SpeakBites

Whiteboard Plans
Initial Landing Page
Restaurant View - Dashboard
Customer View - Order Taking

Inspiration

During the AWS chatbot workshop, we came up with the idea of using voice recognition to solve a problem. Randomly, we came up with the idea to create a more consistent and cost efficient solution for taking restaurant orders and offering an easy-to-use platform for restaurant workers to keep track of their orders.

What it does

SpeakBites uses cutting-edge speech recognition technology to accurately parse and process customer orders in real-time. Once an order is placed, it's instantly displayed on a user-friendly dashboard for kitchen staff, ensuring swift and precise preparation.

How we built it

We built this application using React as our frontend framework and Python Flask as our backend to handle server requests. We used Auth0 for authentication and OpenAI API for order processing.

Challenges we ran into

Our first initial struggle was coming up with a way turn a transcript into clean and consistent JSON to be used for the restaurant staff's order dashboard. We realized that OpenAI's Chat GPT API could handle this easily if it was given the right context. Without the right context and clear instructions on what to output, this API was returning inconsistent JSON that would completely break the application.

Accomplishments that we're proud of

We are proud that we were able to complete a functioning application based off of our initial idea. From a whiteboard to code in less than 24 hours, we were able to implement the base functionality and present a working application.

What we learned

Throughout the past 36 hours, we learned that simplicity is key. During our initial white boarding session, it was easy to go down rabbit holes and come up with complex yet useful features. There was one problem though, there was just simply not enough time to implement these features. We took the time to focus on the user, and how they would be interacting with the platform. What features were the most pertinent? How can we deliver them as simply as possible? What must we have, and what would be nice to have? Finding clear answers to these questions early on was important in leading us to a final working product.

What's next for SpeakBite

After this hackathon, we are going to fix major bugs. The app is not perfect and can use a lot of work. As our goal for the past 36 hours was to simply create a working prototype, perfecting the application was something we could not afford to do.