---

🌱 AgroMind Backend — AI Crop Recommendation, Disease Detection & IoT Farm Intelligence

AgroMind backend is a full-stack FastAPI-powered engine that delivers AI crop recommendations, plant disease detection, real-time IoT soil telemetry, market insights, and an LLM-powered advisory system for farmers. It integrates weather, soil, and market APIs with device-level IoT data and advanced ML models to provide actionable agricultural insights.

---

📌 Features

🌾 AI Crop Recommendation

• Uses soil chemistry, temperature, rainfall, humidity, and market price signals.
• Supports future integration with soil sensors & weather IoT nodes.

🍃 Plant Disease Detection

• Upload leaf images or feed IoT camera inputs.
• Runs ML inference and returns disease + cure recommendations.

📡 IoT Integration

• Device registration, provisioning, and secure token-based access.
• Telemetry ingestion for soil moisture, pH, temperature, etc.
• IoT image ingestion for edge disease detection.

🌦️ External API Integrations

• OpenWeather API for weather data
• SoilGrids API for soil nutrient and texture information
• Market API (mock or real) for commodity pricing data

🤖 LLM Advisory Agent

• Farmers ask questions via text or voice.
• Backend uses an LLM + retrieval pipeline to give precise answers.

🎙️ Voice Assistant

• STT → LLM → TTS pipeline
• Returns both text and audio response.

🛠️ Modular, Extendable Architecture

• Clear separation: routes, services, ML logic, database, integrations
• Ready for Docker, CI/CD, and cloud deployment
• IoT-ready MQTT support (optional)

---

📁 Project Structure

backend/
  app/
    api/
      routes/
        user.py
        auth.py
        crop_recommendation.py
        disease.py
        devices.py
        iot_webhook.py
        feedback.py
        llm_agent.py
        voice.py
        integrations.py
      router.py
    core/
      config.py
      logger.py
      security.py
      middleware.py
    integrations/
      weather_client.py
      soil_client.py
      market_client.py
    services/
      user_service.py
      token_service.py
      storage.py
      utils.py
      iot_service.py
    ml_services/
      crop_service.py
      image_service.py
    models/
      db_models.py
      schemas.py
    workers/
      celery_app.py
      tasks.py
    main.py

---

⚙️ Setup & Installation

1. Clone the repository

git clone https://github.com/<your-username>/agromind-backend.git
cd agromind/backend

2. Create virtual environment

python -m venv venv
source venv/bin/activate   # Windows: venv\Scripts\activate

3. Install dependencies

pip install -r requirements.txt

---

🔐 Environment Variables

Create a .env file using .env.example:

SECRET_KEY=<generated_key>
DATABASE_URL=postgresql://postgres:password@localhost:5432/agromind
OPENWEATHER_API_KEY=<your_key>
SOIL_API_KEY=
MARKET_API_URL=
MARKET_API_KEY=
REDIS_URL=redis://localhost:6379/0

Generate a secure key:

openssl rand -hex 32

---

🏃 Running the Backend

Start FastAPI

uvicorn app.main:app --reload

API docs available at:

http://localhost:8000/docs
http://localhost:8000/redoc

---

🧪 Testing the API

Check health

GET /health/

Predict crop

POST /crop/recommend
{
  "soil_type": "loamy",
  "temperature": 28,
  "humidity": 60,
  "rainfall": 140
}

Detect disease

Upload image using:

POST /disease/predict (multipart/form-data)

IoT telemetry ingestion

POST /iot/telemetry
{
  "device_id": "device123",
  "temperature": 28,
  "moisture": 18,
  "ph": 6.2
}

---

🔌 Integrations

Weather

integrations/weather_client.py

Functions:

• get_current_weather(lat, lon)
• get_forecast(lat, lon, days)
• get_historical(lat, lon, start_unix)

Soil

integrations/soil_client.py

Functions:

• get_soil_properties(lat, lon)

Market

integrations/market_client.py

Functions:

• get_latest_price(commodity, region)
• get_price_history(commodity, region)

---

🤖 LLM Agent

API:

POST /agent/query
{
  "user_id": 1,
  "text": "What should I do if my crop leaves turn yellow?"
}

Uses retrieval + LLM pipeline to generate exact farming advice.

---

🎙️ Voice Assistant

POST /voice/ask
audio_file + user_id

Pipeline:

1. Speech-to-Text
2. LLM advisory
3. Text-to-Speech
4. Returns audio + transcript

---

🗄️ Database

Uses SQLAlchemy ORM.

Tables include:

• Users
• Devices
• Telemetry
• Feedback
• Images
• TokenBlocklist (for refresh-token revocation)

Supports PostgreSQL, SQLite (dev), MySQL.

---

🧰 Services Provided

• user_service → user CRUD & auth
• token_service → JWT + refresh token system
• storage → local/S3 file upload handling
• iot_service → telemetry + device images
• ml_services → crop & disease prediction models
• utils → image validation, job utils

---

🧱 Future Enhancements

• ONNX/TFLite edge deployment
• Full MQTT ingestion pipeline
• Retraining automation via Celery
• Role-based access control
• Region-language support for LLM answers
• Offline-first IoT sync

---

🤝 Contributing

Pull requests are welcome. Open an issue for feature requests or discussions.

---

📜 License

MIT License