As I understand it, CQRS is not about separating into read/write databases (although that is an architectural decision that might be helped by CQRS or not, it is independent of CQRS).
CQRS is about separating (decoupling) read models from write models (models or interfaces, I guess). What that means is that the client no longer writes the data in the same way as it is consumed...
Correct me if I am wrong but at a basic level:
A client can use a Create Command (which might write a new row to a database).
Then, a client can use an EditProperty Command (which might alter the row in the database OR store the resulting delta changes in the database OR even another database, the later being related to Event Sourcing and not CQRS per se).
Then, a client might read a subset of the data as something else (say we store a full record of a user along with profile picture, maybe Client A sees the user as a login, merely username and a password hash while Client B sees the user as a name, surname and profile picture. This part is what makes CQRS so interesting in DDD as it allows for decoupling data from model across domains, offering to different domains different 'representations' of the data).
Disclaimer, I have not yet seen the code, but I thought it was an interesting point to comment on.