Agri-AI Advisory System 🌾

An intelligent agricultural advisory platform designed to bridge the gap between farmers and verified agricultural experts. The system leverages AI-powered crop recommendations, CNN-based image-to-soil detection, real-time location mapping, a multilingual generative voice chatbot, and real-time Socket.io-driven video/audio communication.

🏗️ System Architecture

The application comprises three core components:

React Frontend (Vite): An interactive, multilingual, and highly responsive user interface.
Node.js Express Server: The central logic handler, managing user roles (Farmer, Expert, Admin), live signaling, and community forums.
Python Flask ML Service: A microservice dedicated to running TensorFlow models for crop recommendations and soil classifications.

graph TD
    classDef client fill:#e1f5fe,stroke:#039be5,stroke-width:2px;
    classDef node fill:#efebe9,stroke:#5d4037,stroke-width:2px;
    classDef python fill:#e8f5e9,stroke:#2e7d32,stroke-width:2px;
    classDef db fill:#fbe9e7,stroke:#d84315,stroke-width:2px;
    classDef external fill:#fff8e1,stroke:#f57f17,stroke-width:2px;

    %% Nodes
    React[React Frontend<br/>Vite / Socket.io Client]:::client
    Express[Express Backend<br/>Node.js - Port 5000]:::node
    Flask[ML Microservice<br/>Python Flask - Port 5001]:::python
    Mongo[(MongoDB Database)]:::db
    
    %% External API Nodes
    Gemini[Google Gemini API]:::external
    TTS[Google TTS & Translate]:::external
    OSM[OpenStreetMap Nominatim]:::external
    HF[HuggingFace API]:::external
    Reddit[Reddit API]:::external

    %% Connections
    React <-->|Socket.io / HTTP| Express
    React -->|navigator.geolocation| OSM
    Express <-->|Mongoose| Mongo
    Express <-->|Axios Proxy| Flask
    Express <-->|Google GenAI SDK| Gemini
    Express <-->|Axios Proxy| TTS
    Express <-->|Inference SDK| HF
    Express <-->|Axios| Reddit

    %% Sub-features inside Flask
    subgraph Python Flask Services
        Flask -->|Predict Crop| CropModel[crop_model.h5<br/>TensorFlow Neural Network]
        Flask -->|Classify Soil| SoilModel[soil_model_cnn.h5<br/>ResNet50 CNN]
    end

    %% Sub-features inside Node
    subgraph Express Services
        Express -->|Chat requests / signalling| SocketIO[Socket.io Signaling Server]
        SocketIO <-->|WebRTC Video Calls| React
    end

🌟 Core Features & Technical Implementations

1. Smart Crop Predictor

Feature: Allows farmers to input soil and climate parameters (Nitrogen, Phosphorus, Potassium, pH, humidity, temperature, rainfall) to find the most suitable crop to plant.
Implementation: The parameters are sent to the Python Flask microservice, scaled using a pre-saved scaler.pkl, and classified using a trained neural network (crop_model.h5). The result is mapped back using label_encoder.pkl and returned to the frontend.

2. CNN Image-to-Soil Classification

Feature: Farmers can upload an image of their soil to immediately classify its type (Red, Black, Alluvial, Laterite) and obtain a plausible pH level.
Implementation: Uses a ResNet50 Convolutional Neural Network architecture built inside train_soil_cnn.py. The Flask service accepts the image at POST /predict-soil, resizes it to $224 \times 224$, runs it through the preprocessor, and feeds it into the CNN model soil_model_cnn.h5.

3. Geolocation & Area Mapping

Feature: Automatically maps the farmer's location during registration/profile updates to deduce localized weather and regional soil patterns.
Implementation: Uses the native HTML5 Web Geolocation API (navigator.geolocation) to extract { latitude, longitude } coordinates. The frontend resolves these coordinates via the OpenStreetMap Nominatim reverse geocoding API to identify local divisions (suburb, village, city, county).

