Jtokkit is an excellent repo and I think it could be easily support multi model token estimate, is there any milestone for this?