If the libraries produce 128-bit values, then they could use Postgresql's UUID type. In fact, Postgresql should accept anything as a UUID in text mode as long a

If the libraries produce 128-bit values, then they could use Postgresql's UUID type. In fact, Postgresql should accept anything as a UUID in text mode as long as the value is in hex with a length of either 32 or 36 (32 + 4 dashes), though using binary mode would probably be faster if the UUID library and your driver support that. We've been using sortable epoch-based UUIDs as primary key for two different software products, storing them in Postgresql with the builtin UUID type, utilising binary mode. Performance is good.