Krishi Sakhi

Krishi Sakhi is a cross-platform solution for smart agriculture, featuring a Flutter mobile app and a Flask backend powered by machine learning models. The system provides crop recommendations, yield predictions, plant disease detection, and pest identification to assist farmers in making informed decisions.

Project Structure

Krishi Sakhi/
├── Flask Backend - ML Models/
│   ├── requirements.txt
│   ├── src/
│   │   ├── app.py
│   │   └── models/
│   │       └── ... (ML models and assets)
│   └── tests/
│       └── plant_disease_prediction_test_script.py
├── Flutter App/
│   ├── pubspec.yaml
│   ├── .env
│   ├── lib/
│   │   ├── main.dart
│   │   ├── l10n/                # Localization
│   │   ├── models/              # Data models
│   │   ├── providers/           # State management (provider)
│   │   ├── screens/             # UI screens (home, onboarding, crop recommendation, etc.)
│   │   ├── services/            # API and logic services
│   │   ├── utils/               # Utilities (env_keys, constants, theme)
│   │   └── widgets/             # Reusable widgets
│   └── ... (platform folders: android, ios, web, etc.)
└── README.md

Features

• Crop Recommendation: Suggests optimal crops based on soil nutrients, weather, and rainfall using ML models and fallback logic.
• Yield Prediction: Estimates expected crop yield from user and weather data.
• Plant Disease Detection: Identifies plant diseases from leaf images using deep learning.
• Pest Detection: Detects and classifies pests from images.
• Fertilizer Recommendation: Recommends fertilizers based on soil and crop data.
• Weather Integration: Fetches real-time weather data for location-aware recommendations.
• Schemes & Loans: Provides information on government schemes and agricultural loans.
• Voice Assistant: Voice-based interaction for hands-free usage.
• Chatbot: In-app chatbot for agricultural queries and support.
• Multi-language Support: Localized UI for multiple Indian languages.

Getting Started

1. Backend (Flask + ML Models)

Features

• Crop Recommendation: Predicts optimal crops using a trained ML model.
• Yield Prediction: Estimates crop yield based on area, production, rainfall, fertilizer, crop, season, and state.
• Plant Disease Detection: Classifies plant diseases from images using a PyTorch model and Gemini AI for enhanced accuracy.
• Pest Detection: Identifies pests from images using a deep learning model.
• Fertilizer Recommendation: Suggests fertilizers based on soil and crop type.
• API Endpoints: RESTful endpoints for all predictions, with JSON and image support.
• Environment Variables: Loads sensitive keys (e.g., Gemini API key) from .env.

Structure

Flask Backend - ML Models/
├── requirements.txt         # Python dependencies
├── .env                    # Environment variables (not committed)
├── src/
│   ├── app.py              # Main Flask app and API endpoints
│   └── models/             # ML models and assets (joblib, pth files)
└── tests/
    └── plant_disease_prediction_test_script.py

Setup

1. Navigate to the backend folder:

   cd "Flask Backend - ML Models"

2. Create a Python virtual environment:

   python -m venv venv

3. Activate the environment:
   • Windows:

     .\venv\Scripts\activate

   • macOS/Linux:

     source venv/bin/activate

4. Install dependencies:

   pip install -r requirements.txt

5. Create a .env file in the backend root. Example:

   GEMINI_API_KEY=your_api_key_here
   FLASK_ENV=development
   SECRET_KEY=your_secret_key
   # Add other keys as needed
   

6. Run the Flask server:

   python src/app.py

2. Frontend (Flutter App)

Setup

1. Navigate to the Flutter app folder:

   cd "Flutter App"

2. Get dependencies:

   flutter pub get

3. Create an environment file for API endpoints and keys:
   • Create a .env file in the Flutter App directory. Example:

     # Flutter App/.env
     API_BASE_URL=http://127.0.0.1:5000
     OPENWEATHER_API_KEY=your_openweather_key
     GEMINI_API_KEY=your_gemini_key
     # Add other keys as needed
     

   • The app uses flutter_dotenv to load environment variables at startup. Keys are accessed in code via the EnvKeys class in lib/utils/env_keys.dart:

     import 'package:flutter_dotenv/flutter_dotenv.dart';
     class EnvKeys {
       static String get openWeatherApiKey => dotenv.env['OPENWEATHER_API_KEY'] ?? '';
       static String get geminiApiKey => dotenv.env['GEMINI_API_KEY'] ?? '';
       // ...
     }

   • Update the .env file as needed for your deployment.
4. Run the app:

   flutter run

App Structure Details

• State Management: Uses provider for app-wide state (authentication, language, crop recommendation, weather, etc.).
• API Services: All backend and ML model communication is handled in lib/services/ (e.g., crop_api_service.dart).
• Screens: Each feature (crop recommendation, disease detection, etc.) has its own screen in lib/screens/.
• Localization: Multi-language support via flutter_localizations and custom AppLocalizations.
• Environment Variables: All sensitive keys and endpoints are loaded from .env and never hardcoded.
• UI: Custom themes and reusable widgets for a consistent look.

Notes

• Ensure both backend and frontend .env files are not committed to version control.
• Update API URLs in the Flutter .env file as per your backend deployment.

Contributing

Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.

License

MIT