@prisma/adapter-d1

Fast and Small Query Compilers We've been working on various performance-related bugs since the initial ORM 7.0 release. With 7.3.0, we're introducing a new compilerBuild option for the client generator block in schema.prisma with two options: fast and small. This allows you to swap the underlying Query Compiler engine based on your selection, one built for speed (with an increase in size), and one built for size (with the trade off for speed). By default, the fast mode is used, but this can be set by the user:

Bypass the Query Compiler for Raw Queries Raw queries ($executeRaw, $queryRaw) can now skip going through the query compiler and query interpreter infrastructure. They can be sent directly to the driver adapter, removing additional overhead.

Update MSSQL to v12.2.0 This community PR updates the @prisma/adapter-mssql to use MSSQL v12.2.0. Thanks Jay-Lokhande!

Pin better-sqlite3 version to avoid SQLite bug An underlying bug in SQLite 3.51.0 has affected the better-sqlite3 adapter. We’ve bumped the version that powers @prisma/better-sqlite3 and have pinned the version to prevent any unexpected issues. If you are using @prisma/better-sqlite3 , please upgrade to v7.3.0.

prisma-engines#5745: Cast BigInt to text in JSON aggregation When using relationJoins with BigInt fields in Prisma 7, JavaScript's JSON.parse loses precision for integers larger than Number.MAX_SAFE_INTEGER (2^53 - 1). This happens because PostgreSQL's JSONB_BUILD_OBJECT returns BigInt values as JSON numbers, which JavaScript cannot represent precisely.