Interesting project, and the visual display is appealing, but the results don't seem to be any better than from the recommendation engines of, for eg., Goodreads. In many cases, the other nodes end up being mostly other books by the same author, with one or two books by different authors in the same genre.

One thing that could improve this as an exploration and discovery engine is an option to disallow books by the same author as the original we input. That would allow me to, for eg., check out what books the engine thinks are similar to 'Thief of Time', without the result being just a bunch of other Discworld books (which is not really useful or new information).

Also, why so few related nodes (five) in the result? Is this a constraint placed explicitly or just a feature of the algorithm? Having more books returned would also help increase the probability that the user finds something actually useful and previously unknown to them.