Convert PostgreSQL schemas to Prisma

See genuine Prisma schema output generated from a 13-table PostgreSQL-shaped schema, including relations, composite keys and explicit caveats.

Open the editor Inspect real output

target: Prisma schema
access: Pro export
source specimen: 2 of 13 tables · seed/schema.json

Trellis exports a schema drafted on its canvas using PostgreSQL types. This page does not claim to parse a PostgreSQL dump or a live database.

01 / models

Models with their relation fields attached

The excerpt keeps two complete models from the 13-table export. Back-relations remain in the model where the full generator placed them.

actual exporter output2 of 13 tables · article, article_revision

model article {
/// One row is one piece of writing. The current body lives here; every earlier body is a revision row.
  id String @db.Uuid @default(uuid()) @id
  author_id String @db.Uuid
  title String
  slug String
  body String @default("")
  labels String[] @default([])
  search_vector Unsupported("tsvector")?
  published_at DateTime?
  created_at DateTime @default(now())
  updated_at DateTime @default(now())
  author account @relation("article_author_id_to_account", fields: [author_id], references: [id])
  article_revision_article_id article_revision[] @relation("article_revision_article_id_to_article")
  article_embedding_article_id article_embedding? @relation("article_embedding_article_id_to_article")
  comment_article_id comment[] @relation("comment_article_id_to_article")
  // not expressible in Prisma: primary key (id)
  @@unique([author_id, title])
  @@unique([slug], where: raw("published_at is not null"))
  // not expressible in Prisma: index using gin (search_vector)
  @@index([labels], type: Gin)
  // not expressible in Prisma without raw SQL: index using gin (title gin_trgm_ops)
  // not expressible in Prisma without raw SQL: index (author_id, published_at desc)
  // not expressible in Prisma: check (published_at is null or the author's account must still be active)
}

model article_revision {
/// One row is one earlier version of an article's body. The article and the revision number together identify the row, so there is no surrogate key.
  article_id String @db.Uuid
  revision Int
  title String
  body String
  edited_by String? @db.Uuid
  edited_at DateTime @default(now())
  article article @relation("article_revision_article_id_to_article", fields: [article_id], references: [id])
  edited_by_account account? @relation("article_revision_edited_by_to_account", fields: [edited_by], references: [id])
  @@id([article_id, revision])
  // not expressible in Prisma: primary key (article_id, revision)
  // not expressible in Prisma without raw SQL: index (edited_by, edited_at desc)
}

Columns: Scalar fields retain nullability and safe defaults.
Relations: Foreign keys become named relation fields with references.
Keys: Composite identities become model-level declarations.

02 / review

The generated file marks what needs a hand

Prisma cannot express every PostgreSQL type or raw index construct. Trellis leaves those decisions in the output instead of silently deleting them.

The full 13-table export writes its caveats into the generated file. These examples are copied from it:

// Unsupported types are left as Unsupported("...") so the file still parses.
// not expressible in Prisma: index using gist (path)
// not expressible in Prisma: check (entity_type in ('account', 'article', 'article_revision', 'comment', 'subscription', 'invoice', 'tag'))
// not expressible in Prisma without raw SQL: index (occurred_at desc)

source / generated Prisma schema output