This article details how to utilize Chrome's built-in AI capabilities, which enable web applications to perform AI tasks without deploying or managing AI models independently. These features leverage APIs like the Prompt API, Summarizer API, Writer API, and Rewriter API, utilizing the Gemini Nano model.
Currently, these APIs function only in Chrome, though standardization across browsers is underway. They are designed for desktop and laptop use (Windows 10/11, macOS 13+, Linux) and demand specific hardware: At least 22 GB of storage, a GPU with >4 GB VRAM, and an unmetered internet connection. Mobile device support is not yet available.
Several built-in AI APIs are accessible, some via origin trials and others through an early preview program. Instructions for each API, including local and production deployment, are provided. Local testing on localhost is supported, with troubleshooting steps detailed.
Efforts to standardize these APIs are in progress through the W3C Web Incubator Community Group, involving feedback from major browser developers such as Mozilla and WebKit.
The article encourages developers to share their feedback on the use of Chrome's built-in AI features.