Does StellarGraph support unsupervised GraphSAGE on Neo4j graphs?

[silva-vinicius]

I am working on a project where I have to generate node embeddings in a large directed graph stored on Neo4j, that does not fit on memory and was wondering if I could use StellarGraph's unsupervised GraphSAGE implementation in order to do so. When reading the available documentation, I came across these 2 demos:

1- Node representation learning with GraphSAGE and UnsupervisedSampler

2- Directed GraphSAGE with Neo4j

Demo 1 covers working with unsupervised GraphSAGE, but requires the graph to be in memory.
Demo 2 allows for the graph to be stored in Neo4j, but implements a supervised task.

I think both demos are helpful, but I kind of need the best of both worlds here: Execute unsupervised GraphSAGE in a directed graph that is stored in Neo4j. Ultimately, I could lower the ambition of my project, work with a smaller graph and completely ignore demo 2, but of course it wouldn't be ideal.

Any advice here?
Thanks