Nota

Summarize, translate, and proofread - privately, with Chrome's Prompt API

Comment

Inspiration

The idea for Nota came from a simple but persistent problem — switching between browser tabs and AI tools to summarize or cite online information interrupts focus and flow.
We wanted a way to bring the power of language models directly into Chrome, so users could analyze, summarize, and save information without ever leaving the page.

With the launch of Google’s Gemini Nano and Chrome’s new Prompt API (LanguageModel), we saw an opportunity to build a fully local AI assistant — fast, private, and deeply integrated with the browser itself.

What it does

Nota is an on-device AI assistant for Chrome that lives right inside your browser.
It uses the built-in Gemini Nano model to help users:

  • Summarize long articles or selections
  • Translate text into English
  • Proofread writing for grammar and clarity
  • Find related quotes from literature or thought leaders
  • Extract citation metadata (title, author, date, URL) in JSON format
  • Save and manage highlights directly within the browser

All of this happens locally on your device, with no external API calls, ensuring speed and privacy.

How we built it

  • Built as a Chrome Extension (Manifest V3) with a side panel interface for prompt interaction.
  • Integrated Google’s Gemini Nano through Chrome’s Prompt API (LanguageModel) for natural language understanding.
  • Added context menu actions using the Chrome Context Menus API to process selected text directly from any webpage.
  • Implemented Chrome Storage API to locally store highlights and citations.
  • Used Chrome Scripting API to access the user’s selected text in real-time.
  • Developed a clean, minimal HTML/CSS UI and used JavaScript modules for logic.
  • Bundled everything with Rollup.js for efficient builds.
  • Utilized DOMPurify and Marked.js to safely render AI-generated and Markdown-formatted responses.

Challenges we ran into

  • Understanding and implementing Chrome’s new Prompt API, which is still in early stages and sparsely documented.
  • Handling message passing between the background script, side panel, and content scripts to keep actions synchronized.
  • Debugging context menu triggers and ensuring selected text was accurately retrieved and passed to the AI model.
  • Managing the local session state of the Gemini Nano model to prevent resets or crashes.
  • Designing a clean and responsive side panel UI that fits naturally inside Chrome’s interface.
  • Ensuring on-device privacy while maintaining usability and real-time feedback.

Accomplishments that we're proud of

  • Successfully integrated Gemini Nano through Chrome’s Prompt API — demonstrating real on-device AI inside a browser.
  • Built a fully offline, privacy-first AI assistant that doesn’t rely on any external servers or API keys.
  • Designed a smooth context menu workflow, allowing users to right-click text and instantly see AI-powered responses in the side panel.
  • Added a highlights manager to save, edit, and export useful snippets or citations.
  • Created a functional, extensible AI workspace that can handle multiple tasks with customizable parameters (temperature, top-k).

What we learned

  • How to leverage Chrome’s next-generation AI APIs to run models locally and securely.
  • The importance of clean architecture in Chrome extensions — separating background, UI, and content layers clearly.
  • How to use Markdown rendering and HTML sanitization safely in an AI-generated content environment.
  • The practical limitations and advantages of on-device AI — faster response times, zero data exposure, but limited model size.
  • Designing with privacy and user control as first-class priorities.

What's next for Nota

  • Enhanced model capabilities as Chrome expands Gemini Nano’s prompt API functionality.
  • Improved UI/UX with Material Design 3 styling and visual feedback for AI responses.
  • Offline caching for highlights and citation metadata, synced across sessions.
  • Multi-language support for translation and summarization.
  • Custom prompt templates and “saved prompt” presets for common research tasks.
  • Eventually publishing Nota to the Chrome Web Store as a showcase of on-device AI integration.

Summary

Nota represents a step toward personal, private, and performant AI — right inside the browser.
It brings natural language understanding closer to where users actually work: the web.

Built With

Share this project:

Updates