New Chrome Extension Runs Google's Gemma 4 AI Entirely On-Device With No Data Sent to External Servers
Summary
A new open-source Chrome extension brings Google's Gemma 4 AI entirely on-device using WebGPU, enabling powerful browser automation, semantic web search, and tab management with zero data sent to external servers.
Key Points
- A new open-source Chrome extension called gemma4-browser-extension brings fully on-device AI directly to the browser, using WebGPU and Transformers.js powered by Google's Gemma 4 model, with all processing happening locally and no data sent to external servers.
- The AI agent is equipped with a suite of browser tools enabling tab management, semantic webpage interaction via Retrieval-Augmented Generation, and natural language browsing history search using vector embeddings stored in IndexedDB.
- The extension follows a three-component architecture — a background script handling AI inference, a side panel serving as the persistent chat UI, and content scripts managing direct DOM interaction — ensuring efficient resource use and a responsive user experience.