
• —Missing / suspect data -- critical sensor fields absent, triggers a data-quality flag

Links

• —Platform: github.com/Okyanus/pomona
• —Model: pomona-tomato-risk-reasoner-v0.1.7-lora
• —Dataset: greenhouse-sensor-data
• —Collection: Pomona -- Local AI for Safer Greenhouse Decision Support