CafeBot

CafeBot is an AI ordering agent that cafes license to take and register customer orders. This repo is the marketing site (/, /how-it-works, /features, /pricing, /vs/:competitor, /case-studies, /blog) plus a live product demo at /demo — a fake coffee shop showing CafeBot's order-taking chatbot and staff dashboard running end to end.

Structure

• frontend/ — marketing pages (index.html, how-it-works.html, features.html, pricing.html, case-studies.html, blog/) and marketing.css. No build step — plain static HTML/CSS/JS.
• frontend/demo/ — the fake-cafe product demo: customer site + chat widget (index.html, about.html, menu.html, offers.html, hours.html, app.js, styles.css) and the staff order dashboard (staff.html/staff.css/staff.js).
• backend/ — server code (Express app, chatbot logic, API routes, and the /vs/:competitor comparison route)
• data/ — menu, promotions, order, and comparison-page data
• prompts/ — prompt templates for the demo chatbot

Setup

Requires Node.js.

1. Install dependencies:

   cd backend
   npm install
   

2. Create a .env file in the project root (same level as this README) by copying .env.example, then fill in a real ANTHROPIC_API_KEY:

   cp .env.example .env
   

Run

From backend/:

npm start

This serves the marketing site, the product demo, the staff dashboard, and the API all on one port (default http://localhost:3000):

• / — marketing homepage
• /demo — live product demo (fake cafe site + chat widget)
• /demo/staff — staff order dashboard
• /api/chat — chatbot API

Environment variables

Variable	Required	Default	Description
ANTHROPIC_API_KEY	Yes	—	API key used to power the chatbot
PORT	No	3000	Port the server listens on
TAX_RATE	No	0.0825	Sales tax rate applied to orders
DELIVERY_FEE	No	3.00	Flat delivery fee applied to delivery orders
CHAT_LIMIT_PER_IP_PER_MIN	No	15	Max chat messages per visitor IP per minute
CHAT_LIMIT_PER_IP_PER_DAY	No	200	Max chat messages per visitor IP per day
CHAT_DAILY_BUDGET_USD	No	25	Estimated daily API spend after which the chat pauses until the window resets

Notes

Orders are persisted to data/orders.json on disk. This works for local development but is not suitable for production on platforms with ephemeral or read-only filesystems (e.g. Vercel serverless functions) — writes there are not guaranteed to persist. Use a real database before deploying to such a platform.