Nano Prompt UI – Local-Only Gemini Nano Side Panel for Chrome

TL;DR: A privacy-first Chrome extension that runs entirely on-device using Chrome’s built-in Gemini Nano language model. Now updated to a Side Panel experience for persistent, uninterrupted AI assistance alongside your browsing.

Unlike standard popups that close when you click away, Nano Prompt UI lives in your Side Panel.

• True Multitasking: Read an article on the left while the AI summarizes it on the right.
• Persistent Sessions: Copy text from a website and paste it into the chat without the window closing or the AI forgetting its thought process.
• Background Processing: Long generation tasks (like summarizing a long PDF) continue even if you switch tabs.

• 100% Local & Private: Runs on the window.ai (Prompt API) built into Chrome. No data leaves your device.
• Smart Context Engine:
  • One-Click Summarization: Instantly reads the active tab to generate concise summaries.
  • Smart Truncation: Intelligently chunks long articles to fit the model's context window.
  • Context Optimization: Automatically detects casual chats (e.g., "Hi") and skips loading page context to save battery and reduce latency.
• Robust Session Management:
  • Auto-Saving: Chats are saved locally and sync metadata across devices.
  • Session Control: Rename, delete, or switch between multiple chat threads via the dropdown menu.
  • Markdown Support: Export full chat history to .md files.
• Rich Input & Media:
  • Multimodal Support: Attach images to your prompts (auto-resized for performance).
  • Voice Mode: Built-in speech-to-text for dictating prompts.
  • Templates: Quick-start prompts for common tasks (Translation, Proofreading, etc.).
• Advanced Configuration:
  • Temperature & TopK: Fine-tune the creativity and vocabulary of the model.
  • System Prompt: Customize the AI's persona (e.g., "You are a cynical senior engineer").

1. Download/Clone this repository.
2. Open chrome://extensions in your browser.
3. Toggle Developer mode (top-right corner).
4. Click Load unpacked and select the folder containing these files.
5. Pin the Extension: Click the puzzle piece icon in Chrome and pin "Nano Prompt UI" for easy access.

To use this extension, you must enable Chrome's experimental AI features:

1. Open chrome://flags and enable the following:
   • Prompt API for Gemini Nano: chrome://flags/#prompt-api-for-gemini-nano
   • Optimization Guide On Device Model: chrome://flags/#optimization-guide-on-device-model (Select "Enabled BypassPrefRequirement")
2. Relaunch Chrome.

1. Go to chrome://components.
2. Find Optimization Guide On Device Model.
3. Click Check for update.
4. Wait until you see a version number (e.g., 2024.5.21.1) and Status: Up-to-date.
   • If it says "0.0.0.0", the model is still downloading. Keep Chrome open.

• Open the Panel: Click the extension icon to toggle the Side Panel.
• Check Context: Click the Document Icon (📄) in the top header to see exactly what text the AI is reading from your current tab.
• Stop Generation: If the AI starts rambling, hit the Stop button immediately.
• Image Analysis: You can attach images! The extension automatically resizes them to ensure they fit within storage limits.

• "Model Unavailable": Ensure you have relaunched Chrome after enabling flags. If it persists, the model might still be downloading in the background.
• "Context Empty": Some system pages (chrome://, about:blank) or complex PDF viewers cannot be read by extensions for security reasons. The extension will alert you if this happens.

The Unlicense — see LICENSE.txt.

Built by Vimal "Vibe Coded" with AI.