AI-Select is a zero-backend Chrome extension that adds a small floating action button next to text selected on any webpage. Pressing it triggers Chrome's built-in on-device AI (Summarizer API via Gemini Nano) to instantly generate concise, actionable summaries.
- Contextual UI: Hover and click the lightweight
✨floating button right alongside your active text selection. - Privacy-First: Operates 100% locally on your machine. No server-side analytics, no remote API calls, no cloud latency.
- Highly Configurable: Use the extension popup to customize output preferences (
TLDR,Teaser,Key Points,Headline), Lengths (Short,Medium,Long), and Formats (Markdown,Plain text). - Real-time Streaming: Toggle the streaming capability to see the text generated chunk-by-chunk natively.
- Smart Dismissal: Hides the UI cleanly when not in use or when clicking away, equipped with dedicated exit icons.
- Open Chrome and navigate to
chrome://extensions. - Enable Developer mode in the top right corner.
- Click Load unpacked.
- Select the
AI-Selectfolder. - Pin the extension on your toolbar to configure settings.
Running on-device AI requires substantial underlying hardware support:
- Google Chrome version 138+.
- Hardcap requirements on Device Memory (>16GB CPU or >4GB GPU).
- At least 22GB of storage.
- An unmetered connection for the one-time local model download fallback.
The extension executes a fallback download listener if the AI nano model hasn't been instantiated yet natively in Chrome.
If you get an error stating "The device is not eligible for running on-device model.", you need to manually enable the API flags.
Do this:
- Go to
chrome://flags/#optimization-guide-on-device-model - Set it to Enabled BypassPerfRequirement. (Note: Plain "Enabled" often fails to expose the API if the benchmark hasn't run).
- Go to
chrome://flags/#prompt-api-for-gemini-nano - Set it to Enabled.
- Relaunch Chrome completely.
Chrome has other built-in AI APIs which you can check out on the getting started guide:
- Translator API
- Language Detector API
- Summarizer API
- Writer API and Rewriter API
- Proofreader API
- Prompt API
For deeper architectural outlines, mechanical overviews of the Summarizer implementation, and contribution context, please refer to the Developer Guide.
