Just use iGraph.
If you’re doing community detection, make sure to get the louvain-igraph module that adds the most cutting edge algorithms to iGraph. You can get it here.
NetworkX is the other big network analysis library, but it’s much slower than iGraph. The reason is that iGraph is written in C, so it’s orders of magnitudes faster than NetworkX, which is entirely written in native Python (much, much slower). For small graphs, NetworkX is fine, but for moderate sized networks (10,000 nodes or more) you really want to use iGraph.
If you get stuck working with iGraph, the mailing list community is amazingly helpful and supportive.