Open-Source Local AI Browser: A Privacy-First Alternative to Perplexity Comet

A new open-source project aims to bring the capabilities of advanced web agents to local environments, utilizing Electron, React, and Ollama to execute autonomous navigation and OS-level interactions without relying on cloud-based APIs.

Autonomous Web Navigation and OS-Level Execution

A new developer has introduced an open-source AI browser designed to function as a local web agent, drawing inspiration from the functionality of Perplexity Comet. Unlike traditional AI assistants that merely summarize static page content, this tool is engineered to actively navigate the web on behalf of the user.

The system operates by interpreting natural language commands and interacting directly with the browser's Document Object Model (DOM). By analyzing the screen and the underlying page structure, the agent can execute real OS-level interactions, including sendInputEvent for mouse clicks and simulated keystrokes, allowing for true autonomous browsing and task completion.

Technical Stack and Local Integration

The application is built using a modern frontend and desktop framework stack to ensure cross-platform compatibility and responsiveness:

• Frameworks: Developed using Electron for the desktop shell and React for the user interface.
• LLM Backend: The project leverages Ollama, ensuring that the Large Language Model (LLM) runs 100% locally on the user's hardware.
• Privacy: By removing the need for external API calls, the project provides a fully transparent environment where data remains on the local machine.

Hardware Requirements and Beta Testing

Due to the computational demands of running both the browser environment and the local LLM, the developer is currently seeking testers with high-end hardware. Specifically, a minimum of 16GB of VRAM is required to ensure stable performance and efficient model inference.