Dinee - Voice AI Agent for Restaurants

Dinee is a smart assistant for restaurants that manages calls, orders, and callbacks like a real human, letting restaurant focus on great service.

Comment

Inspiration

AI voice agents are impacting every industry, from banking to software. To extend this initiative, I built a voice AI agent for restaurants that specializes in handling multiple calls simultaneously, adding and updating orders, calling back customers, providing updates, and all of this without human intervention.

What it does

The main aim of the software is simple: to ease the workload of restaurant owners so they can focus on serving their customers.

A few key features Dinee provides:

  1. Handling Calls: The agent assists users on both inbound and outbound calls.
  2. Order Taking and General Query: Dinee excels at taking orders, updating transcripts of the conversation, and answering general questions.
  3. Calling Back the Customer: Restaurants can command the agent to call back customers to clarify or provide updates about orders.
  4. Handling Multiple Calls Simultaneously: While staff are limited in how many calls they can take at once, Dinee can handle any number of calls all at the same time.

How we built it

  1. The application was built entirely in Kiro from frontend to backend, integrating OpenAI Live API, migrating to shadcn-style design, and even generating commit messages (thanks to Kiro hooks).
  2. We used Kiro IDE to generate systematic plans, organize code, modularize everything, and even clean up unused files in the codebase.
  3. Technology and tools used:
    • Frontend: Next.js, TailwindCSS
    • Backend: Fastify (Node.js), Next.js, Convex, Twilio
    • LLM: Gemini, OpenAI

Challenges we ran into

  • One of the biggest challenges was displaying calls, orders, and other information in real time, so restaurants are notified immediately when Dinee handles a new call or an order is placed. We solved this using Convex DB’s native realtime API, which is perfect for applications like this.
  • Another big challenge was integrating the OpenAI Live API with Twilio. Our solution forwards all incoming and outgoing calls to a custom websocket server, which handles all bi-directional communication.

Accomplishments we're proud of

  • This hackathon turned me into a professional Vibe coder special thanks to Kiro IDE.
  • Without referencing docs for Fastify, OpenAI Live API, or Twilio, I was able to integrate all three together. All thanks to Kiro for making this possible.

What we learned

  • First and foremost, I found a new IDE I can rely on. This will increase the speed of my future software releases.
  • I also gained hands-on experience with new technologies such as Twilio and OpenAI Live API.

What's next for Dinee - Voice AI Agent for Restaurants

  • The vision for Dinee is to integrate with third-party vendors like Clove, making it even easier for restaurant owners to use this software built with Kiro.

Built With

Share this project:

Updates