Vertex Browser

Graph Structure
Cluster Description
Semantic Search
Similarity Weights on Edges

Inspiration

I was rushing to submit an assignment due in a few minutes when I realized I couldn't find one of my sources for the paper. I hurriedly clicked through each tab, went through my history, and tried searching for it on the web, but couldn't find it in time. I submitted the paper late and realized afterward that it was in one of the 40 tabs I had open. This made me invest in finding a better way to organize tabs so I can better optimize intensive web searching.

What it does

Vertex Browser is the First-Ever AI-Powered Graph-Based Browser. It organizes your tabs in a graph, with edges weighted by the similarity of the pages. The graph is color-coded by clusters of similar tabs, and you can search to find a cluster you're looking for. There's also a cluster summarization of each tab you click on the graph to help you find what you're looking for.

How we built it

We used the Anthropic API for Claude to judge the similarity between webpages. We clustered tabs with high similarity scores and implemented semantic search on them. We also used Claude API calls for multi-document cluster summarization and topic tag generation, which aided in the search functionality.

Challenges we ran into

We ran into challenges optimizing semantic search efficiency and the speed of calculating similarity scores (an API call to Anthropic always led to a small delay).

Accomplishments that we're proud of

We're proud to have created a unique browser that effectively solves a real pain point for us (and many others). We're proud that this is something we would use in the real world to make browsing more efficient. We're also very proud to have worked through all the challenges to create a genuinely helpful product.

What we learned

We learned how to integrate AI tools like Claude into web applications and debug alongside the AI. We learned how to ideate features for an MVP and collaborate with others.

What's next for Vertex Browser

We're very excited about the future of Vertex Browser and plan to build it out into a full web application after this hackathon. We plan to optimize speed on features like semantic search and add new features that help you browse effectively.