4. Multilingual Generative Voice Chatbot

Feature: Voice-enabled AI chatbot allowing farmers to speak queries in English, Hindi, or Telugu and receive tailored responses from distinct AI expert "personas".
Implementation: Utilizes the browser-based Web Speech API for live transcription. The text is sent to the Node.js backend which coordinates requests to the Google Gemini API (gemini-2.5-flash) using customized prompts tailored for Weather, Fertilizer, or Disease experts. The response is spoken back via a custom node proxy endpoint /api/tts querying Google Translate's TTS stream.

5. Open Knowledge Community Forum

Feature: A collaborative forum where farmers post agricultural questions and verified experts post helpful answers.
Implementation: Managed by Mongoose Schemas (Post and Answer). It supports asynchronous comments, chronologically ordered feeds, and user-to-expert linkages.

6. Synchronous Expert Live Conferencing & Signaling

Feature: Farmers can request live consultations, message chat, or launch WebRTC video/audio sessions with verified experts.
Implementation: Built with Socket.io over HTTP. Standard chat messages are cached to a MongoDB collection (Message) for persistence. Audio/video calls use a custom WebRTC signaling protocol: when a farmer initiates a call, Socket.io manages call routing, offering, answering, and ICE candidate exchange. An expert approval mechanism restricts excessive calls through the database collection CallRecord.

7. Farmer-to-Expert Upgrade Pipeline (Admin Panel)

Feature: A workflow where experienced farmers can apply to become verified experts, reviewed and approved by administrators.
Implementation: Farmers submit applications from their profile dashboard. Admins can view pending applications, audit credentials, and trigger a PUT request to update their role to expert, automatically instantiating their Expert details in the database.

📁 Repository Structure

agri-ai-advisory-system/
├── backend/                  # Node.js Express server
│   ├── agents/               # AI Agent prompts & instructions
│   ├── config/               # Database and API setups
│   ├── controllers/          # Request handler functions
│   ├── models/               # MongoDB Mongoose Schemas (User, Post, Answer, Expert, etc.)
│   ├── routes/               # Express endpoints (auth, crops, admin, etc.)
│   ├── services/             # Third-party integrations (Gemini, HuggingFace)
│   ├── uploads/              # Temp storage for soil image classification
│   ├── utils/                # Helper files
│   ├── server.js             # Main server entrypoint
│   └── .env                  # Server environment configuration
│
├── frontend/                 # React client application (Vite)
│   ├── src/
│   │   ├── api/              # API caller functions (axios)
│   │   ├── assets/           # Global assets & styling sheet
│   │   ├── components/       # Reusable components (Navbar, Chat, etc.)
│   │   ├── pages/            # Application pages (Forum, Market, Advisor, etc.)
│   │   ├── App.jsx           # Main routing & app composition
│   │   └── main.jsx          # React entrypoint
│   ├── index.html
│   └── vite.config.js
│
├── ml-service/               # Python ML Flask microservice
│   ├── Crop_recommendation.csv
│   ├── app.py                # Flask entry point (Port 5001)
│   ├── crop_model.h5         # Crop classification weights
│   ├── soil_model_cnn.h5     # Soil ResNet50 CNN model weights
│   ├── train_model.py        # Model training script for Crops
│   ├── train_soil_cnn.py     # ResNet50 soil classifier trainer
│   └── requirements.txt      # Python dependencies
│
├── README.md                 # Project README
└── package.json              # Root configurations

🚀 Setup & Installation

1. Database Setup (MongoDB)

Set up a MongoDB cluster (local or Atlas) and copy the URI.
Note: The database seeds three default administrator accounts upon initial server startup:
- Admins: admin1@harvestmate.com, admin2@harvestmate.com, or admin3@harvestmate.com
- Password: admin123

2. Python ML Microservice Setup

Navigate to the machine learning service directory:

cd ml-service

Create a virtual environment and activate it:

