How to avoid CQRS from breaking our microservice

We have microservice that is getting near real-time writes and reads updates against a neo4j graph datasource.

That microservice's logic became bigger and bigger and the writing and reading logic became fatter.

For the sake of performances and Domain Driven Design - we would like to separate the read and write logic(aka cqrs)

But writing and reading must be from the same datasource(we cant afford a "sync" latency time and having two datasources - it wont meet our up-to-date requirements)

How can I apply CQRS without breaking my microservice architecture by sharing the same datasource?

Thanks, ray.