The problem: In a typical GraphQL API, your resolver computes all fields for an object, even if the client only asked for 2 out of 15. This means unnecessary database calls, API requests, and CPU time — all thrown away.

The solution: LazyQL uses JavaScript Proxy to intercept field access. You write a class with getter methods (getStatus(), getCustomerEmail(), etc.), and LazyQL ensures each getter only runs when GraphQL actually reads that field.

  @LazyQL(OrderDTO)
  class Order {
    constructor(private id: number, private db: Database) {}

    getEntityId() { return this.id; }
    getStatus() { return this.db.getOrderStatus(this.id); }
    async getCustomerEmail() { return this.db.getCustomerEmail(this.id); }
    async getShippingAddress() { return this.db.getShippingAddress(this.id); }
  }

  // Query { entity_id, status } → only getEntityId() and getStatus() run
  // getCustomerEmail() and getShippingAddress() never execute

Key design decisions: - Works transparently with Apollo, Mercurius, or any GraphQL server — they don't know LazyQL exists - Naming convention maps snake_case DTO fields to getCamelCase methods automatically - @Shared() decorator caches expensive operations that multiple getters depend on - Validates at startup that all required DTO fields have matching getters (fail fast) - Zero runtime dependencies beyond reflect-metadata

It's been running in production for a few weeks now with zero issues. The whole point was to build something invisible — it sits there, does its job, and doesn't interfere with anything.

~400 lines of code, MIT licensed. Would love to hear your thoughts.