Agri-AI

Agri-AI is a Flask-based backend API that provides an agriculture assistant chatbot powered by Google Gemini, along with crop information and sample order-tracking endpoints.

It is designed for easy integration with mobile or web clients and includes:

• Conversational farming guidance
• Intent-aware responses (crop info, pest/disease, fertilizers, yield, order tracking)
• Local agriculture dataset support (crops + orders)
• REST endpoints for chat, search, statistics, and utility features

Tech Stack

• Python 3.x
• Flask + Flask-CORS
• Google Generative AI (Gemini)
• python-dotenv for environment configuration

Features

• Chat endpoint with per-user conversation history
• Lightweight intent detection to improve response relevance
• Knowledge-base augmentation for common agriculture topics
• Data-driven context injection from local JSON files:
  • Crop details (season, soil, climate, cultivation, pests, diseases)
  • Order details (status tracking by order ID and user)
• Health check and API statistics endpoints
• CORS enabled for client integration

Project Structure

Agri-AI/
├── app.py                      # Flask app factory and server startup
├── config.py                   # Environment + model configuration
├── check_models.py             # Utility script to list available Gemini models
├── requirements.txt            # Python dependencies
├── data/
│   ├── crops.json              # Crop dataset
│   ├── orders.json             # Orders dataset
│   └── products.csv            # Product reference data
├── models/
│   └── agriculture_data.py     # Data loading/search/format helpers
├── routes/
│   └── chat_routes.py          # REST API endpoints
└── services/
    ├── chatbot_service.py      # Chat orchestration + Gemini integration
    └── knowledge_base.py       # Built-in agriculture knowledge snippets

API Overview

Base URL (local): http://localhost:5000

Root endpoint:

• GET / - API metadata and endpoint summary

Chat and session endpoints:

• POST /api/chat
• POST /api/clear-history
• GET /api/conversation-history/<user_id>

Crop endpoints:

• GET /api/crops
• GET /api/crop/<crop_name>
• GET /api/search-crops?q=<query>

Order endpoints:

• GET /api/order/<order_id>
• GET /api/user-orders/<user_id>

Utility endpoints:

• GET /api/health
• GET /api/quick-questions
• GET /api/statistics

Setup and Run

1. Clone and move into the project

git clone <your-repo-url>
cd Agri-AI

2. Create and activate a virtual environment

python3 -m venv .venv
source .venv/bin/activate

3. Install dependencies

pip install -r requirements.txt

4. Configure environment variables

Create a .env file in the project root:

GOOGLE_API_KEY=your_google_ai_api_key
SECRET_KEY=replace_with_secure_value
FLASK_DEBUG=True
PORT=5000

Notes:

• GOOGLE_API_KEY is required for chatbot responses.
• OPENAI_API_KEY appears in configuration but is not required by the current Gemini-based runtime flow.

5. Start the server

python app.py

Server will start on:

• http://localhost:5000

Health check:

• http://localhost:5000/api/health

Request/Response Examples

Chat

curl -X POST http://localhost:5000/api/chat \
  -H "Content-Type: application/json" \
  -d '{
    "user_id": "USER123",
    "message": "How to grow rice?"
  }'

Sample response:

{
  "success": true,
  "response": "...",
  "intent": "crop_info"
}

Track an Order

curl http://localhost:5000/api/order/ORD001

Search Crops

curl "http://localhost:5000/api/search-crops?q=kharif"

Data Included

Bundled sample datasets include:

• 5 crops in data/crops.json
• 3 orders in data/orders.json
• Product reference rows in data/products.csv

These can be expanded to fit your local market and crop conditions.

Configuration Notes

Main configuration lives in config.py:

• Model name: models/gemini-pro-latest
• Temperature: 0.7
• Max tokens: 500
• API server port from PORT environment variable

You can verify available Gemini models with:

python check_models.py

Error Handling

The API returns JSON error objects with success: false and an error message for:

• Missing or empty request input
• Resource not found (e.g., unknown crop/order)
• Internal runtime exceptions

Global handlers exist for:

• 404 endpoint not found
• 500 internal server error

Limitations and Future Improvements

• Conversation history is in-memory (not persistent).
• No authentication/authorization layer yet.
• No test suite currently included.
• products.csv is not currently used directly by API logic.

Recommended next steps:

• Add persistent storage (SQLite/PostgreSQL/Redis)
• Add automated tests (unit + API integration)
• Add auth and rate limiting for production
• Add logging and observability

License

Add your preferred license here (for example, MIT).