db-schema

Generate Database ERD Diagram

Analyze the project's database models and schema files, then produce a Mermaid erDiagram saved to docs/schema.md (or the path in $ARGUMENTS if provided).

Step 1 — Detect ORM

Search for known config and schema files in this order:

Prisma:

Glob: **/*.prisma (exclude node_modules)

If found → ORM is Prisma.

Drizzle:

Glob: {src,db,database,lib}/**/{schema,drizzle}.{ts,js}

Check for imports of drizzle-orm, pgTable, mysqlTable, sqliteTable → ORM is Drizzle.

TypeORM:

Glob: {src,lib}/**/*.entity.{ts,js}

Check for @Entity() decorator → ORM is TypeORM.

Sequelize:

Glob: {src,lib,models}/**/*.{ts,js}

Check for DataTypes, Model.init, sequelize.define → ORM is Sequelize.

Mongoose:

Glob: {src,lib,models}/**/*.{ts,js}

Check for new Schema(, mongoose.model( → ORM is Mongoose.

Raw SQL:

Glob: **/*.{sql,migration.ts,migration.js}

If CREATE TABLE statements found → parse SQL directly.

If multiple ORMs detected, list all and process each.

Step 2 — Extract Models and Fields

Prisma

Read each .prisma file. For every model block extract:

• Model name
• Fields: name, type (String, Int, Boolean, DateTime, Float, Json)
• Relations: @relation fields — extract fields, references
• Optional (?) vs required fields
• @id, @unique, @@unique constraints

Drizzle

Read schema files. For every table definition extract:

• Table name (first argument to pgTable/mysqlTable/sqliteTable)
• Columns: name and type (varchar, integer, boolean, timestamp, text, json)
• .primaryKey(), .notNull(), .unique() modifiers
• references(() => otherTable.column) for foreign keys

TypeORM

Read each .entity.ts file. Extract:

• Class name (model name)
• @Column() fields with type
• @PrimaryGeneratedColumn() or @PrimaryColumn()
• @OneToMany, @ManyToOne, @OneToOne, @ManyToMany relations and their targets