UpCycle

Inspiration

Everybody's home has clutter - lots of clutter. And most people don't know what to do with it. Normally it would take extensive research and several tabs to figure out whether something is even worth selling, and UpCycle is the solution. Even in our own homes, this problem is still prominent. From old monitors all the way to vintage clothes, UpCycle can help you with anything.

What it does

UpCycle has a user submit an image or take a photo, then uses that and optionally submitted additional information to give an analysis of the price, as well as previous sale prices and a recommended platform to sell it on.

How we built it

We created our frontend to handle the photo upload and send it to OpenAI's GPT-4o, which identifies the item and writes the listing. At the same time, our backend calls SerpApi to pull real sold prices from eBay. We combine those two results to show the user accurate pricing and a ready-to-post listing. Frontend is deployed on Vercel, backend is on Railway, both connected to GitHub so every push automatically redeploys.

Challenges we ran into

One problem we ran into was the price accuracy and making sure that we were pulling real, sold prices. Early on, the app was pulling from active listings, leading to inaccuracies and major discrepancies in the prices. Another challenge was name cleaning ChatGPT's descriptions. We originally had very descriptive names for the items, but that would lead to the backend having trouble finding prices for that specific item. A final important issue was outlier listings like listings that were listed extremely high. We fixed this by adding an outlier filtering value.

Accomplishments that we're proud of

Seeing as this is our first Hackathon, we were very proud that we were able to create a fully-functional app including a backend. We were very surprised with the alertness and accuracy of the app, as well as the genuine usefulness. We were also able to fully deploy the app and solved a real problem.

What we learned

This was our first hackathon, so we learned a lot in terms of frontends, backends, APIs, and how to implement them. I would say the biggest thing was just the experience and process of creating a website, something that I have never done before. I probably had to create accounts for at least five code-related websites, and we've overall gotten a lot more confident navigating computers like this

What's next for UpCycle

While the site works well as is, we would like to clean up the UI and especially work on making the price estimates more accurate, probably by finding a way to give Open AI access to other platforms like Amazon. Additionally, we want to make it into a mobile app for easy installation and access. The end goal is to minimize button presses and typing, while maximizing accuracy.

Built With

• github
• html
• javascript
• serpapi
• vscode