SidebookLM

Inspiration

I wanted a fast, private study assistant that doesn’t rely on the cloud. Chrome’s new Prompt API with Gemini Nano made it possible to build an AI that runs entirely on-device.

What it does

SidebookLM converts any PDF into a summary, quiz, and flashcards, all inside Chrome’s side panel — no internet or upload required.

How we built it

Built with Manifest V3, PDF.js for document parsing, and Chrome’s Prompt API for local AI inference using Gemini Nano. The UI is made with HTML, CSS, and JS, styled for a clean side panel experience.

Challenges we ran into

Integrating PDF parsing with the Prompt API context size limits and managing UI state between modes (Summary, Quiz, Flashcards) were the toughest parts.

Accomplishments that we're proud of

Running generative AI entirely offline in a browser extension — fast, private, and self-contained.

What we learned

How to leverage Chrome’s on-device Gemini Nano and structure prompts for efficient summarization and question generation.

What's next for SidebookLM

Publish it to the store

Built With

• chrome
• css
• css-frameworks-&-libraries:-pdf.js
• gemini
• html
• javascript
• pdf.js
• storage
• tailwind
• tailwindcss-(for-styling)-platform:-chrome-extension-(manifest-v3)-with-side-panel-api-ai-engine:-chrome?s-built-in-prompt-api-powered-by-gemini-nano-(on-device)-build-tools:-vite-(for-bundling)-apis:-chrome-runtime
• vite