Show HN: On-device browser agent (Qwen) running locally in Chrome (github.com)

A new Chrome extension has been introduced that enables AI-powered web automation directly on devices, utilizing the WebLLM framework with WebGPU acceleration for local large language model (LLM) inference. This on-device capability means that all automation tasks—such as navigating the web, clicking links, and extracting data—are executed without reliance on cloud APIs, prioritizing user privacy by keeping data local. After an initial model download of about 1GB, the extension can function offline, making it both versatile and secure for users. This development is significant for the AI/ML community as it demonstrates a practical implementation of multi-agent systems in browser automation, featuring a Planner Agent and Navigator Agent that work together to perform complex tasks efficiently. This approach allows for strategic planning and tactical execution within web environments, expanding the range of automation tasks easily achievable by end-users without compromising their data security. The extension is a proof-of-concept built on modern web technologies and aims to inspire further advancements in on-device AI applications, potentially shaping the future landscape of web interaction.