Create feature-centric and prompt-centric visualizations for sparse autoencoders (like those from Anthropic's published research). - View it on GitHub