bevy | Devpost

Drink Card
Search Function
Devyn AI
Search bar
Drink grid
Swipe feature

Inspiration

We were brainstorming on the couch, trying to think of what we could build for HackGT, when our friend walked in with coffee she just bought. It was the worst cup of coffee I've ever had, but it got me thinking. If only there was some way to find the drinks you love and share your favorites. Real people giving real opinions so that you can show off your taste or find something new curated to your preferences.

What it does

Bevy is where drink lovers come together. Whether it’s your morning coffee, a boba, or a new twist on a classic, Bevy makes it easy to share what you’re sipping and discover fresh ideas from a community of enthusiasts. Bevy is designed to help you find individual drinks rather than locations, making it easy for you to see what flavors and types match for you perfectly. It also includes an AI agent that can recommend you drinks based on what you like and prefer in your profile.

How we built it

We started with UI mockups in Figma to quickly iterate on design ideas before jumping into development. From there, we built the full stack using React, Tailwind, and Vite for the frontend, with Supabase powering our backend and database needs. For deployment, we used Vercel to ensure a smooth and fast hosting experience. On the AI side, we experimented with CedarOS and Mastra for chatbot functionality, but ultimately integrated OpenAI directly to provide reliable and responsive conversations. Each team member had a hands-on role in developing all the different layers of the stack, and together we were able to stitch everything into a cohesive, interactive web app. Using vertical integration, everyone got a taste of fullstack development while implementing their features.

Challenges we ran into

One of the biggest challenges was overambition. We had dozens of exciting feature ideas, but with only 36 hours, it was difficult to narrow our focus and deliver something both functional and polished. We also struggled with setting up CedarOS and Mastra for our AI chatbot integration, eventually deciding to pivot and rely solely on OpenAI. Beyond technical hurdles, managing time effectively between ideation, development, and debugging proved to be just as challenging as the code itself. We learned that biting off more than you can chew and thinking up things that are out of scope can be your downfall as a developer, while still dreaming big and using what we know to build.

Accomplishments that we're proud of

Despite the prior hurdles, we successfully built a working AI-powered agent that helps users discover drinks near them. We’re especially proud of the “swiping” feature that lets users explore new drink recipes created by others, saving favorites to their wishlist in a fun and intuitive way. We also take pride in the fact that we implemented an entire full-stack app, from UI mockups to deployment, within the hackathon timeframe. The AI Agent, despite its ups and downs, also proved to be really impactful, and shows that we were able to implement technologies that are relevant now to imrpove our app.

What we learned

We deepened our understanding of full-stack development, especially how frontend, backend, and AI integrations work together. We learned how to connect TypeScript with Supabase for data persistence, how to design scalable components with Tailwind, and how to handle real-world limitations like API integration and feature prioritization. Beyond the technical aspects, we learned about effective team communication, time management, and the importance of iterating quickly while remaining flexible. With different backgrounds for all our team members, and some not even being CS, it was a huge adjustment learning how to manage time and communicate effectively.

What's next for bevy

This is just version 1 of Bevy, and we’re excited to keep building. Next steps include refining our AI agent to provide even smarter drink recommendations, developing algorithms that utilize user preferences for enhanced recipe and explore page curation, and expanding social features such as global recipe rankings and collaborative drink challenges. We also plan to revisit CedarOS and Mastra once we have more time to experiment, aiming to make Bevy not just a discovery tool, but a true community hub for drink lovers everywhere. We'd love to hear what ideas you have for us!