Upcycle

Home page
Entering Items
Results
Specific Result Expansion
Saved Crafts

Inspiration

We have always been passionate about crafting, whether building popsicle stick houses when we were young or robotics and woodworking now. It hurts to see other people waste what we could put to good use, especially since we always have ideas about cool things to make. We hope this can inspire other people to do the same by building a website that will give them personalized ideas.

What it does

The user enters potentially wasted or scrap items on the website. These get added to an internal list, that is then submitted to the backend, which returns a response with the materials required (these materials are common household items or the items the person entered), instructions, a title, and a difficulty level. There are three responses, one for each difficulty level (Easy, Medium, Hard). This allows the user to make a well-informed decision when choosing a project based on their ambition.

How we built it

We used Reactjs for the front end, Supabase to store saved crafts for logged-in users, and the Cohere LLM API to generate personalized craft ideas.

Challenges we ran into

Firstly, we ran into several stylistic issues, which we are still debating. We spent a lot of time determining a common theme, style, or feel we were going for that also represented how the user should feel. Then, we ran into issues around using the generative API LLM, as this was our first experience using Cohere. The biggest challenge we ran into was parsing the data from the Cohere LLM API, as we spent quite a bit of time taking out the user-pertinent data and removing any extra special characters.

Accomplishments that we're proud of

We were really proud when the LLM API worked end-to-end, from retrieving the response of the prompt to formatting it well for the user. We're also proud of building a full website through React and making it look aesthetic.

What we learned

We learned how to various frontend and backend frameworks. We furthered our skills in front-end design, and skills around time management and efficient division of tasks. In addition, this was our first overnight hackathon, so we learned how to budget time to maximize rest and minimize tiredness.

What's next for Upcycle

We want to use image detection to let people take a picture and extract items from the picture. We also want to use image generation models to create step-by-step visuals and a nice picture of the final product. We hope to grow past just us three by building the website and promoting on socials to Cupertino and beyond.