update-schema

When the user asks to modify the Prisma schema, follow this sequence.

Step 1 — Edit prisma/schema.prisma

Adding a new model:

model <Name> {
    id        String   @id @default(cuid())
    // required fields
    // optional fields (mark with ?)
    createdAt DateTime @default(now())
    updatedAt DateTime @updatedAt

    // relations
    userId String
    user   User   @relation(fields: [userId], references: [id], onDelete: Cascade)

    @@index([userId])
    @@index([/* other commonly filtered fields */])
    @@map("<snake_case_table_name>")
}

Adding fields to an existing model:

• Nullable new fields (safest for existing data): fieldName Type?
• Non-nullable with a default: fieldName Type @default(value)
• Non-nullable without a default: requires a migration with a --create-only step to backfill data before applying

Changing relations:

• Adding a one-to-many: add the array field on the "one" side and the FK field + scalar on the "many" side
• Always set onDelete: explicitly — use Cascade for child records that must not outlive the parent, Restrict or SetNull otherwise

Naming conventions:

• Table names: @@map("snake_case") on every model
• Column names: @map("snake_case") on every field that differs from the Prisma field name
• All IDs: @id @default(cuid())
• Always add @@index on foreign key columns

---

Step 2 — Run the migration

pnpm prisma:migrate

This creates a new migration file in prisma/migrations/ and applies it to the dev database.

If the migration involves a non-nullable column without a default (data backfill needed):

pnpm exec prisma migrate dev --create-only   # generate SQL without applying
# edit the generated SQL to add UPDATE statement for existing rows
pnpm exec prisma migrate deploy              # apply the edited migration

---

Step 3 — Regenerate the Prisma client

pnpm prisma:generate

This must run after every schema change. The generated client lives in src/generated/prisma/.