AgriBoost 🌱

AgriBoost is an AI-powered precision agriculture platform that analyzes soil health reports and farm-specific data to generate actionable, data-driven recommendations for improving crop yield, soil health, and long-term sustainability.

The system combines computer vision (for soil card analysis), large language models (Gemini), and an event-driven backend (Inngest) to provide farmers with scientifically grounded insights in a simple, accessible interface.

---

Key Features

• 📸 Soil Health Card Analysis (image upload + OCR/vision via Gemini)
• 📊 Personalized Crop Yield Prediction
• 🌾 Crop Recommendation Engine
• 🧪 Fertilizer & Nutrient Planning
• 🐛 Pest Management Strategy
• 💧 Irrigation Optimization
• 🌍 Climate-Specific Tips
• ♻️ Sustainability & Soil Improvement Plans
• 📄 PDF Report Generation
• 🔐 Secure Authentication with Clerk
• 🌐 Multi-language Support (next-intl)
• ⚡ Event-driven AI pipeline using Inngest

---

Architecture Overview

AgriBoost follows a modern event-driven, serverless-friendly architecture:

Frontend (Next.js App Router)
      |
      v
/api/results  --->  Inngest Event  --->  AI Pipeline (Gemini)
      |                    |
      |                    v
      |              ImageKit Upload
      |                    |
      v                    v
/api/inngest/status  <---  PostgreSQL (Drizzle ORM)

• Next.js 15 (App Router) for UI and API routes
• Clerk for authentication and session management
• Inngest for background job orchestration
• Google Gemini for vision + reasoning
• ImageKit for secure image storage
• PostgreSQL + Drizzle ORM for persistence

---

Tech Stack

• Framework: Next.js 15, React 18, TypeScript
• Styling: Tailwind CSS, shadcn/ui
• Auth: Clerk
• AI: Google Gemini (Vision + Text)
• Jobs / Workflow: Inngest
• Storage: ImageKit
• Database: PostgreSQL (Drizzle ORM)
• i18n: next-intl
• PDF Generation: Puppeteer

---

Local Setup

1. Clone the repository

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

2. Install dependencies

pnpm install

3. Environment Variables

Create a .env.local file:

# Auth
CLERK_SECRET_KEY=
NEXT_PUBLIC_CLERK_PUBLISHABLE_KEY=

# AI
GEMINI_API_KEY=

# Inngest
INNGEST_SIGNING_KEY=

# Database
DATABASE_URL=

# ImageKit
IMAGEKIT_PUBLIC_KEY=
IMAGEKIT_PRIVATE_KEY=
IMAGEKIT_URL_ENDPOINT=

4. Run services

pnpm dev:inngest

This runs:

• Next.js app
• Inngest Dev Server (event dashboard)

Open:

• App: http://localhost:3000
• Inngest UI: http://localhost:8288

---

AI Pipeline Flow

1. User uploads soil test image + farm details

2. /api/results triggers ai/generate-crop-yield event

3. Inngest executes steps:

   • Upload image to ImageKit
   • Call Gemini Vision + Text
   • Normalize structured JSON output
   • Store results in PostgreSQL

4. Frontend polls /api/inngest/status until completion

5. Results are cached and rendered

6. PDF report can be generated and downloaded

---

Folder Highlights

app/                → Next.js routes & UI
inngest/            → Background workflows
services/           → API clients (ImageKit, Inngest, etc.)
configs/            → DB & schema
components/ui/      → shadcn UI components
lib/                → Utilities & PDF templates

---

Security Model

• Browser authentication handled by Clerk
• Server-to-server communication secured via API keys
• Inngest control-plane never exposed to frontend
• Image uploads use signed URLs
• Dev-only API bypass available for Postman testing

---

Roadmap

• Real-time weather API integration
• Regional crop price prediction
• Satellite NDVI-based soil health scoring
• Offline-first mobile PWA
• Farmer-to-farmer knowledge graph

---

License

This project is licensed under the GNU GPL v3.0.

---

AgriBoost is built as a full-stack, production-grade AI system — not a demo — with proper orchestration, security boundaries, and data pipelines suitable for real-world agricultural decision support.