architect-pro（建筑师专业版）

Architect Pro: Advanced Software Architecture (2026 Edition)

Master the art of building maintainable, scalable, and testable systems using proven architectural patterns. This skill is optimized for the Next.js 16.2 / React 19.3 / Node.js 24 ecosystem, focusing on decoupling business logic from infrastructure and enabling Autonomous AI Agents to work effectively within your codebase.

🎓 Prerequisites & Skill Level

• Level: Senior / Lead Engineer / Software Architect
• Prerequisites:

- Deep understanding of TypeScript 5.5+ (Generics, Discriminated Unions, Template Literal Types). - Familiarity with Next.js App Router and Server Components. - Basic knowledge of SQL (PostgreSQL) and NoSQL (Redis/MongoDB) databases. - Experience with automated testing (Jest, Vitest, Playwright). - Interest in Domain-Driven Design (DDD).

📌 Table of Contents

1. Core Philosophy
2. Why Clean Architecture in 2026?
3. Architecture for AI Agents
4. Quick Start: The 2026 Stack
5. Modern Data Flow: React 19 + Next.js 16
6. Standard Pattern: Clean Architecture + Server Actions
7. Tactical DDD: Entities, Value Objects, and Aggregates
8. Domain Services vs Application Services
9. Advanced Pattern: Event-Driven DDD
10. Event Storming & Domain Modeling
11. Architecture Decision Records (ADR)
12. Context Mapping & Strategic DDD
13. Testing Strategy for Architects
14. Architectural Enforcement & Guardrails
15. Performance & Scalability at the Architectural Level
16. Glossary of Architectural Terms
17. The Do Not List (Anti-Patterns)
18. Deep-Dive References
19. Repository Analysis with Repomix

🏛️ Core Philosophy

In 2026, software is increasingly built and refactored by AI Coding Agents. For an agent to be effective, the codebase must have predictable boundaries and explicit contracts.

• Decoupling: Business logic (Domain) must not depend on the database, UI, or frameworks.
• Testability: You should be able to test your core logic without spinning up a database or mocking complex framework internals.
• Independence: The system should be agnostic of its delivery mechanisms (Web, CLI, Mobile, AI Agents).
• Predictability: Files and folders should follow a strict convention so that both humans and AI can locate logic instantly.

🧠 Why Clean Architecture in 2026?

The "Move Fast and Break Things" era has evolved into the "Move Fast with AI and Maintain Sanity" era. Without a solid architecture:

1. AI Agents Hallucinate: When business logic is mixed with Prisma queries and React hooks, AI often loses context and introduces subtle bugs.
2. Framework Churn: Next.js and React versions move fast. A clean domain layer protects you from the next breaking change in the app router or server actions.
3. Complexity Debt: What starts as a simple "CRUD" app often grows into a complex domain. Clean Architecture allows for "Gradual Complexity" where you can add layers only when needed.

🤖 Architecture for AI Agents

In 2026, we design for two audiences: Humans and Agents.

• Deterministic File Paths: Using a strict Clean Architecture folder structure helps AI find the "Source of Truth" (Domain) instantly.
• Strong Typing (TS 5.5+): Explicit return types in Use Cases act as "API Documentation" for AI agents.
• Small Context Windows: Decoupled layers allow AI to work on a single file (like a Use Case) without needing to understand the entire database schema.
• Repomix Context Packing: We use tools to pack high-density context for AI agents, and a clean architecture makes this packing 10x more efficient.

⚡ Quick Start: The 2026 Stack

The most common production stack in 2026 uses Next.js 16.2 with Server Functions, TypeScript 5.x, and Bun/pnpm monorepos.

