Inspiration
Modern web browsing often results in dozens of open tabs across work, learning, entertainment, and social sites. Managing them manually is overwhelming, distracting, and reduces productivity. I wanted to create a tool that automatically organizes tabs and provides meaningful summaries, helping users stay focused while keeping all data private and offline.
What it does
Tabber is a Chrome Extension that leverages Chrome’s built-in Prompt API (Gemini Nano) to classify and summarize tabs locally on your device. Key features include:
- AI-powered tab classification into categories like Work, Learning, Social, and Entertainment.
- Accordion view to expand or collapse tab categories for easy navigation.
- Focus mode with checkboxes to keep only relevant tabs open.
- Scroll position memory so you never lose your place in a tab.
How we built it
The extension is built using JavaScript, HTML, and CSS, bundled with Vite for fast builds. It uses Chrome Manifest V3 for the extension framework and relies on npm scripts to manage builds and copy static assets. The AI classification runs entirely locally in Chrome via the Prompt API, ensuring privacy, offline access, and fast performance.
Challenges we ran into
- Learning the nuances of Manifest V3 and background/service workers.
- Ensuring AI classifications were meaningful and consistent across different types of tabs.
- Managing tab state and scroll positions efficiently across multiple open tabs without impacting performance.
Accomplishments that we're proud of
- Successfully integrated client-side AI with Gemini Nano for on-device tab classification.
- Built a lightweight, fully offline extension that improves user productivity.
- Implemented scroll memory and focus mode, features rarely seen in tab management tools.
What we learned
- How to integrate AI into Chrome Extensions using the Prompt API.
- Best practices for tab management and UI state preservation.
- Optimizing Vite builds for Chrome Extensions and structuring static assets for deployment.
What's next for Tabber AI
- Build a self-learning tab classifier that improves over time using user interactions.
- Train a separate AI model for tab classification while leveraging Gemini Nano for guidance and on-device inference.
- Develop an advanced web scraper and analyzer to extract meaningful content from a wide variety of websites, enabling even more accurate categorization and summarization.
- Enhance automation and personalization, so Tabber can adapt to individual browsing habits and provide smarter recommendations.
- Explore ways to maintain privacy and offline functionality while improving AI intelligence and multi-site understanding.
Log in or sign up for Devpost to join the conversation.