Prisma Driver Adapter Runtime
The two-runtime client setup Prisma ORM 7 forces once the Rust query engine is gone and every database needs an explicit driver adapter. Its reason to exist is the upgrade guide's warning that adapters now inherit the underlying Node driver's pool defaults, which may differ significantly from v6: pg ships no connection timeout at all (0) where v6 used 5 seconds, so pool exhaustion stops surfacing as a database error and starts presenting as a hung service. The Node client pins max, connectionTimeoutMillis, idleTimeoutMillis, maxLifetimeSeconds and allowExitOnIdle explicitly, attaches the onPoolError and onConnectionError handlers that keep an idle-client error from taking the process down, and survives hot reload on globalThis so watch mode cannot leak a pool per save. The edge client uses PrismaPostgresAdapter over the Prisma Postgres serverless driver, where a single connectionString is the whole surface and there is no pool to inherit anything from. A prisma.config.ts rounds it out with the dotenv import that v7 no longer does for you and the direct, unpooled migration URL that replaces the removed directUrl.
npx shadcn@latest add https://ui.aryank.space/r/prisma-driver-adapter-runtime.jsonInstalls from ui.aryank.space. To add it by hand, copy the files in Files below, or register the @compronents namespace via the docs.
Prisma 7 hands pooling to the pg driver, and pg ships connectionTimeoutMillis: 0, wait forever. The database hiccups, the checkout query sits on the pool queue with no deadline, and the request above it hangs until the client gives up. Watch the timer climb with no failure and no answer, the worst outcome a query can have.
[email protected]@prisma/[email protected]@prisma/[email protected]@prisma/[email protected]@prisma/[email protected][email protected][email protected]