vla-viz

Real-time visualization dashboard for monitoring Vision-Language-Action model inputs during training and inference.

Features

• Live WebSocket connection to training server (default: ws://localhost:8765)
• View observation images, query images, and ICL examples in real-time
• Timeline navigation with history playback
• Mock mode for development (?mock=true)

Quick Start

npm install
npm run dev