Architecture

Request Flow

Layered Design

The CRUD implementation follows a four-layer architecture:

1. Transport Layer (Hono)

• Location: features/crud/crud.routes.ts
• Purpose: Maps HTTP requests to CRUD operations using zValidator
• Generated endpoints: Standard REST-like POST endpoints per model

2. Service Layer (Effect)

• Location: features/crud/services/crud.service.ts
• Purpose: Orchestrates the Operation Pipeline
• Interface: execute<T>(model, op, params, context)

3. Logic Layer (Hooks)

• Location: features/crud/services/crud-hook.service.ts
• Purpose: Manages lifecycle hooks via hook-executor.ts
• Stages:
  • runBeforeHooks: Modify params before DB operation
  • runAfterSuccessHooks: Side-effects after successful write
  • runAfterFailureHooks: Error handling

4. Data Layer (Prisma)

• Location: features/crud/services/crud-repository.service.ts
• Purpose: Low-level database access
• Features:
  • Dynamic delegate selection: prisma[model]
  • Retry logic via withDbRetry
  • Error mapping to CrudError

Hook Registration

Hooks are registered using registerCrudHook in factory.ts:

import { registerCrudHook } from "./factory";

registerCrudHook(
  { model: "contact", op: "createOne" },
  {
    before: async (ctx) => {
      // Validate or enrich params
      return ctx.params;
    },
    after: async (ctx) => {
      // Side-effect after creation
    },
  }
);

Dependency Injection

Services are composed in layers.ts using Effect-TS:

export const CrudServiceLive = Layer.provide(
  CrudService.of({ ... }),
  [CrudRepositoryLive, CrudHookServiceLive, MetricsServiceLive]
);

Shared Pipeline

CRUD leverages the shared OperationPipeline from features/shared/operation-pipeline.ts, ensuring Dynamic Objects and other features share the same:

• Observability (metrics, tracing)
• Lifecycle hooks
• Error handling
• Retry logic