Restaurant Analytics AI

AI-powered analytics platform that generates comprehensive business insights for restaurants by analyzing reviews, menus, and competitive positioning.

🚧 Work in Progress

Currently in active development. Core functionality works, but not yet deployed for public use.

What It Does

Restaurant owners get a full business analytics report by simply providing their restaurant's URL. The system:

1. Collects Reviews: Scrapes and aggregates customer reviews from multiple platforms
2. Analyzes Menus: Extracts menu items, pricing, and categorization
3. Generates Insights: Uses AI to identify trends, sentiment patterns, competitive positioning, and actionable recommendations
4. Delivers HTML Reports: Clean, printable analytics reports (export to PDF in browser)

Think of it as a DIY version of what consulting firms charge thousands for - automated market research and business intelligence for restaurants.

The Problem

Small restaurant owners don't have access to the kind of data-driven insights that larger chains get from expensive analytics platforms. They're flying blind on customer sentiment, menu optimization, and competitive positioning.

Tech Stack

Architecture:

• Monorepo with npm workspaces
• Shared TypeScript types across frontend/backend
• Clean separation of concerns

Backend:

• Node.js + Express + TypeScript
• MongoDB Atlas (data storage)
• AI integration for analysis (LLM prompts via env config)
• Stripe webhooks for payment processing
• State machine for async job processing

Frontend:

• Astro (fast, modern static site generator)
• Server-side rendering for reports
• Clean, printable HTML output

Deployment:

• Backend: Heroku (single dyno, no extra services)
• Database: MongoDB Atlas
• Frontend: Vercel

Architecture Highlights

Status Pipeline:

pending → processing → reviews_collecting → reviews_collected → 
menu_urls_collecting → menu_urls_collected → menus_extracting → 
menus_extracted → data_collection_complete → completed → failed

Key Design Decisions:

• Reports served as HTML (not stored as files) - generates on demand
• Idempotent job runner (Stripe webhooks + redirects hit same endpoint)
• In-process coordinator (no separate job queue service needed)
• External API keys/prompts configured via env vars only

Current Status

✅ Working:

• Review collection from multiple sources
• Menu extraction and analysis
• AI-powered insight generation
• Report generation pipeline
• Payment processing via Stripe

🚧 In Progress:

• UI polish and error handling
• Deployment configuration
• Performance optimization for large datasets

📋 TODO:

• Public beta launch
• Additional data sources
• Competitive analysis features
• Multi-location support

Local Development

# Ensure MongoDB is running locally or via Atlas
# Copy env templates
cp packages/backend/.env.example packages/backend/.env
cp packages/frontend/.env.example packages/frontend/.env

# Install dependencies
npm install

# Run both frontend and backend
npm run dev

Why I'm Building This

Two reasons:

1. Problem validation: I love food and have friends who run restaurants. They're consistently surprised by insights hidden in their own review data.

2. Technical exploration: Wanted to build something real with AI that generates actual business value, not just a chatbot demo. This combines web scraping, data pipeline orchestration, LLM integration, and product design.

What I'm Learning

• Building robust web scraping that handles rate limits and varied site structures
• Designing state machines for complex async workflows
• Prompt engineering for consistent, actionable business insights
• Balancing AI capabilities with practical product constraints
• Monorepo architecture for shared type safety

---

Built by @arist0tl3