I already tried the third option with the python library networkx standard implementation of pagerank, and the speed difference was almost zero.
Not available for neo4j, it is available in memgraph, buuuut I checked the algo and it's horribly implemented. It returns wrong results (I've opened an issue on their gh and talked to the their team, they confirmed it's wrong).
I’m glad you’re taking a look at memgraph - I know it exists but I’ve never tried it out.
Have you tried any other foss centrality algo libraries?
My current plan is to expose an API that will provide a variety of WoT scores: DoS (number of hops), personalized PageRank, GrapeRank, etc. Then clients can access the APU and use different scores for different purposes.
The API can provide scores that are “global” or personalized, with personalized being more cypherpunk. Results can be scores of one pubkey at a time or can provide scores for a batch of pubkeys. Example: the top 100 pubkeys by GrapeRank score, not including pubkeys you’re already following, which could define a feed that you’d use when you’re bored with your existing feed. And if you get bored with that, maybe switch to the next 100 on the list, and so on.
