Inspiration
Researching online often feels chaotic — too many tabs, unclear credibility, and no way to visualize how sources connect. We wanted a tool that helps students and researchers see relationships and trust between the sites they explore.
What it does
GraphTrust is a Chrome extension that lets you add any website to a personal research map. It uses local AI summarization to extract key ideas, connects related pages through shared keywords, and renders an interactive D3.js web graph — all stored fully offline for privacy.
How we built it
I built a Chrome Extension (Manifest V3) using:
- D3.js for dynamic graph visualization
- LocalStorage for on-device data
- Chrome’s built-in AI summarizer for keyword extraction
- Vanilla JavaScript and HTML for a lightweight interface All computation happens locally, so users can research safely without sharing data online.
Challenges we ran into
- Getting consistent summaries from the local AI model across different pages
- Managing dynamic graph updates without heavy performance cost
- Handling storage limits and persistence between sessions
- Designing a clean UI that balances readability and graph complexity
Accomplishments that we're proud of
- Built a fully local AI tool that respects privacy
- Achieved smooth real-time D3 visualization inside a Chrome extension
- Created an intuitive workflow for mapping research sources visually
- Demonstrated how graph thinking can enhance critical evaluation of information
What we learned
I learned how to combine local AI, visualization, and browser APIs to create useful research tools. I also discovered that visualizing web relationships helps users think critically about credibility, not just content.
What's next for GraphTrust
- Add trust scoring based on domain references and link density
- Support graph export/import for collaboration
- Integrate optional citation databases (e.g., Semantic Scholar, OpenAlex)
- Explore edge semantics — showing why two sites are connected
Log in or sign up for Devpost to join the conversation.