Living Map Extension

Inspiration

Web pages are dense and filled with ads and who knows what.

What it does

Semantically chunks a webpage into a side panel. The user can then cull it while immediately able to use it for navigation. Once polished, the user can use it to navigate, highlight sections, and create tl;dr summaries using the built-in AI Summarizer API.

How I built it

Using Firebase Studio, Gemini, AI Studio, and my ibgib library.

Challenges I ran into

My computer is old and the performance on Summarizer is atrocious. This was no fault to the API, but it meant that everything was geared towards working around this. So I had to build a custom semantic chunker instead of relying on the key-points Summarizer capabilities, especially since the max characters even possible would be 5k (and a typical Wikipedia article is 30k+ characters).

There were also challenges with some of the tooling with Gemini, though the model itself is Awesome.

Accomplishments that I'm proud of

The UI is pretty neat, with highlighting and navigation being extremely useful. The summaries also help the more dense topics.

What I learned

I need a new computer.

What's next for Living Map Extension

I need to expand the AI capabilities, including blending it with my own ibgib.com blank-gib web app and my ibgib architecture for replication and distributed collaboration.