ZenStack’s goal is to become the unified data layer of the application. It has been advocating a model-first approach, which involves using a rich and coherent schema as the single source of truth of the application and automatically deriving as many workpieces as possible, such as access control, RESTful APIs, frontend hooks, and Zod schemas.

ORM is an essential part of it. To avoid reinventing the wheel, we started our journey as an extension package for Prisma ORM. However, as we added more features, we felt more constrained by the foundation, Prisma. Therefore, we made a bold decision for ZenStack V3 to reimplement the ORM part using Kysely (https://kysely.dev/). Its type-safe query builder gives us enough flexibility to go along the road. At the same time, we aim to maintain Prisma's excellent DX.

Here comes the ZenStack v3 Beta release! We've spent a lot of time working on the ORM's parity with Prisma, implementing bug fixes, and enhancing performance. Here are the highlights of the achieved results:

- The ORM client is essentially compatible with PrismaClient.

- Features that go beyond Prisma:

    [query builder API]

    [database-side computed fields]

    [polymorphic models]

    [strongly-typed JSON]

- Runtime plugin system that allows you to intercept queries at different levels.

- A new VSCode extension

Finally, the comprehensive yet coherent schema is a perfect match for both vibe coding and AI-assisted coding. It not only saves you time juggling many tools in a fragile setup, but also gives you a more deterministic result due to the slim code base.