What are the steps to add replication to a non-replicated database?

A heads up, multi-master replication can be really complex, especially if one instance succeeds and another fails. You might have to rollback changes on the ones that succeeded to keep things consistent. This is where powerful systems like Cassandra with QUORUM and LOCAL_QUORUM checks show their value, although they usually rely on eventual consistency.

One common approach is to use log shipping. Basically, when you make a write to the database, the change goes into a write-ahead log (WAL). You can then send these changes to another server. When you're ready to commit the transaction, you commit on both databases, but you should implement a 2-phase commit for reliability. Just keep in mind that actually making this work isn’t straightforward, especially for embedded databases like SQLite which might not derive much benefit from added complexity. Other databases like MySQL and Postgres already have built-in replication features.

Interesting project you're working on! Some databases may not have a WAL like SQLite does, which complicates things a bit. You'll need to figure out how to handle changes from users seamlessly, especially in a local-first distributed setup. I think the idea of having all users interconnected on the same graph network is ambitious and could be quite rewarding!