Changelog

All notable changes to this project are documented here. The format is based on Keep a Changelog, and the project adheres to Semantic Versioning.

[4.0.0] - 2026-07-08

Repackaged the Python distribution. The library is imported as the hydra_image_processor package (conventionally import hydra_image_processor as HIP), which wraps the compiled Hydra CUDA extension and exposes a Pythonic, snake_case API (e.g. gaussian, device_count).
Unified the build on scikit-build-core. A single pip install . now drives CMake, builds the extension, and installs the package — the same path used locally, in CI, and by the conda-forge feedstock. This retires the recipe-side patches the feedstock had to carry.
Made the source self-sufficient for conda-forge: CUDA C++ standard raised to 17, hardcoded GPU architectures removed in favor of CMAKE_CUDA_ARCHITECTURES (default all-major), and a package-relative install target added.
Single source of truth for the version (pyproject.toml); __version__ is now read from the installed package metadata.

Top-level HIP and Hydra compatibility shims so legacy import HIP / import Hydra code keeps working.
Windows + Linux CI build-gate (.github/workflows/ci.yml) and a tag-driven release workflow with a version guard (.github/workflows/release.yml).
Python TIFF accuracy suite (src/Python/Test/test_accuracy.py) mirroring the MATLAB AccuracyTest.m; it is the GPU-correctness gate for releases (see TESTING.md).

make_ellipsoid_mask produced an off-center, asymmetric structuring element (shifted one voxel), so morphological ops using it diverged from the MATLAB/C++ results.
The command-framework headers now include <cstdint> explicitly instead of relying on transitive includes, fixing the build with newer gcc/libstdc++ toolchains.

The in-repo conda-build recipe (recipe/) and the anaconda.org upload workflow. Distribution is now solely via conda-forge.