GPI stands for Graphical Programming Interface which is a development environment for scientific algorithms that provides a visual workspace for assembling algorithms. Algorithm elements (i.e. nodes) can be linked together to form a flow diagram. Each node is executed according to the hierarchy of the diagram.

• Website

• Documentation

• Issues

Installing

macOS / Linux

Create a new conda environment for gpi:

conda create -n gpi python=3.9 fftw eigen qt compilers llvm-openmp -c conda-forge
conda activate gpi

Windows

Windows requires the MinGW-w64 GCC 13+ toolchain (for C++17 support in pybind11 nodes):

conda create -n gpi python=3.9 fftw eigen qt zlib gxx_win-64 distutils-activate-mingw -c conda-forge
conda activate gpi

Package notes:

• gxx_win-64 — provides MinGW-w64 GCC 13+ (C++17/C++20), gendef, and dlltool needed to build C/C++ extensions
• distutils-activate-mingw — configures Python's distutils/setuptools to use the MinGW compiler
• zlib — provides zlib.h required by some pybind11 nodes (cnpy.h)
• compilers and llvm-openmp are macOS/Linux packages and should not be used on Windows
• pthreads are provided by the MinGW sysroot (no separate pthreads-win32 needed)

All platforms

Install gpi from source:

git clone https://github.com/gpilab/framework.git gpi_source
cd gpi_source
pip install .

Build the core nodes:

gpi_init

You can now run gpi from your conda environment:

gpi

To build a node with C dependencies:

macOS / Linux:

cd /path/to/node
gpi_make --all

Windows:

cd C:\path\to\node
gpi_make

On Windows, gpi_make automatically passes --all via the gpi_make.cmd wrapper — no extra flags are needed. The MinGW-w64 toolchain (installed in the conda environment) is used to compile C/C++ extensions. If you encounter compiler errors on first run, re-run gpi_init to re-run the MinGW environment setup.