Recommended Folder Structure

src/
├── domain/           # Core Logic (The "Truth")
│   ├── entities/     # Identifiable objects with identity and lifecycle
│   ├── value-objects/# Data wrappers with validation (Email, Money)
│   ├── services/     # Pure logic involving multiple entities
│   ├── events/       # Domain Events (OrderPlaced, UserDeactivated)
│   └── exceptions/   # Custom domain exceptions
├── application/      # Orchestration (The "User's Intent")
│   ├── use-cases/    # Specific business actions (Commands)
│   ├── ports/        # Interfaces (contracts for Infrastructure)
│   ├── dtos/         # Data Transfer Objects
│   └── queries/      # Read-only operations (if using CQRS)
├── infrastructure/   # Tools (The "Implementation")
│   ├── adapters/     # Implementations of Ports (Prisma, Stripe, AWS)
│   ├── persistence/  # DB Schemas, Migrations, Seeders
│   └── external/     # Third-party API clients
├── presentation/     # UI (The "View")
│   ├── components/   # React 19 Server/Client components
│   ├── actions/      # Next.js Server Actions (Primary Adapters)
│   └── pages/        # Route Handlers and Page definitions
└── shared/           # Cross-cutting concerns (Constants, Utils)

🌊 Modern Data Flow: React 19 + Next.js 16

The interaction model has changed. We no longer just "fetch from API". We "Invoke a Service".

1. View (React 19)

Uses the new useActionState and useOptimistic for a seamless UX.

// presentation/components/ProjectForm.tsx
'use client'

import { useActionState } from 'react';
import { createProjectAction } from '../actions/project-actions';

export function ProjectForm() {
  const [state, formAction, isPending] = useActionState(createProjectAction, null);

  return (
    <form action={formAction}>
      <div className="flex flex-col gap-4">
        <label htmlFor="projectName">Project Name</label>
        <input
          id="projectName"
          name="projectName"
          className="border p-2 rounded"
          required
        />
        <button
          className="bg-blue-600 text-white p-2 rounded"
          disabled={isPending}
        >
          {isPending ? 'Creating...' : 'Create Project'}
        </button>
        {state?.error && <p className="text-red-500">{state.error}</p>}
      </div>
    </form>
  );
}

2. Action (Next.js 16.2)

Acts as the Primary Adapter. It handles the transition from HTTP (FormData) to Application logic.

// presentation/actions/project-actions.ts
'use server'

import { CreateProjectUseCase } from '@/application/use-cases/CreateProject';
import { PrismaProjectRepository } from '@/infrastructure/adapters/PrismaProjectRepository';

export async function createProjectAction(prevState: any, formData: FormData) {
  const name = formData.get('projectName') as string;

  // Dependency Injection: In 2026, we prefer explicit composition over complex DI containers
  const repo = new PrismaProjectRepository();
  const useCase = new CreateProjectUseCase(repo);

  try {
    const project = await useCase.execute({ name });
    return { success: true, data: project };
  } catch (e) {
    // Audit log or Error tracking integration
    console.error("[Action Error]", e);
    return { success: false, error: e.message };
  }
}

🛠️ Standard Pattern: Clean Architecture + Server Actions

1. The Domain Entity

Entities represent business concepts with identity.

// src/domain/entities/User.ts
import { Email } from '../value-objects/Email';

export class User {
  constructor(
    public readonly id: string,
    public readonly email: Email,
    private _isActive: boolean = true
  ) {}

  public deactivate() {
    this._isActive = false;
  }

  get isActive() { return this._isActive; }
}

2. The Use Case (Application Layer)

Use Cases orchestrate the domain objects to perform a specific task.

// src/application/use-cases/DeactivateUser.ts
import { IUserRepository } from '../ports/IUserRepository';

export class DeactivateUserUseCase {
  constructor(private userRepo: IUserRepository) {}

  async execute(userId: string): Promise<void> {
    // 1. Fetch from repository (Port)
    const user = await this.userRepo.findById(userId);
    if (!user) throw new Error("User not found");

    // 2. Perform domain logic (Entity)
    user.deactivate();

    // 3. Persist changes (Port)
    await this.userRepo.save(user);
  }
}

3. The Repository Implementation (Infrastructure)

The "Adapter" that talks to the database.

