What is the Repository pattern, and how does it help decouple business logic from a specific database technology?

The Repository pattern defines an abstraction (an interface or a consistent set of functions) for accessing and persisting domain objects, hiding the specific database technology and query syntax behind that abstraction -- business logic calls methods like findById() or save() without knowing whether the underlying implementation uses MongoDB, PostgreSQL, or an in-memory store, which makes both testing (via a fake in-memory repository) and future database migrations significantly easier.

class UserRepository {
  async findById(id) { return db.collection('users').findOne({ _id: id }); }
  async save(user) { return db.collection('users').insertOne(user); }
}

// In tests, swap in a fake implementation with the same interface
class FakeUserRepository {
  constructor() { this.users = new Map(); }
  async findById(id) { return this.users.get(id); }
  async save(user) { this.users.set(user.id, user); return user; }
}