What prisma/orm shipped

V8.0.0-rc.23 is out: MongoDB migrations track indexes by name, PostgreSQL scalars become convenience types, and SQLite and MySQL get convenience date and time types. Also this week: MySQL create, update, upsert and delete return written rows, and migrate dev no longer hangs witho

This week's release candidate, v8.0.0-rc.23, brings named MongoDB indexes to migrations, convenience scalar types on PostgreSQL, and convenience date and time types on SQLite and MySQL. MySQL create, update, upsert and delete now return the written rows, as they do on Postgres. Migration statements can rename, remove and fill value object fields and convert and delete variants on Postgres and SQLite. Prisma ORM no longer refuses to start when a second copy is loaded, migrate dev no longer prompts for a name without a TTY, and a failed MySQL KILL no longer turns into an unhandled rejection.

This week: citext and Postgres extension support in db sign, MySQL statements in one round trip, MongoDB index merging, and connect() that is safe to call twice. Publishing to npm is fixed after failing since 23 October.

This week's work on prisma/orm: db sign now adopts Prisma 7 databases that use citext columns or extensions such as pgvector, PostGIS and ParadeDB. MySQL statements go out in one round trip through the mariadb package, and a MongoDB field with both @unique and @@index gets one unique index. Calling connect() on a client again is now safe. Publishing to npm, which had failed on main since 23 October, is fixed. The convenience date and time types project and the MongoDB authoring project both closed out.

8.0.0-rc.20 is out: ORM commands run on cli-engine 0.9.0, so piped output is tab-separated and JSON needs --json. MySQL can now emit a contract, read a database's structure, migrate, verify and sign. Migrations rename or delete enum values and update the rows, and type aliases ca

Release 8.0.0-rc.20 is out. ORM commands now run on cli-engine 0.9.0, so a piped command prints tab-separated data lines and scripts that read JSON must pass --json. Type aliases in the schema can set @onCreate and @onUpdate, and the contract stores each column's real type instead of the alias name. MySQL gains a target that emits a contract and reads a database's structure, plus migrate, verify and sign against MySQL 8.4, with each runner saying what a failed migration left behind. Migration plan and db update now rename or delete an enum value and update the rows that hold it, and Prisma 8 can migrate a changed relation action or two-column primary key. Smaller fixes: SQLite binds each parameter to its own numbered placeholder, and promise-returning members of a closed client reject inst

This week: @noCheck removed and nullable list elements are written T?[]; migrations skip an operation only when its full effect is in place; contract infer reads column types from data types; defineContract gets the same ORM types as emitted contracts; database adapters take a lo

This week in the Prisma ORM repository: @noCheck is removed and a list that allows null elements is now written T?[]; migrations skip an operation only when its full effect is in place; contract infer reads column types from data types and no longer emits Unsupported(...) for hstore or tsquery; a contract written with defineContract gets the same ORM types as an emitted one and filters accept plain strings; the database adapters postgres(), postgresServerless(), sqlite() and supabase() take a log option so runtime warnings and swallowed hook errors are no longer discarded; middleware can observe or answer a grouped activity; MongoDB migration plan renames collections and fields and asks before removing a field; and packages can contribute collection methods computed from a model.

Row locks on ORM collections, migration plan and db update refusing data loss until each drop is answered, --rename for models and fields, sql values in contracts, and multi-field Postgres full-text indexes. 8.0.0-rc.17 and 8.1.0-dev releases also landed.

This week in the project: the ORM client locks the rows a collection reads with forUpdate, forNoKeyUpdate, forShare and forKeyShare until the transaction ends. Migration plan and db update refuse to lose data until each drop is answered with --delete or --rename, and --rename makes a renamed model or field a rename instead of a drop. Raw SQL in a TypeScript contract is now a sql value, a Postgres full-text index can span several weighted fields, and a contract written with defineContract gives ORM rows their field types. firstOrThrow was added to SQL and Mongo collections. Releases 8.0.0-rc.17 and 8.1.0-dev.9 also shipped.

Prisma 8.0.0-rc.14 ships hover docs in PSL files, completion for generic blocks, multi-file schema support, and fixes for nullable lists, table renames, and Postgres defaults.

Prisma 8.0.0-rc.14 is out. This week: hover shows declarations and docs in schema files; language server completes generic-block values; nullable scalar-list elements work correctly; tables rename without losing rows; multi-file schemas in LSP; serverless connect() gains db.orm and db.transaction; and Postgres migrations handle autoincrement() and float values properly.

V8.0.0-rc.12-dev ships MongoDB timestamps, Int64/Decimal128/Binary/Json types, multi-file schemas, advanced orderBy, and Prisma 7 migration improvements.

V8.0.0-rc.12-dev continues building out MongoDB feature parity with timestamps and rich type support, adds multi-file schema organization, enhances ORM querying with related-row sorting, and improves the Prisma 7 to 8 upgrade path. Fixes ensure composite keys and PostgreSQL identifiers work correctly.

Prisma 8 now uses verbatim table names by default, writes SQL defaults as sql literals, and supports prepared queries for faster row reads.

This week Prisma 8 aligns table names with what you write (no more silent case transforms), modernizes SQL column defaults from dbgenerated() to sql literals, and ships prepared queries for optimized row reads and includes. The contract infer command now correctly prints nullable arrays. Prisma 7 users can upgrade by pointing at their existing schema instead of hand-fixing inferred output.