Local Embeddings Demo

Generate vector embeddings for Binance tickers using Hugging Face's Transformers.js running directly in your browser.

Ticker Embeddings

Model Status: Initializing...

Implementation Details

Web Worker Processing

  • • Offloads embedding generation to a dedicated thread
  • • Keeps UI responsive during processing
  • • Handles batched processing for better performance
  • • Real-time progress updates

Transformers.js Integration

  • • Uses all-MiniLM-L6-v2 model for embeddings
  • • Quantized model for efficient browser execution
  • • Automatic model caching
  • • 384-dimensional embeddings

Applications

Semantic Search

Find similar trading pairs based on their market behavior and characteristics using vector similarity search.

Pattern Recognition

Identify trading patterns and correlations by analyzing the relationships between token embeddings.

Market Analysis

Use embeddings with machine learning models for advanced market analysis and prediction tasks.

Performance Considerations

  • Initial model download: ~50MB (cached for future use)
  • Processing speed: ~100-200 tickers per second
  • Memory usage: ~100MB for model + processing
  • Recommended for modern browsers with WebAssembly support