Local Image Finder 🔍

A desktop application that allows searching through local images using natural language queries, powered by AI.

video1185266171.mp4

✨ How It Works

• Uses AI to understand and index your local images
• Converts images into searchable vector embeddings
• Enables natural language search queries
• Works completely offline after initial setup

🛠️ Tech Stack

• Frontend: React + Electron
• Backend: Python Flask
• AI Model: Moondream
• Vector Storage: ChromaDB

📋 Prerequisites

• Windows 10 or later
• Node.js v16+
• Python 3.8+
• Dedicated NVIDIA GPU

🚀 Quick Start

Backend Setup

# Navigate to backend directory
cd local_img_app/backend

# Create and activate virtual environment
python -m venv venv
.\venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Download Moondream model
# Place moondream-2b-int8.mf in the backend directory

#run python server
python backend.py

Frontend Setup

# Navigate to project root
cd local_img_app

# Install dependencies
npm install

# Start development server
npm run dev

Model Setup

• Moondream 2B
• Moondream 0.5B

💡 Usage

• Launch the application
• Select folder with images to encode
• Wait for indexing to complete
• Search using natural language
• Press enter then VOILÀ

Note

Encoding image takes a while(20-30s/image currently). If you just want to quickly test the app out, in backend folder, there is a test folder with a few image already encoded. Feel free to pick a few and test out the app capabilities.