Installation

Install TypeSeg from PyPI with Python 3.9 or later:

pip install typeseg

The package includes the general U-Net and Mamba models. NumPy and ONNX Runtime are installed automatically. No API key, model download step, or GPU is needed for CPU inference.

For throughput-oriented text analysis, start with typeseg.fast(text) or typeseg --model fast file.txt.

Optional GPU support

On a compatible NVIDIA CUDA system:

pip install "typeseg[gpu]"

The same Python API selects an available GPU backend. CUDA support uses ONNX Runtime for U-Net and CuPy for Mamba. macOS uses CPU inference; the GPU extra does not provide an Apple GPU backend.

Check the installation

python -c "import typeseg; print(typeseg.fast('SELECT name FROM users;'))"

Continue with Quick Start for CLI and Python examples, or try the browser demo without installing anything.