Inspiration Café orders often pass through several disconnected handoffs: cashier, customer, payment, kitchen, pickup, and inventory. We wanted to make that flow feel like one connected system. WristOrder was inspired by the idea of giving café staff a hands-free, voice-first ordering assistant through a Vexo wristband while keeping the customer and kitchen informed in real time. What it does WristOrder connects the entire café workflow:
- Staff can type or speak orders using the cashier screen or Vexo band.
- AI understands the order, applies menu items and modifiers, and reads it back for confirmation.
- Customers see their order and live total on a separate screen.
- Customers can pay by scanning a QR code, with mock Stripe payments available for demos.
- Paid orders automatically appear in the kitchen queue.
- Kitchen staff can mark orders ready, updating the pickup display and cashier notification.
- Inventory is reserved during checkout and deducted after payment.
- Owners can view sales, popular items, ingredient usage, inventory levels, reorder suggestions, and AI-generated reports. How we built it WristOrder is a local Node.js and Express application with a vanilla JavaScript frontend. We used:
- Google Gemini for order parsing and owner questions
- Deepgram for speech-to-text and text-to-speech
- The Vexo browser kit for wearable interaction
- Stripe Checkout and QR codes for payments
- SQLite for orders, inventory, and sales history
- Server-Sent Events for real-time updates across screens
- Resend and scheduled reports for owner communication Every external integration includes a mock mode so the complete experience can be demonstrated without API keys, live payments, or specialized hardware. Challenges we ran into One challenge was turning imperfect spoken language into accurate café orders, especially when customers mentioned quantities, modifiers, or similar-sounding menu items. We added menu key terms, confirmation readbacks, and cashier editing tools to make the process reliable.
We also had to keep multiple interfaces synchronized. A single order could be changing on the cashier, customer, kitchen, pickup, and owner screens at the same time. Server-Sent Events helped us broadcast updates without requiring a complicated frontend framework.
Inventory was another challenge. We needed to reserve ingredients before payment, consume them only after successful payment, and release them if an order was cancelled or expired. Designing those state transitions carefully helped prevent stock from being deducted twice. Accomplishments that we're proud of We built a complete café experience from voice order to pickup in a short hackathon window. The same order can move through ordering, confirmation, payment, kitchen preparation, pickup, sales reporting, and inventory tracking.
We are especially proud of:
- Integrating a wearable voice interface with a practical business workflow
- Creating a live multi-screen experience for staff and customers
- Adding inventory-aware ordering and low-stock suggestions
- Supporting both real integrations and reliable mock services
- Building a working owner dashboard instead of stopping at the ordering interface What we learned We learned that voice interfaces need confirmation and graceful correction, not just transcription. Showing the interpreted order and allowing the cashier to edit it makes the experience much more trustworthy.
We also learned that real-time state management is central to operational tools. Payment, inventory, kitchen status, and pickup notifications all depend on clear transitions and a shared source of truth.
Finally, we learned that mock integrations are extremely valuable during rapid development. They allowed us to test the full product flow even when live API credentials, network conditions, or physical hardware were unavailable. What's next for WristOrder Next, we would improve WristOrder by:
- Testing and refining the experience with live Gemini, Deepgram, Stripe, and Resend integrations
- Completing more physical Vexo band testing
- Supporting multiple registers and locations
- Adding bilingual and multilingual ordering
- Improving correction handling for ambiguous orders
- Adding staff accounts, permissions, and audit history
- Expanding analytics and forecasting for inventory and demand
- Connecting to real café point-of-sale and inventory systems
Built With
- deepgram
- deepmind
- vexo
