VeloGraphX v0.8.2 introduces production-ready Python packaging and PyPI distribution for the existing C++20 graph analytics engine.
Python packaging
VeloGraphX can now be installed with:
python -m pip install velographx
The Python package exposes the native VeloGraphX C++20 engine through pybind11 bindings.
Cross-platform wheels
Release CI builds and smoke-tests wheels for CPython 3.9–3.14 across:
- Linux x86_64 / manylinux
- Windows
- macOS Intel
- macOS Apple Silicon
A validated source distribution is also generated.
Python API
The package provides Python access to core VeloGraphX functionality including:
- Graph and UpdateBatch
- Incremental BFS
- Connected Components
- Triangle Counting
- k-core
- PageRank
- Weighted SSSP
Release validation
The packaging pipeline includes:
- Installed-wheel smoke testing
- Source-distribution build and installation testing
- Dynamic BFS correctness validation
- Package/version validation
- Release-tag validation
- Cross-platform CI
Secure PyPI publishing
PyPI publication uses GitHub Actions OIDC Trusted Publishing rather than a long-lived API token.
The publishing workflow is restricted to published GitHub releases and uses the protected pypi GitHub environment.
C++ engine
The existing VeloGraphX C++20 engine remains available for exact analytics on evolving graphs, including:
- BFS / unweighted SSSP
- Connected Components
- Triangle Counting
- k-core
- Weighted SSSP
- PageRank
This release primarily improves installation, distribution, and Python accessibility while preserving the existing graph-analytics architecture.
Previous release
v0.8.1 — CI Reliability & Release Workflow Fixes