Rayeva AI Platform 🌿

Production-ready AI-powered sustainable commerce platform.

🚀 Live Deployment

• Frontend (Netlify): https://rayeva-ai-system.netlify.app
• Backend API (Render): https://rayeva-ai-system.onrender.com

---

Overview

Rayeva is a B2B/B2C sustainable commerce platform that uses OpenAI to automate product categorization, generate corporate proposals, and assist customers via WhatsApp. Built with a clean, layered architecture separating AI concerns from business logic.

---

Architecture

rayeva-ai/
├── backend/
│   ├── server.js              # Express entry point
│   ├── routes/
│   │   └── aiRoutes.js        # API route definitions
│   ├── controllers/
│   │   ├── categoryController.js   # Module 1 input validation + response
│   │   └── proposalController.js   # Module 2 input validation + response
│   ├── services/
│   │   ├── categoryService.js      # Module 1 business logic
│   │   └── proposalService.js      # Module 2 business logic
│   ├── ai/
│   │   ├── promptBuilder.js   # Centralized prompt factory (system + user)
│   │   ├── openaiService.js   # OpenAI API wrapper (JSON mode)
│   │   └── aiLogger.js        # Dual logger (Winston file + MongoDB)
│   ├── models/
│   │   ├── AILog.js           # AI interaction audit trail
│   │   ├── ProductCategory.js # Module 1 results storage
│   │   └── B2BProposal.js     # Module 2 results storage
│   ├── middleware/
│   │   ├── validate.js        # express-validator error formatter
│   │   ├── errorHandler.js    # Global error handler
│   │   └── requestLogger.js   # HTTP request logging
│   └── utils/
│       ├── logger.js          # Winston app + AI logger instances
│       └── responseHelper.js  # Standardized response wrappers
├── frontend/
│   └── src/
│       ├── pages/
│       │   ├── ProductUploadPage.jsx   # Module 1 UI
│       │   └── ProposalPage.jsx        # Module 2 UI
│       ├── components/
│       │   ├── ResultCard.jsx          # Reusable AI result card
│       │   └── LoadingSpinner.jsx      # Animated spinner
│       └── services/
│           └── api.js                  # Axios API client
└── logs/
    ├── combined.log       # All app logs
    ├── error.log          # Error-level logs only
    └── ai-prompts.log     # AI prompt+response structured logs

---

Quick Start

Prerequisites

• Node.js 18+
• MongoDB (local or Atlas)
• OpenAI API key

1. Clone & Configure

# Copy and fill environment variables
cp rayeva-ai/.env.example rayeva-ai/.env

Edit .env:

MONGODB_URI=mongodb://localhost:27017/rayeva
OPENAI_API_KEY=sk-your-key-here
PORT=5000
FRONTEND_URL=http://localhost:5173

2. Start Backend

cd rayeva-ai/backend
npm install
npm run dev

Server starts on http://localhost:5000

3. Start Frontend

cd rayeva-ai/frontend
npm install
npm run dev

Frontend runs on http://localhost:5173

---

API Reference

Module 1 — Category Generator

POST /api/ai/category

Request:

{
  "product_name": "Bamboo Toothbrush",
  "description": "Eco-friendly toothbrush made from bamboo handle with BPA-free bristles"
}

Response:

{
  "success": true,
  "message": "Category and tags generated successfully",
  "data": {
    "id": "65f...",
    "primary_category": "Personal Care",
    "sub_category": "Oral Care",
    "seo_tags": ["bamboo toothbrush", "eco friendly toothbrush", "plastic free dental care", "sustainable oral care"],
    "sustainability_filters": ["plastic-free", "compostable", "vegan"],
    "ai_log_id": "65f..."
  }
}

GET /api/ai/category?page=1&limit=10

---

Module 2 — B2B Proposal Generator

POST /api/ai/proposal

Request:

{
  "budget": 5000,
  "client_type": "Corporate office",
  "event_type": "Employee welcome kits"
}

Response:

{
  "success": true,
  "message": "B2B proposal generated successfully",
  "data": {
    "id": "65f...",
    "recommended_products": [
      { "name": "Reusable Coffee Cup", "quantity": 200, "estimated_cost": 1500, "sustainability_note": "BPA-free, replaces 200 disposable cups daily" },
      { "name": "Bamboo Notebooks", "quantity": 200, "estimated_cost": 2000, "sustainability_note": "FSC-certified bamboo, fully biodegradable" }
    ],
    "budget_allocation": {
      "product_cost": 3500,
      "packaging": 500,
      "logistics": 1000
    },
    "impact_summary": "This proposal replaces single-use plastics and supports sustainable sourcing.",
    "ai_log_id": "65f..."
  }
}

GET /api/ai/proposals?page=1&limit=10

---

Prompt Design

Why JSON Mode?

We use OpenAI's response_format: { type: 'json_object' } so the API guarantees valid JSON output — no regex parsing, no markdown stripping.

Temperature Strategy

• Module 1 (Category): temperature: 0.3 — low randomness for consistent categorization
• Module 2 (Proposal): temperature: 0.4 — slightly higher for product variety

Prompt Structure

SYSTEM PROMPT = Role definition + Output schema + Hard rules
USER PROMPT   = Specific input data + JSON keys expected

The system prompt defines what JSON keys to output and hard rules (e.g., budget allocation must sum to total). This drastically reduces schema validation errors.

---

Logging

All AI interactions are logged in two places:

1. logs/ai-prompts.log — Structured JSON, timestamped, module-tagged
2. MongoDB ailogs collection — Queryable audit trail with token usage

Sample log entry:

{
  "timestamp": "2026-03-05 18:30:00",
  "level": "info",
  "message": "AI Interaction",
  "module": "category-generator",
  "status": "success",
  "tokensUsed": 312,
  "processingTimeMs": 1842
}

---

Module 3 — Impact Reporting (Architecture)

Goal: Aggregate sustainability impact across all orders.

Endpoints:

• GET /api/impact/report — Full platform impact report
• GET /api/impact/report/:month — Monthly breakdown

Logic:

1. Aggregate ProductCategory docs → count sustainability_filters
2. Aggregate B2BProposal docs → sum product quantities
3. Apply conversion constants (e.g. 1 bamboo product ≈ 0.3kg plastic saved)
4. Call OpenAI to generate narrative summary
5. Return: plastic_saved_kg, carbon_avoided_kg, products_sustainably_sourced, impact_narrative

New Models: ImpactReport { month, plastic_saved_kg, carbon_avoided_kg, products_count, narrative, generatedAt }

---

Module 4 — WhatsApp Support Bot (Architecture)

Goal: Automated customer support via WhatsApp Cloud API.

Endpoints:

• GET /api/whatsapp/webhook — Verify webhook (token challenge)
• POST /api/whatsapp/webhook — Receive & process messages

Intent Routing:

User Message	Bot Action
"Where is my order?"	Look up order in DB → reply with status
"What is your return policy?"	Return static FAQ response
"I want a refund"	Create support ticket → escalate to human
Default	Friendly fallback + escalation offer

New Models:

• WhatsAppConversation { phone_number, messages[], intent, status, createdAt }
• SupportTicket { phone_number, issue_type, status, conversation_id }

New Services: whatsappService.js, intentClassifierService.js, orderLookupService.js

---

Environment Variables

Variable	Description
PORT	Backend server port (default: 5000)
NODE_ENV	development or production
MONGODB_URI	MongoDB connection string
OPENAI_API_KEY	OpenAI API key
WHATSAPP_TOKEN	WhatsApp Cloud API token
WHATSAPP_PHONE_NUMBER_ID	WhatsApp Business phone ID
WHATSAPP_VERIFY_TOKEN	Webhook verification token
FRONTEND_URL	Frontend URL for CORS

---

Tech Stack

Layer	Technology
Frontend	React 18 + Vite + TailwindCSS
Backend	Node.js + Express
Database	MongoDB + Mongoose
AI	OpenAI GPT-4o (JSON mode)
Logging	Winston (file + console)
Messaging	WhatsApp Cloud API (Module 4)
Env	dotenv