nx-parallel is a backend for NetworkX, a widely used pure-Python library for graph analysis. While NetworkX includes hundreds of useful algorithms (like TSP, PageRank, Dijkstra’s shortest paths, etc.) it can get really slow on large graphs since it runs on a single CPU core. nx-parallel accelerates these algorithms using joblib, running them in parallel across multiple CPU cores, threads, or nodes—depending on the parallel backend. To use it as a backend, simply pass backend="parallel" to supported NetworkX functions (e.g. nx.all_pairs_dijkstra(G, backend="parallel")), or input a nxp.ParallelGraph object instead of nx.Graph, or set the environment variable NETWORKX_BACKEND_PRIORITY="parallel".
