You can implement the same features of this customised version using standard Redis sets. I've been using Redis to implement some large directed graphs (e.g. 300,000+ nodes, > 1m directed edges) using the built in Redis data structures pretty efficiently.
The author should explain the benefits of this approach.