Swift Embeddings

A macOS application that generates text embeddings using BERT, built with SwiftUI and the Apple Embeddings framework.

Overview

Swift Embeddings is a simple desktop application that allows users to generate vector embeddings from text input. It uses Apple's Embeddings framework to run BERT (Bidirectional Encoder Representations from Transformers) locally on your machine.

Features

• Clean, native macOS interface
• Real-time text embedding generation
• Display of embedding vector preview (first 5 values)
• Error handling and loading states
• Sandboxed application for security

Requirements

• macOS 14.0 or later
• Xcode 15.0 or later

Installation

1. Clone the repository
2. Make sure BERT model files are placed them in the app's resource directory:
   • config.json
   • pytorch_model.bin
   • tokenizer.json
   • vocab.txt
   • tokenizer_config.json
   • special_tokens_map.json
3. Open the project in Xcode
4. Build and run the application

Usage

1. Launch the application
2. Enter text in the input field
3. Click "Generate Embedding" to create the embedding vector
4. View the first 5 values of the resulting embedding

Technical Details

• Built with SwiftUI
• Uses Apple's Embeddings framework for BERT model integration
• Implements async/await for non-blocking operations
• Includes comprehensive error handling and debug logging