Architecture
Prisma Multi-Schema
Isolate your data like an Enterprise
Most SaaS boilerplates give you a single schema.prisma file. This works fine for a todo-list app, but breaks down when you're building a multi-tenant enterprise system with independent microservices (like Stripe does).
The Sailor Approach
Nebutra Sailor utilizes Prisma 7's multiSchema preview feature extensively.
In packages/platform/db/prisma/schema.prisma, you will see something like this:
generator client {
provider = "prisma-client-js"
previewFeatures = ["multiSchema"]
}
datasource db {
provider = "postgresql"
url = env("DATABASE_URL")
schemas = ["public", "auth", "stripe", "ai"]
}
// ... Models spread across different schemas ...Why schemas?
By physically isolating your tables into schemas (e.g., auth.Users, stripe.Subscriptions, ai.Embeddings), you achieve:
- Security boundaries: A bug in the AI service cannot accidentally drop the
stripebilling schema. - Microservice portability: If your AI service gets huge, it's trivial to move the
aischema to a completely different physical database later. - Cleaner Cognitive Load: You aren't scrolling through 200 models in one giant block.
When using create-sailor, if you opt-out of "Stripe integration", the CLI handles automatically pruning the stripe schema entirely before your first boot!
How is this guide?
在 GitHub 上编辑此页面
最后更新于