// src/infrastructure/adapters/PrismaUserRepository.ts
import { IUserRepository } from '@/application/ports/IUserRepository';
import { User } from '@/domain/entities/User';
import { Email } from '@/domain/value-objects/Email';
import { prisma } from '../prisma';

export class PrismaUserRepository implements IUserRepository {
  async findById(id: string): Promise<User | null> {
    const data = await prisma.user.findUnique({ where: { id } });
    if (!data) return null;

    return new User(
      data.id,
      Email.create(data.email),
      data.isActive
    );
  }

  async save(user: User): Promise<void> {
    await prisma.user.upsert({
      where: { id: user.id },
      update: { isActive: user.isActive },
      create: {
        id: user.id,
        email: user.email.value,
        isActive: user.isActive
      }
    });
  }
}

🧩 Tactical DDD: Entities, Value Objects, and Aggregates

Value Objects (The Power of Types)

Don't use string for everything. Use Value Objects for validation and business logic.

// domain/value-objects/Money.ts
export class Money {
  private constructor(public readonly amount: number, public readonly currency: string) {}

  static create(amount: number, currency: string = 'USD'): Money {
    if (amount < 0) throw new Error("Amount cannot be negative");
    return new Money(amount, currency);
  }

  add(other: Money): Money {
    if (this.currency !== other.currency) throw new Error("Currency mismatch");
    return new Money(this.amount + other.amount, this.currency);
  }
}

Aggregates (The Consistency Boundary)

An Aggregate ensures that business rules are always met across multiple related objects.

// domain/entities/Order.ts
export class Order {
  private items: OrderItem[] = [];

  addItem(product: Product, quantity: number) {
    const totalItems = this.items.reduce((sum, item) => sum + item.quantity, 0);
    if (totalItems + quantity > 50) {
      throw new Error("An order cannot have more than 50 items");
    }
    this.items.push(new OrderItem(product.id, quantity));
  }
}

⚙️ Domain Services vs Application Services

One of the most common points of confusion in DDD.

• Domain Service: Contains business logic that belongs to the domain but doesn't fit in a single entity (e.g., TransferService between two bank accounts). It works with entities and value objects.
• Application Service (Use Case): Contains NO business logic. It only orchestrates. It calls repositories, triggers domain services, and sends emails.

🚀 Advanced Pattern: Event-Driven DDD

Capture intent and decouple side-effects.

Domain Service Example

// domain/services/TransferService.ts
export class TransferService {
  async execute(from: Account, to: Account, amount: Money) {
    from.withdraw(amount);
    to.deposit(amount);

    // Domain Events are captured here
    return new FundsTransferredEvent(from.id, to.id, amount);
  }
}

🌪️ Event Storming & Domain Modeling

In 2026, we don't start with "Tables". We start with "Events".

1. Domain Events: What happens in the business? (e.g., OrderPlaced, InventoryReduced).
2. Commands: What triggers these events? (e.g., PlaceOrder).
3. Aggregates: What objects are involved in making sure the rules are followed?
4. Read Models: How do we want to display this information to the user?

📝 Architecture Decision Records (ADR)

In 2026, we record *why* we made a choice, not just *what* we chose.

# ADR 005: Use Clean Architecture for AI Core

## Status
Accepted

## Context
We need a way for AI agents to refactor code safely without breaking business rules. The previous "Spaghetti Monolith" resulted in AI hallucinating database queries in React components.

## Decision
We will use Clean Architecture with strict separation between Domain and Infrastructure.

## Consequences
- Pros: Predictable file structure, high testability, zero framework leakage in domain.
- Cons: More boilerplate for simple CRUD operations, higher learning curve for junior devs.

🗺️ Context Mapping & Strategic DDD

How different parts of the system talk to each other.

• Customer/Supplier: One team depends on another.
• Shared Kernel: Two teams share a common model (rarely recommended).
• Anticorruption Layer (ACL): A layer that translates between a legacy system and your clean new system.
• Separate Ways: No relationship between contexts.

🧪 Testing Strategy for Architects

In a Clean Architecture, you have different "Testing Grooves":

1. Domain Unit Tests: Test entities and value objects. ZERO mocks needed. 100% code coverage expected here.
2. Use Case Tests: Test orchestration. Use In-Memory Adapters (Mocks/Fakes) for the database.
3. Integration Tests: Test the real adapters (e.g., Prisma repository against a real Test Container).
4. E2E Tests: Test the full Next.js flow using Playwright.

