DB Seed Generator

Before generating any output, read config/defaults.md and adapt all patterns, imports, and code examples to the user's configured stack.

Process

1. Read schema.prisma and extract all models, fields, types, constraints, relations, and enums.
2. Determine dependency order: models with no required relations first, then models that depend on them.
3. Generate a prisma/seed.ts file with realistic fake data that respects all constraints.
4. Output a ready-to-run seed file.

Dependency Ordering

Build a directed graph of model dependencies from @relation fields. Topologically sort so parent records are created before children.

User (no deps)           → create first
Organization (no deps)   → create first
Post (depends on User)   → create second
Comment (depends on User, Post) → create third

If circular dependencies exist, use a two-pass approach: create records with required fields first, then update with optional relation fields.

Data Generation Rules

Field Name Conventions

Generate realistic values based on field names:

Pattern	Generated value
firstName	"Elena", "Marcus", "Priya"
lastName	"Rodriguez", "Chen", "Okafor"
name (on User)	"Elena Rodriguez"
email	"elena.rodriguez@example.com"
phone	"555-0101", "555-0102"
avatarUrl, imageUrl	"https://example.com/avatars/1.jpg"
url, website	"https://example.com"
title (on Post/Article)	Realistic short sentences
content, body, description	1-3 realistic sentences
slug	Derived from title: "realistic-short-sentence"
price, amount	Realistic numbers like 29.99, 149.00
quantity, count	Small integers: 1-100
createdAt, updatedAt	Dates spread over the last 90 days
publishedAt	Either null or a date after createdAt
isActive, isPublished	Mix of true/false
role (enum)	Distribute across valid enum values
password	Pre-hashed placeholder: "$2b$10$..." — never plaintext

Constraints

• @unique: Ensure every value is unique. Use index suffixes: "user1@example.com", "user2@example.com".
• @id @default(cuid()): Omit from create calls — let the database generate them.
• @default(now()): Omit or set explicit dates for predictable ordering.
• @default(value): Omit to use the default, or set explicitly for variety.
• Enums: Only use values defined in the enum.
• Int, Float: Respect @db.SmallInt, @db.Decimal(10,2) etc.

Relations

• One-to-many: Create the "one" side first, reference its ID on the "many" side.
• Many-to-many (implicit): Use connect syntax after both sides exist.
• Self-referencing: Create root records first (no parent), then children pointing to them.
• Required relations: Always satisfy — never leave a required FK null.
• Optional relations: Mix of connected and null.

Output Template

import { PrismaClient } from "@prisma/client";

const prisma = new PrismaClient();

async function main() {
  // Clear existing data (reverse dependency order)
  await prisma.comment.deleteMany();
  await prisma.post.deleteMany();
  await prisma.user.deleteMany();

  // Seed in dependency order
  const users = await Promise.all([
    prisma.user.upsert({
      where: { email: "elena.rodriguez@example.com" },
      update: {},
      create: {
        email: "elena.rodriguez@example.com",
        name: "Elena Rodriguez",
        // ... fields
      },
    }),
    // ... more users
  ]);

  // Use createMany for bulk inserts where relations allow
  await prisma.post.createMany({
    data: [
      { title: "...", authorId: users[0].id },
      // ...
    ],
  });

  console.log("Seeded: X users, Y posts, Z comments");
}

main()
  .catch((e) => {
    console.error(e);
    process.exit(1);
  })
  .finally(() => prisma.$disconnect());

Scale Parameter

When the user specifies a count (e.g., "seed 50 users"), generate that many records. Default to 5-10 records per model. Use createMany for bulk inserts where possible. Wrap large seeds in a prisma.$transaction().

Upsert Pattern

Use upsert on a unique field to make the seed idempotent — safe to run multiple times.

Reference

See references/seed-patterns.md for dependency ordering details, realistic data patterns, and performance tips.