Self-serve logical replication
You can now enable logical replication for your Neon project from the Settings > Beta page in the Neon Console. This feature lets you replicate data changes from Neon to external data services and platforms.
Get started with one of our logical replication guides.
This feature is currently in Beta. If you've got requests or feedback, let us know via the Feedback form in the Neon Console or our feedback channel on Discord.
Clearer language around compute types
Words matter. We've changed our naming convention around compute types: from RW compute
and RO Replica
to a cleaner, more straightforward Primary compute
and Read replica
.
With Neon's unique architecture, where we separate storage from compute for copy-on-write branching, you can choose the size and features for the compute that powers your branch's database independently from your data.
- Primary compute — When you create a branch in Neon, a primary compute is automatically created alongside it. You can think of your primary compute as the main engine for your branch. It supports read-write operations, though you can modify database permissions using traditional Postgres roles.
- Read replicas — When you’re ready to scale your application, you can direct read-only traffic to one or more read replicas. Unlike traditional systems where data is physically replicated, Neon’s read replicas access the same data source as the primary read-write compute — at no additional storage cost.
For more information, see:
create-app
improvements
Neon CLI The Neon CLI create-app
command initializes a new Neon project and bootstraps a full-stack application using your preferred package manager. This feature is very new, but it's maturing fast:
- It now supports Prisma as another ORM option for your bootstrapped application.
- It also now creates two separate branches for your app's Neon project: your main branch and a development branch.
To learn more, see Neon CLI commands — create-app.