Example: Use Case Test (No DB)

// tests/use-cases/DeactivateUser.test.ts
import { DeactivateUserUseCase } from '@/application/use-cases/DeactivateUser';
import { InMemoryUserRepository } from '../mocks/InMemoryUserRepository';

test('should deactivate user', async () => {
  const repo = new InMemoryUserRepository();
  await repo.save({ id: '1', email: 'test@test.com', isActive: true });

  const useCase = new DeactivateUserUseCase(repo);
  await useCase.execute('1');

  const user = await repo.findById('1');
  expect(user.isActive).toBe(false);
});

🛡️ Architectural Enforcement & Guardrails

Use tools to ensure the "Dependency Rule" is never broken.

dependency-cruiser config (2026 best practice)

{
  "forbidden": [
    {
      "name": "domain-not-import-infra",
      "from": { "path": "src/domain" },
      "to": { "path": "src/infrastructure" }
    },
    {
      "name": "usecase-not-import-infra",
      "from": { "path": "src/application" },
      "to": { "path": "src/infrastructure" }
    },
    {
      "name": "presentation-not-import-infra",
      "from": { "path": "src/presentation" },
      "to": { "path": "src/infrastructure" }
    }
  ]
}

📈 Performance & Scalability at the Architectural Level

• Read-Write Segregation (CQRS): Use a fast NoSQL database for reads and a robust SQL database for writes.
• Caching as a First-Class Citizen: Don't put cache logic in your Use Case. Use a Caching Adapter or the Next.js use cache directive in the presentation layer.
• Job Queues: Offload heavy work (Email, AI Processing) to background jobs via Domain Events.

📖 Glossary of Architectural Terms

• Bounded Context: A specific boundary (usually a module or service) where a domain model is consistent.
• Port: An interface defined by the application layer that infrastructure must implement.
• Adapter: An implementation of a port (e.g., a SQL database implementation of an IUserRepository).
• Inversion of Control (IoC): The principle where the domain defines its needs (Ports), and the high-level layers provide the implementations (Adapters).
• Aggregate Root: The main entity in an aggregate that controls all access and enforces business rules.
• Ubiquitous Language: A common language used by all team members to describe the domain.

🚫 The Do Not List (Anti-Patterns)

1. Framework Leakage

BAD: Using @prisma/client types as return types in your Domain or Application layers. GOOD: Map database rows to Domain Entities in the Infrastructure layer.

2. Anemic Domain Model

BAD: Putting all your logic in UserService and having User as just a type. GOOD: Move business rules (e.g., user.canJoinProject()) into the User class.

3. The "Shared" Dumpster

BAD: Putting everything in shared/ because you are too lazy to decide where it belongs. GOOD: Keep shared/ minimal. If it's business-related, it's domain/. If it's tool-related, it's infrastructure/.

4. Premature Microservices

BAD: Splitting into 10 repos before you understand the domain boundaries. GOOD: Build a Modular Monolith using Clean Architecture first. Boundaries are easy to cut later if they are clean.

5. Lazy Naming

BAD: data, info, manager, process. GOOD: PaymentRecord, UserAuthenticationDetails, OrderOrchestrator.

📚 Deep-Dive References

Explore our specialized documentation for deeper technical insights:

• Clean Architecture Guide
• Hexagonal Architecture (Ports & Adapters)
• DDD Patterns & Tactical Design
• Performance & Scalability 2026
• Migration & Refactoring Strategies
• Repository Analysis Guide

🔍 Repository Analysis with Repomix

Before refactoring, always analyze the existing codebase structure.

# Optimized command for architecture discovery
npx repomix@latest --include "src/**/*.ts" --compress --output /tmp/arch-discovery.xml

Using Repomix allows you to "pack" your architecture into a single file that can be fed into an AI for a full audit.

💎 Best Practices Checklist

• Is my Domain Layer free of external dependencies?
• Are my Use Cases doing only orchestration?
• Do I have interfaces (Ports) for all external services (DB, Auth, Mail, Stripe)?
• Am I using Value Objects for things like Email, Money, and Address?
• Is my code testable without a real database?
• Have I mapped my infrastructure errors to domain exceptions?
• Have I used ADRs to document major architectural decisions?
• Is my ubiquitous language reflected in the code?

*Updated: January 22, 2026 - 15:18*