Google Prompt API

Join the Google Early Preview Program

Read this for details: https://github.com/explainers-by-googlers/prompt-api

Overview

This Next.js application demonstrates integration with Chrome's built-in AI capabilities using the experimental Prompt API. It provides a chat interface where users can interact with an AI model directly in their browser.

Prerequisites

• Node.js (v14 or later)
• Google Chrome (version 127 or later, Canary or Dev channel recommended)
• At least 22 GB of free storage space
• GPU with at least 4 GB VRAM
• Non-metered internet connection

Chrome Setup

1. Download Chrome Dev or Canary channel
2. Ensure Chrome version is 128.0.6545.0 or newer
3. Set Chrome flags:
   • Navigate to chrome://flags/#optimization-guide-on-device-model
   • Set to "Enabled BypassPerfRequirement"
   • Go to chrome://flags/#prompt-api-for-gemini-nano
   • Set to "Enabled"
   • Relaunch Chrome

Verifying Gemini Nano Availability

1. Open Chrome DevTools console
2. Run: await window.ai.canCreateTextSession();
   • If it returns "readily", you're set
   • If not, run: await window.ai.createTextSession();
3. Relaunch Chrome
4. Go to chrome://components
5. Check if Optimization Guide On Device Model is present with version ≥ 2024.5.21.1031
   • If not, click "Check for update"

Project Setup

1. Clone the repository:
2. Install dependencies: npm install
3. Start the development server: npm run dev
4. Open http://localhost:3000 in your Chrome browser

Key Components

• AIStatus: Displays the current status of the AI model
• ChatInterface: Manages the chat session and message handling
• ChatArea: Renders the chat messages
• MessageInput: Handles user input for sending messages

Styling

This project uses Tailwind CSS for styling. Custom styles can be added in app/globals.css.

Limitations and Notes

• This app uses an experimental API and is for demonstration purposes only
• The Prompt API is intended for local prototyping, not production use
• AI functionality may not work in Incognito or Guest mode
• The app currently only supports English (US) language

Troubleshooting

If you encounter issues:

• Verify Chrome flags are correctly set
• Ensure Gemini Nano is downloaded (check chrome://components)
• Confirm your device meets the hardware requirements
• Try relaunching Chrome or clearing browser data

Feedback and Contributions

We welcome feedback and contributions. Please open an issue or submit a pull request on our GitHub repository.

License

MIT

---

This project is part of the Chrome Built-in AI Early Preview Program. For more information, please refer to the [https://forms.gle/DWYuEkLYePsFfCZ7A](official documentation).