Backend-agnostic array operations for embedding experiments. Seamless switching between NumPy, MLX (Apple Silicon), and PyTorch (CUDA/MPS) with zero code changes. - View it on GitHub