JalDrishti (जलदृष्टि)
AI-Powered Precision Agronomy & Hydrological Irrigation Engine
JalDrishti (जलदृष्टि) is a climate-smart agronomy and precision irrigation advisory platform engineered specifically for Indian smallholder agriculture. It combines FAO-56 Penman-Monteith Evapotranspiration modeling, persistent daily root-zone water balance tracking, ISRIC SoilGrids pedotransfer physics, real-time Open-Meteo satellite meteorology, and a 100% Groq-native multilingual AI assistant (JalSathi AI). Operating without physical in-situ sensors, JalDrishti provides smallholders with scientific decision support to optimize water usage, eliminate diesel and electricity pump waste, mitigate microclimate disease outbreaks, and track verified cumulative financial savings (ROI). Following an audited 6-phase engineering overhaul in September 2026, the platform features a 9-table PostgreSQL schema, asynchronous Redis caching, a 44-test Pytest suite, and a cross-platform Flutter client.
Decoupled Microservice Topology with Asynchronous FastAPI, Supabase PostgreSQL, Redis Cache, and Groq-Native RAG Cascade
Calculates ETo via Stefan-Boltzmann net radiation and logarithmic 10m-to-2m wind reduction without in-situ soil probes.
Suppresses irrigation runs when forecast precipitation (≥3mm/24h or ≥5mm/48h) is detected, preventing waterlogging.
Groq-native translation converts Bengali and Hindi farmer queries into agronomic terms for dense ChromaDB retrieval.
Strict ICAR Package of Practices grounding with automatic KVK fallback replacing ungrounded agrochemical advice.
Sentence-Transformers all-MiniLM-L6-v2 running on-device for memory-efficient semantic search.
Pytest test suite covering hydrology math, soil depletion state persistence, security contracts, and API routes.
Engineering Highlights
- Rigorous FAO-56 Penman-Monteith Evapotranspiration and root-zone water balance modeling without physical sensors
- Smart Rain Hold protection saving ₹150–₹500 in fuel/power per avoided irrigation run
- JalSathi AI: 100% Groq-native multilingual RAG voice copilot with Bengali and Hindi speech support
- Weather-driven pest & disease early warning system with ICAR-grounded IPM chemical and biological recommendations
- Audited 9-table PostgreSQL relational schema with persistent SoilDepletionState and 44/44 passing Pytest suite