# Windows
python -m venv venv
venv\Scripts\activate

# macOS/Linux
python3 -m venv venv
source venv/bin/activate

Install requirements:

pip install -r requirements.txt

Start the ML service (runs on http://localhost:5001):

python app.py

3. Backend Express Server Setup

Open a new terminal tab, navigate to the server directory:

cd server

Install dependencies:

npm install

Create a .env file in the server directory:

MONGO_URI=your_mongodb_connection_string
GEMINI_API_KEY=your_google_gemini_api_key
HF_TOKEN=your_huggingface_token
WEATHER_API_KEY=your_openweathermap_api_key
OPENAI_API_KEY=your_openai_api_key
PORT=5000

Start the server:

npm start

The server will boot on http://localhost:5000 and establish a MongoDB connection.

4. Frontend React Client Setup

Open a new terminal tab, navigate to the frontend directory:

cd frontend

Install packages:

npm install

Start the dev server:

npm run dev

Open your browser and visit http://localhost:5173.

🔌 API Documentation (Key Endpoints)

backend Express API

Method	Endpoint	Description
`POST`	`/auth/register`	Register new Farmers or Experts
`POST`	`/auth/login`	Login and acquire JWT
`GET`	`/experts`	Retrieve all verified experts
`POST`	`/request-chat`	Farmer registers a private chat request with an expert
`GET`	`/api/tts`	Proxy to play back generated spoken audio (bypassing CORS)
`POST`	`/api/detect-soil`	Multipart request that proxies soil image to Python Flask API
`GET`	`/api/daily-tip/:farmerId`	Generate a personalized daily tip using Gemini API
`GET`	`/api/agri-news`	Pull live hot topics from agriculture subreddits

Python Flask ML API

Method	Endpoint	Description
`POST`	`/predict`	Accept soil statistics (N, P, K, etc.) and predict optimal crop
`POST`	`/predict-soil`	Accept soil image stream and perform classification

🤝 Contributing

Contributions are highly encouraged! Please make sure to test your code locally, update configurations accordingly, and submit a pull request with clear description details.

📄 License

This project is licensed under the MIT License.

Name		Name	Last commit message	Last commit date
frontend		frontend
ml-service		ml-service
server		server
test		test
train_data		train_data
.gitignore		.gitignore
Feature_List.txt		Feature_List.txt
README.md		README.md
package-lock.json		package-lock.json
package.json		package.json
port_update.js		port_update.js

Navigation Menu

Uh oh!

Repository files navigation

Agri-AI Advisory System 🌾

🏗️ System Architecture

🌟 Core Features & Technical Implementations

1. Smart Crop Predictor

2. CNN Image-to-Soil Classification

3. Geolocation & Area Mapping

4. Multilingual Generative Voice Chatbot

5. Open Knowledge Community Forum

6. Synchronous Expert Live Conferencing & Signaling

7. Farmer-to-Expert Upgrade Pipeline (Admin Panel)

📁 Repository Structure

🚀 Setup & Installation

1. Database Setup (MongoDB)

2. Python ML Microservice Setup

3. Backend Express Server Setup

4. Frontend React Client Setup

🔌 API Documentation (Key Endpoints)

backend Express API

Python Flask ML API

🤝 Contributing

📄 License

About

Releases

Packages

Contributors

Languages

Uh oh!

Latest commit

History

Folders and files

Repository files navigation

Agri-AI Advisory System 🌾

🏗️ System Architecture

🌟 Core Features & Technical Implementations

1. Smart Crop Predictor

2. CNN Image-to-Soil Classification

3. Geolocation & Area Mapping

4. Multilingual Generative Voice Chatbot

5. Open Knowledge Community Forum

6. Synchronous Expert Live Conferencing & Signaling

7. Farmer-to-Expert Upgrade Pipeline (Admin Panel)

📁 Repository Structure

🚀 Setup & Installation

1. Database Setup (MongoDB)

2. Python ML Microservice Setup

3. Backend Express Server Setup

4. Frontend React Client Setup

🔌 API Documentation (Key Endpoints)

backend Express API

Python Flask ML API

🤝 Contributing

📄 License

About

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages