Agentic Coffee Chat Shop
Project type
Agentic AI / Chat Commerce Demo
Date
2025
Location
Remote
An agentic coffee shop that runs on an AI barista. The bot greets you, walks you through a craft coffee menu, collects your name and preferences, then spins up a Stripe Checkout for instant payment. Voucher codes unlock free drinks while operational alerts flow to email and Telegram—no human in the loop required.
What I Built
- •Next.js 15 app router with a chat-first UI and live menu reference. The assistant keeps conversations on-topic and summarizes orders (drink, temperature, quantity) before checkout.
- •Typed OpenAI GPT-4o-mini prompt via Vercel AI SDK that emits structured JSON. Server-side validation maps drinks to canonical menu items and calculates totals from a single source of truth.
- •On-demand Stripe Checkout sessions with metadata (items, customer name, barista notes) flowing through to webhooks for automated fulfillment.
- •Multi-channel notifications: Resend emails and Telegram push notifications trigger on successful payments. Voucher-based orders skip Stripe but hit the same notification rails.
- •UX polish: message bubbles with loading states, deep links to Stripe rendered as buttons, and a modal for voucher entry so users can claim free drinks without restarting the conversation.
Why it's interesting
- •End-to-end agentic workflow: model-constrained JSON output, server validation, payment orchestration, and multi-channel notifications all working together.
- •Production-minded hardening: origin detection for redirects, webhook signature verification, and environment-driven secrets management.
- •Polished UX: responsive chat-first ordering with menu context, typing indicators, voucher flow, and clear success/cancel states.
Tech Stack
Next.js 15TypeScriptVercel AI SDKOpenAI GPT-4o-miniStripeResendTelegram Bot APITailwind CSSshadcn/ui