Cropia: AI-Powered Early Crop Health Detection System

Cropia is an intelligent agricultural platform designed to bridge the gap between farmers and administrative bodies. It empowers farmers with real-time, offline-first disease detection and decision support, while providing government officials and NGOs with a "God View" of regional crop health to prevent outbreaks.

Built as a high-performance monorepo using Turborepo, Bun, and Hono, Cropia leverages Google's Gemini 2.5 Flash Lite for speed and accuracy in low-resource environments.

---

🌐 Live Demos

• Landing Page: https://cropia-phi.vercel.app/
• Admin Dashboard (Government/NGO): https://cropia-admin.vercel.app
• Farmer App: https://cropia-farmer.vercel.app

---

🚀 Key Features

👨‍🌾 For Farmers (The Field App)

• AI Crop Scanner ("The Gatekeeper"):

• Instantly validates if an image is a crop (filtering out selfies/documents).

• Detects diseases, pests, and nutrient deficiencies in real-time.

• Provides a "Confidence Score" and "Visual Severity" rating (Healthy vs. Warning vs. Critical).

• Decision Support Dashboard:

• Smart Spray Planner: A 12-hour timeline analyzing Wind Speed (<15km/h) and Rain Probability (<30%) to recommend safe spraying windows.

• Root Health X-Ray: Visualizes soil moisture at two depths (Surface 0-10cm vs. Deep 10-30cm) to detect "false dryness" and prevent over-irrigation.

• Water Balance Calculator: Tracks net water deficit (Rain - Evapotranspiration) over the last 3 days.

• Cropia Assistant: A conversational AI chatbot that can answer questions about recent scans, weather forecasts, and agronomy advice.

🏢 For Admin/Government ("The Watchtower")

• The God View (Live Map):

• Interactive geospatial map visualizing every crop scan in the jurisdiction.

• Color-coded markers (Green=Healthy, Red=Critical) for spotting regional outbreaks.

• Regional Analysis Agent:

• Aggregates data from thousands of scans to generate "News Headlines" (e.g., "Sugarcane Rust cases up 40% in Kolhapur").

• Provides statistical breakdowns of disease distribution per district/taluka.

• Jurisdiction Control: Granular RBAC allowing officials to view data specifically for their State, District, or Taluka.

🛠 Tech Stack

Core Infrastructure

• Monorepo: Turborepo
• Runtime/Manager: Bun (v1.2.20)
• Dependency Management:
  • Catalog: Centralized versions in root package.json.
  • Peer Dependencies: Preventing React duplication in shared UI packages.
• Language: TypeScript (v5.9)

Frontend (Web & PWA)

• Framework: React (Vite)
• Marketing Site: Astro (Static/Zero-JS)
• Styling: Tailwind CSS + Shadcn UI (@repo/ui)
• State/Routing: TanStack Query + TanStack Router
• Maps: Mapbox / Leaflet (via React)

Why Astro for Landing Page? We chose Astro over Next.js/React for the public site to ensure maximum performance:

• Zero JS Runtime: Ships pure HTML/CSS by default.
• No Overhead: Avoids the ~30kb React hydration cost for static pages.
• Islands Architecture: Interactive components (React) are only loaded when needed.

⚡ Proof of Concept: Perfect 100/100

To validate our choice, the Cropia landing page achieves a perfect score on Google PageSpeed Insights (Desktop).

Tip

View the Live Report: PageSpeed Insights - Cropia

Backend (API & Edge)

• Framework: Hono (Lightweight, Edge-ready)
• Communication: Hono RPC (Type-safe client-server contract)
• Database: MongoDB (via Prisma ORM)
• Auth: BetterAuth (Multi-tenant Organization support)

Artificial Intelligence

• Vision Model: Google Gemini 2.5 Flash Lite (Optimized for low latency).
• Framework: Vercel AI SDK (@ai-sdk/google).
• Agents:

1. Gatekeeper Agent: Validates image integrity and diagnoses specific crop issues.
2. Analysis Agent: Summarizes regional data into actionable natural language insights.

---

🏗 Architecture & Repository Structure

The project follows a modern monorepo structure:

.
├── apps
│   ├── farmer       # The Farmer PWA (Vite + React)
│   ├── admin        # The Government Dashboard (Vite + React)
│   ├── landing-page # Marketing Website (Astro)
│   └── api          # Hono Backend (Business Logic + AI Agents)
├── packages
│   ├── db           # Prisma Schema & MongoDB Client
│   ├── ui           # Shared Shadcn UI Components
│   ├── config       # Shared TS/Eslint Configs
│   └── scripts      # Utility scripts
└── turbo.json       # Build pipeline configuration

Dependency Graph

graph TD
    subgraph Apps
        F[Farmer App]
        A[Admin Dashboard]
        L[Landing Page]
        S[API Server]
    end

    subgraph Packages
        UI["@repo/ui"]
        DB["@repo/db"]
        API["@repo/api"]
        TS["@repo/typescript-config"]
    end

    F --> UI
    A --> UI
    L --> UI
    F --> API
    A --> API
    S --> API
    API --> DB
    
    %% Shared Configs
    UI -.-> TS
    DB -.-> TS
    API -.-> TS

Loading

The AI Pipeline

Cropia uses a Conditional Agentic Workflow:

1. Input: Image uploaded by farmer.
2. Gatekeeper (Gemini 2.5 Flash Lite):

• Check: Is this a plant? (Reject if false).
• Analyze: Identify Crop -> Detect Issue -> Assess Severity.
• Label: Generate semantic filename (e.g., sugarcane_rust_timestamp).

3. Persistence: Data stored in MongoDB with a "Location Snapshot" (State/District/Village frozen at upload time) for fast geospatial queries.
4. Async Analysis: The "Analysis Agent" runs periodically to aggregate these snapshots into regional reports for the Admin.

---

💾 Database Schema (Key Highlights)

• User/Auth: Uses BetterAuth with Organization and Member tables for managing Co-ops and Government hierarchies.
• Scan Model: Stores the imageUrl, aiMetadata (diagnosis, confidence), and a Location Snapshot (state, district, taluka) to avoid expensive joins during analytics.
• Jurisdiction Type: A custom composite type defining an admin's scope (e.g., { state: "Maharashtra", district: "Kolhapur" }).

---

🚀 Getting Started

Prerequisites

• Bun (v1.0+)
• MongoDB (Local or Atlas URL)
• Cloudinary Account (for image storage)
• Google Gemini API Key

Installation

1. Clone the repository:

git clone https://github.com/your-username/cropia.git
cd cropia

2. Install dependencies:

bun install

3. Environment Setup: Create a .env file in the root (see .env.example):

DATABASE_URL="mongodb+srv://..."
GOOGLE_GENERATIVE_AI_API_KEY="AIzaSy..."
CLOUDINARY_CLOUD_NAME="..."
BETTER_AUTH_SECRET="..."

4. Database Push:

bun run db:push

5. Run Development Server:

bun run dev

• Landing Page: http://localhost:4321
• Farmer App: http://localhost:5173
• Admin App: http://localhost:5174
• Backend API: http://localhost:3000

---

🤝 Contributing

1. Fork the repo.
2. Create your feature branch (git checkout -b feature/amazing-feature).
3. Commit your changes (git commit -m 'Add amazing feature').
4. Push to the branch (git push origin feature/amazing-feature).
5. Open a Pull Request.

---

📄 License

Distributed under the MIT License. See LICENSE for more information.