AgniSutra

Smart India Hackathon Submission Project

AgniSutra is an AI-powered smart agriculture platform that helps farmers improve crop decisions using IoT sensor data, satellite insights, disease detection, weather context, and multilingual advisory support.

Problem We Address

Farmers often make critical crop decisions with delayed, fragmented, or unavailable field intelligence. AgniSutra unifies real-time farm data and AI-driven recommendations to support timely, data-backed actions.

Core Features

• IoT-based field monitoring (moisture, temperature, humidity, NPK)
• Real-time alerting over WebSockets for critical field conditions
• AI disease detection from crop images (CNN + GPT fallback + RAG-based advisory)
• Yield and fertilizer recommendation workflows
• Krishi Saathi assistant for contextual advisory
• Satellite/NDVI integration for crop-health-aware insights
• Authentication and farmer profile management
• Offline-first support in mobile app (Hive caching)
• Multilingual interface support

High-Level Architecture

• Backend: FastAPI + SQLAlchemy + PostgreSQL
• Web App: Next.js (React)
• Mobile App: Flutter
• AI/ML Stack: TensorFlow/Keras, CatBoost/Scikit-learn, LangChain, FAISS, OpenAI integration
• Data Inputs: IoT device payloads, satellite signals, user-entered agronomic data

Repository Structure

.
├── backend/                # FastAPI APIs, DB models, ML/LLM integration
├── frontend/               # Next.js web interface
├── mobile/                 # Flutter mobile application
├── API_DOCUMENTATION.md    # API endpoint reference
├── DB_SCHEMA.md            # Database schema overview
├── INTEGRATION_GUIDE.md    # Frontend/IoT integration guide
├── SETUP.md                # Detailed setup instructions
└── judge_evidence/         # Benchmark and model evidence artifacts

Quick Start

1) Backend

cd backend
python -m venv venv
source venv/bin/activate   # Windows: venv\Scripts\activate
pip install -r requirements.txt
python -m uvicorn app.main:app --reload --host 0.0.0.0 --port 8000

Backend docs: http://localhost:8000/docs

2) Web Frontend

cd frontend
pnpm install
pnpm dev

Web app runs at: http://localhost:3000

3) Mobile App

cd mobile
flutter pub get
flutter run

Set backend URL in mobile/lib/constants.dart to your local machine IP for physical devices.

Evidence for Evaluation

Model/training evidence is available in:

• judge_evidence/1_learning_curve.png
• judge_evidence/3_error_dist.png
• judge_evidence/4_benchmark.png
• judge_evidence/5_shap_explanation.png
• judge_evidence/final_training_data.csv

Documentation Index

• Setup Guide: SETUP.md
• Integration Guide: INTEGRATION_GUIDE.md
• API Documentation: API_DOCUMENTATION.md
• Database Schema: DB_SCHEMA.md

Submission Note

This repository represents the working prototype and integration baseline prepared for Smart India Hackathon evaluation.