FRONTEND LAYER

USER INTERFACE

Flutter / Kotlin Mobile App

• Voice/Chat Interface (Local Language)

• Image Upload (Disease Detection)

• Crop Recommendations View

• Offline Mode (SQLite Storage)

Mocal or Venle

• Folitri starting

• Speech-to-Text bite

• Text-to-Speech

BACKEND LAYER

(API & DECISION LOGIC)

FastAPI — Core integration layer

• API Gateway for frontend communication

• Decision Engine → Combines ML results

• User Feedback Handler

• Database Access Layer

🗄️ PostgreSQL + PostGIS (spatial data)

Docker

GitHub Actions (CI/CD)

DATA & AI LAYER

(MODEL + DATA HANDLING)

Data Sources

• Soil Data (SoliGrids, Bhuvan)

• Weather Data (IMD, OpenWeather)

• Market Data (AGMiRKNET)

• Image Data (Farmer camera)

Preprocessing

Cleaning, merging, geotagging

ML Models

• XGBoost – Soil parameter prediction

• Random Forest - Crop recommendation

• CNN (TFLite) – Disease detection

• LSTM/ARIMA – Price forecasting

Step 1: Data Collection

Collects multi-source data from:

• Soil Data → SoilGrids, Bhuvan

• Weather Data → IMD, OpenWeather

• Market Data → AGMARKNET

• Image Data → Farmer’s mobile (plant disease)

• IoT Sensors (real-time soil moisture, NPK)

🛠 Tools: Python, REST APIs, IoT sensors

👥 Team: Data Engineer + IoT

Step 2: Data Fusion & Storage

Clean, merge, and geotag data

Store spatial data using PostgreSQL + PostGIS

🛠 Tools: Python (Pandas, GeoPandas), PostgreSQL + PostGIS

👥 Team: Data Science

Step 3: AI & ML Models

• XGBoost → Soil parameter prediction (pH, NPK, Moisture)

• Random Forest → Crop recommendation

• CNN (TFLite) → Disease detection

• LSTM/ARIMA → Market price forecasting

🛠 Tools: TensorFlow, Scikit-learn, TFLite

👥 Team: AI/ML

Step 4: Decision Engine

Combine soil, weather & market results

Rank crops based on profitability & sustainability

🛠 Tools: FastAPI backend, PostgreSQL

👥 Team: Backend + AI

Step 5: User Interface

• Mobile app with multilingual voice & chat support

• Offline mode via SQLite

• Image upload for disease detection

🛠 Tools: Flutter / Kotlin, FastAPI, SQLite

👥 Team: App Dev + UI/UX

Step 6: Deployment & Feedback Loop

Deploy using Docker + GitHub CI/CD

Collect farmer feedback → retrain models

🛠 Tools: Docker, GitHub Actions

👥 Team: DevOps + QA

Overall Data Flow & Goal

Soil / Weather / Market / Image Data → Data Fusion & Preprocessing → AI Models → Decision Engine → Mobile App → Feedback → Retraining

🎯 Goal: Provide farmers with real-time, localized, profitable, and sustainable crop recommendations.