Reeflect AI

Main Screen in AI mode with more points
Main screen in AI Mode
Main screen in Google Search Mode
Main screen in AI Mode with cache response returned
Brand new query made, response cached
Accepted Cache Response
Win screen to plant coral on your behalf and restart

Inspiration

We noticed a problem... by 2027, AI's global annual water consumption is projected to be between 4.2 and 6.6 billion cubic meters. That is a lot of water usage that can affect ocean life and other parts of our environment! We wanted a way to get people to lessen their AI use for everyday questions while still giving them a way to access it, all while positively impacting the environment.

What it does

Encourages users to use RAG Cache Responses rather than creating a new response for an already made one. This is done by rewarding points to grow coral, and once you get to a certain threshold of coral, Reeflect AI will plant coral on your behalf and reset your points. Points are also rewarded for just making a google search on our website rather than querying an AI model.

How we built it

We built the front end and the back end using Next JS, and store the common queries and their responses into a Vector Database using an embedding model from hugging face. That way if someone asks a similar query that is not word for word equivalent, they will still get the list of cached responses for that query. We used Clerk for authentication as well as saving user points and coral planted.

Challenges we ran into

Our biggest challenge was working with the vector database, as it was something new for all us to work with. Luckily Vercel offers easy connection to Upstash, so all we needed to do was apply the appropriate embedding model and create the comparison logic!

What we learned

This project allowed us to all further our knowledge in working with Next JS (Client/Server side interaction) and working with vector databases. Our entire team has never worked with vector databases and embedding models, so it was super fun to be able to read into documentation and learn a lot more!

What's next for Reeflect AI

Partner with coral reef relief organizations in order to plant coral grown by users monthly in order to make a positive impact on the environment, as well as add a friend system to compete with friends and see who plants the most coral!