Phase 4: API Design/Implementation + Zero Script QA

Backend API implementation and script-free QA

Purpose

Implement backend APIs that can store and retrieve data. Validate with structured logs instead of test scripts.

What to Do in This Phase

1. API Design: Define endpoints, requests/responses
2. API Implementation: Write actual backend code
3. Zero Script QA: Log-based validation

Deliverables

docs/02-design/
└── api-spec.md             # API specification

src/api/                    # API implementation
├── routes/
├── controllers/
└── services/

docs/03-analysis/
└── api-qa.md               # QA results

PDCA Application

• Plan: Define required API list
• Design: Design endpoints, requests/responses
• Do: Implement APIs
• Check: Validate with Zero Script QA
• Act: Fix bugs and proceed to Phase 5

Level-wise Application

Level	Application Method
Starter	Skip this Phase (no API)
Dynamic	Use bkend.ai BaaS (see below)
Enterprise	Implement APIs directly

Dynamic Level: bkend.ai BaaS API Implementation

Step 1: MCP Setup

claude mcp add bkend --transport http https://api.bkend.ai/mcp

Step 2: Table Design (via MCP tools)

Natural language request: "Create a users table with name(required), email(required, unique), age fields" -> MCP backend_table_create auto-invoked

Step 3: Service API Integration

Method	Endpoint	Description
GET	/v1/data/{table}	List (filter, sort, page)
POST	/v1/data/{table}	Create data
GET	/v1/data/{table}/{id}	Get single
PATCH	/v1/data/{table}/{id}	Partial update
DELETE	/v1/data/{table}/{id}	Delete

Required Headers: x-project-id, x-environment, Authorization

Step 4: Auth Implementation

Reference MCP tools 3_howto_implement_auth and 6_code_examples_auth

Step 5: Zero Script QA

• Check bkend REST API call logs in browser DevTools Network tab
• Verify API behavior via response code/body

What is Zero Script QA?

Instead of writing test scripts, validate with structured debug logs

[API] POST /api/users
[INPUT] { "email": "test@test.com", "name": "Test" }
[PROCESS] Email duplicate check → Passed
[PROCESS] Password hash → Complete
[PROCESS] DB save → Success
[OUTPUT] { "id": 1, "email": "test@test.com" }
[RESULT] ✅ Success

Advantages:
- Save test code writing time
- See actual behavior with your eyes
- Easy debugging

RESTful API Principles

What is REST?

REpresentational State Transfer - an architecture style for designing web services.

6 Core Principles

Principle	Description	Example
1. Client-Server	Separation of concerns between client and server	UI ↔ Data storage separated
2. Stateless	Each request is independent, server doesn't store client state	Auth token included with each request
3. Cacheable	Responses must indicate if cacheable	Cache-Control header
4. Uniform Interface	Interact through consistent interface	Detailed below
5. Layered System	Allow layered system architecture	Load balancer, proxy
6. Code on Demand	(Optional) Server can send code to client	JavaScript delivery

Uniform Interface Details

The core of RESTful APIs is a uniform interface.

1. Resource-Based URLs

✅ Good (nouns, plural)
GET    /users          # User list
GET    /users/123      # Specific user
POST   /users          # Create user
PUT    /users/123      # Update user
DELETE /users/123      # Delete user

❌ Bad (using verbs)
GET    /getUsers
POST   /createUser
POST   /deleteUser/123

2. HTTP Method Meanings

Method	Purpose	Idempotent	Safe
GET	Read	✅	✅
POST	Create	❌	❌
PUT	Full update	✅	❌
PATCH	Partial update	❌	❌
DELETE	Delete	✅	❌

Idempotent: Same result even if requested multiple times Safe: Doesn't change server state

3. HTTP Status Codes

2xx Success
├── 200 OK              # Success (read, update)
├── 201 Created         # Creation success
└── 204 No Content      # Success but no response body (delete)

4xx Client Error
├── 400 Bad Request     # Invalid request (validation failure)
├── 401 Unauthorized    # Authentication required
├── 403 Forbidden       # No permission
├── 404 Not Found       # Resource not found
└── 409 Conflict        # Conflict (duplicate, etc.)

5xx Server Error
├── 500 Internal Error  # Internal server error
└── 503 Service Unavailable  # Service unavailable

4. Consistent Response Format

// Success response
{
  "data": {
    "id": 123,
    "email": "user@example.com",
    "name": "John Doe"
  },
  "meta": {
    "timestamp": "2026-01-08T10:00:00Z"
  }
}

// Error response
{
  "error": {
    "code": "VALIDATION_ERROR",
    "message": "Email format is invalid.",
    "details": [
      { "field": "email", "message": "Please enter a valid email" }
    ]
  }
}

// List response (pagination)
{
  "data": [...],
  "pagination": {
    "page": 1,
    "limit": 20,
    "total": 100,
    "totalPages": 5
  }
}

URL Design Rules

1. Use lowercase
   ✅ /users/123/orders
   ❌ /Users/123/Orders

2. Use hyphens (-), avoid underscores (_)
   ✅ /user-profiles
   ❌ /user_profiles

3. Express hierarchical relationships
   ✅ /users/123/orders/456

4. Filtering via query parameters
   ✅ /users?status=active&sort=created_at
   ❌ /users/active/sort/created_at

5. Version management
   ✅ /api/v1/users
   ✅ Header: Accept: application/vnd.api+json;version=1

API Documentation Tools

Tool	Features
OpenAPI (Swagger)	Industry standard, auto documentation
Postman	Testing + documentation
Insomnia	Lightweight API client

---

API Design Checklist

• RESTful Principles Compliance
  • Resource-based URLs (nouns, plural)
  • Appropriate HTTP methods
  • Correct status codes
• Unified error response format
• Authentication/authorization method defined
• Pagination method defined
• Versioning method (optional)

Templates

• templates/pipeline/phase-4-api.template.md
• templates/pipeline/zero-script-qa.template.md

Next Phase

Phase 5: Design System → APIs are ready, now build UI components