Networkx

If you're working with any kind of network or graph data in Python, this is the standard toolkit. It handles everything from social networks to biological pathways to citation graphs. You get the four main graph types (directed, undirected, multi-edge variants), a comprehensive set of algorithms for centrality, shortest paths, community detection, and graph generators for creating synthetic networks when you need test data. The pandas and numpy integration is solid, so you can go from DataFrames to graphs without friction. Visualization works through matplotlib with various layout algorithms, though for production dashboards you'll want to export to something more interactive. The API is intuitive and the 3-clause BSD license means you can use it anywhere.