Tabwise Chrome Extension

Tabwise is a Chrome extension designed to enhance your browsing experience by categorizing tabs into common categories (e.g., e-commerce, sports, social) and generating summaries using AI models. The extension leverages local AI inference as well as Gemini Nano's experimental summarization capabilities.

---

Features

• Tab Categorization: Classifies websites into predefined categories.
• Summarization API: Provides AI-generated summaries for each open tab using the experimental Gemini Nano API in Chrome.
• Local Inference: Runs AI models locally for privacy-first categorization.

---

System Requirements

Tabwise General Requirements

• Operating System: macOS (tested on Mac M1/M2 silicon), Windows 10/11, or Linux.
• Python Version: Python 3.8 or newer.
• Browser: Google Chrome version 129.0.6639.0 or newer (Canary recommended).

Gemini Nano Requirements (for Summarization)

• Operating Systems:
  • macOS: Version ≥ 13 (Ventura).
  • Windows: 10 or 11.
  • Linux: Requirements not specified.
• Storage:
  • At least 22 GB of free storage on the Chrome profile volume.
  • Note: After the download, Chrome deletes the model if storage drops below 10 GB.
• GPU:
  • Integrated or discrete GPU with 4 GB minimum VRAM.
• Network Connection: Non-metered internet connection.
• Important Notes:
  • Gemini Nano is currently not supported on Chrome for Android, iOS, or ChromeOS.
  • The requirements may change as Gemini Nano is under active development.

---

Installation

Step 1: Set Up Chrome with Gemini Nano

1. Download and Install Chrome Canary:

   • Visit the Chrome Canary download page and install the latest version.
   • Confirm your version is 129.0.6639.0 or newer.

2. Check System Requirements:

   • Ensure your device meets the requirements for Gemini Nano, including storage and GPU capabilities.

3. Enable Gemini Nano:

   • Open a new tab and navigate to:

     chrome://flags/#optimization-guide-on-device-model
     

   • Set the flag "BypassPerfRequirement" to Enabled.
   • Relaunch Chrome.

4. Enable the Summarization API:

   • Open a new tab and navigate to:

     chrome://flags/#summarization-api-for-gemini-nano
     

   • Set the flag to Enabled.
   • Relaunch Chrome.

5. Initialize Gemini Nano:

   • Open Chrome DevTools (F12) and run the following commands in the console:

     await ai.summarizer.create();

     • This forces Chrome to schedule a model download.

   • Wait 3–5 minutes for the download to complete, then run:

     await ai.summarizer.capabilities();

     • Wait until the response changes to "readily".

   • Troubleshooting:

     • If you encounter "The model was available but there was not an execution config available for the feature.", wait for 24 hours and try again.

---

Step 2: Install Tabwise Backend

1. Create a Python Virtual Environment:

   python3 -m venv env
   source env/bin/activate

2. Install Dependencies:

   pip install -r requirements.txt

3. Install ctransformers for Mac M1/M2 (if applicable):

   CT_METAL=1 pip install ctransformers --no-binary ctransformers

4. Download Required Models: Navigate to the backend folder and run the Model Manager to download TinyLlama:

   cd backend
   python model_manager.py ensure

   The Model Manager CLI supports several commands:

   • ensure: Downloads the required models(including TinyLlama).
   • download --model MODEL_NAME: Download a specific model
   • verify: Check if all models are valid
   • list: Show all available models
   • info --model MODEL_NAME: Show detailed information about a specific model Optional arguments:
   • --force: Force download even if model exists
   • --models-dir PATH: Custom directory for storing models (default: backend/models)
   • --cache-dir PATH: Custom directory for model cache (default: backend/cache)
   • --max-cache-size GB: Maximum cache size in GB (default: 4.0)

5. Run the Backend Server:

   python server.py

   The server should start on http://localhost:8000.

---