VeloGraphX — Dynamic Graph Analytics & Incremental Graph Algorithms

VeloGraphX is a high-performance C++20 and Python engine for dynamic graph analytics and incremental graph algorithms on large, continuously evolving graphs. It supports exact maintained BFS/unweighted SSSP, connected components, triangle counting and k-core; exact weighted shortest-path distances with conservative recomputation fallback; and PageRank-related maintenance with residual/tolerance validation.

The engine combines mutable graph storage with localized maintenance and full-recomputation paths. Its central systems question is practical: when should an evolving graph analytic be repaired incrementally, and when should the engine recompute the exact result?

Dynamic graph algorithm guides

These focused guides explain the major concepts and APIs while providing stable entry points for developers and researchers searching for dynamic and incremental graph analytics:

• Dynamic BFS — exact dynamic breadth-first search
• Incremental BFS — maintain exact BFS after graph updates
• Dynamic graph algorithms in C++20 and Python
• Incremental graph analytics for evolving graphs
• Dynamic connected components — exact connectivity
• Dynamic triangle counting — exact evolving-graph counts
• C++ graph analytics library with Python bindings

User guides

• Architecture
• Python bindings
• Benchmark methodology
• Current limitations
• Reproducibility guide
• Paper artifact guide
• Workflow catalog
• Submission archival status
• Security policy
• Contributing

Generated C++ API reference

The public C++ API is generated from include/velographx/ with Doxygen. From a repository checkout:

doxygen Doxyfile
python3 tools/build_docs_seo.py build/docs/html

The generated HTML entry point is:

build/docs/html/index.html

The published site adds page-specific titles and descriptions, canonical URLs, Open Graph and Twitter metadata, a generated sitemap.xml, and a robots.txt file as part of the GitHub Pages build.

Python API

The Python package exposes the native engine through pybind11. Installation, local builds, interoperability with NumPy/SciPy/Arrow, and supported Python versions are documented in the Python bindings guide.

Install from PyPI with:

python -m pip install velographx

Research evidence

Publication-facing quantitative claims should be read together with their provenance and limitations. The authoritative claim mapping is maintained in PAPER.md, paper/results-ledger.md, and benchmarks/paper-evidence.json; historical workflow output is not automatically treated as current paper evidence.