I seriously thought you knew about this. It is the icing on the cake, the final peice that makes all of the other work worth the effort.
It expands the heap storage past just memory to using file storage. In other words if you want to have a few terbytes of predicates active, which obviously don’t fit into memory, then with RocksDB, the SWI-Prolog rocksdb package and SWI-Prolog rocksdb-predictes package you can just start creating large amonts of predicates. AFAIK it has only been tested with facts, no clauses with bodies, but the ground work was put in place into rockdb-predicates to do this.
I pushed it to 60.01 GB (ref) which was way past the 16GB of memory on my machine. But that only worked on Ubuntu on WSL on Windows, it was not on Windows native which is what I am after because that is what many companies use at the user level.
It is not aviable via the normal means as it is currently a prototype created by Jan W and is located in his GitHub repository only.
The way it is suppose to load is with a Git URL but the - in the URL seems to be a problem.
pack_install('https://github.com/JanWielemaker/rocks-predicates.git').
Note: rocks-predicates is not a proper package so will not work with pack_install/1. In the future it might become a package but for now it is just a library.
I ran into a few other problems today because of it and just had to walk away out of frustraion to work on it more tommorow. However on the ChatGPT side of problem solving am making some really intresting progress but since it does not currenlty relate to SWI-Prolog it is not noted here, nor anywhere else at the moment.