API Design Skill

API Design Skill

Design APIs that developers love to use.

Core Principle

A good API is intuitive, consistent, and hard to misuse. Design for the consumer, not the implementation.

REST Fundamentals

Resource Naming

HTTP Methods

*PATCH can be idempotent if designed carefully

Status Codes

Contract-First Design

The Process

1. Define the contract (OpenAPI/Swagger)
2. Review with consumers before coding
3. Generate server stubs from contract
4. Implement business logic
5. Validate responses against contract

OpenAPI Skeleton

openapi: 3.0.3
info:
  title: My API
  version: 1.0.0
paths:
  /users:
    get:
      summary: List users
      parameters:
        - name: limit
          in: query
          schema:
            type: integer
            default: 20
      responses:
        '200':
          description: Success
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/UserList'
components:
  schemas:
    User:
      type: object
      required: [id, email]
      properties:
        id:
          type: string
        email:
          type: string
          format: email

Versioning Strategies

Recommendation: URL path for major versions. It's explicit and easy.

Version Lifecycle

1. Active: Current version, fully supported
2. Deprecated: Works but sunset announced
3. Sunset: Removed, returns 410 Gone

Pagination Patterns

Offset-Based

GET /users?offset=40&limit=20

{
  "data": [...],
  "pagination": {
    "offset": 40,
    "limit": 20,
    "total": 150
  }
}

• ✅ Simple, familiar
• ❌ Inconsistent on fast-changing data

Cursor-Based

GET /users?cursor=abc123&limit=20

{
  "data": [...],
  "pagination": {
    "next_cursor": "def456",
    "has_more": true
  }
}

• ✅ Consistent, performant
• ❌ No random access

Error Response Design

Consistent Structure

{
  "error": {
    "code": "VALIDATION_ERROR",
    "message": "Invalid request parameters",
    "details": [
      {
        "field": "email",
        "message": "Must be valid email format"
      }
    ],
    "request_id": "req_abc123"
  }
}

Error Principles

• Machine-readable code for programmatic handling
• Human-readable message for debugging
• Request ID for support correlation
• Never expose stack traces or internal details

Request/Response Design

Consistency Rules

• Use camelCase or snake_case—pick one, stick to it
• Timestamps in ISO 8601: 2026-02-01T14:30:00Z
• IDs as strings (future-proof for UUIDs)
• Envelope responses: { "data": ..., "meta": ... }

Partial Responses

GET /users/123?fields=id,name,email

Reduces payload, improves performance.

Bulk Operations

POST /users/bulk
{
  "operations": [
    { "method": "create", "data": {...} },
    { "method": "update", "id": "123", "data": {...} }
  ]
}

Caching Strategies

HTTP Cache Headers

Cache-Control Directives

# Public, cacheable for 1 hour
Cache-Control: public, max-age=3600

# Private (user-specific), cacheable for 5 minutes
Cache-Control: private, max-age=300

# No caching at all
Cache-Control: no-store

# Cache but revalidate every time
Cache-Control: no-cache

# Stale content OK while revalidating
Cache-Control: max-age=60, stale-while-revalidate=30

ETag Validation Flow

# First request
GET /users/123
→ 200 OK
   ETag: "v1-abc123"
   Cache-Control: private, max-age=60

# Subsequent request (after cache expires)
GET /users/123
If-None-Match: "v1-abc123"
→ 304 Not Modified (cache still valid)
   OR
→ 200 OK with new ETag (content changed)

Caching Decision Matrix

Application-Level Caching

// Cache key design
const cacheKey = `${resource}:${id}:${version}`;

// Cache-aside pattern
async function getUser(id: string) {
  const cached = await cache.get(`user:${id}`);
  if (cached) return cached;

  const user = await db.users.findById(id);
  await cache.set(`user:${id}`, user, { ttl: 300 });
  return user;
}

// Cache invalidation on write
async function updateUser(id: string, data: UserUpdate) {
  const user = await db.users.update(id, data);
  await cache.delete(`user:${id}`);
  await cache.delete(`users:list:*`); // Invalidate list caches
  return user;
}

Cache Invalidation Strategies

Rate Limiting

Why Rate Limit?

• Protect resources: Prevent server overload
• Ensure fairness: No single client monopolizes
• Cost control: Limit expensive operations
• Security: Mitigate DoS and brute force

Common Algorithms

Token Bucket Example

class TokenBucket {
  private tokens: number;
  private lastRefill: number;

  constructor(
    private capacity: number,      // Max burst size
    private refillRate: number,    // Tokens per second
  ) {
    this.tokens = capacity;
    this.lastRefill = Date.now();
  }

  consume(tokens: number = 1): boolean {
    this.refill();
    if (this.tokens >= tokens) {
      this.tokens -= tokens;
      return true;
    }
    return false;
  }

  private refill() {
    const now = Date.now();
    const elapsed = (now - this.lastRefill) / 1000;
    this.tokens = Math.min(
      this.capacity,
      this.tokens + elapsed * this.refillRate
    );
    this.lastRefill = now;
  }
}

Rate Limit Response Headers

HTTP/1.1 200 OK
X-RateLimit-Limit: 100
X-RateLimit-Remaining: 95
X-RateLimit-Reset: 1706792400

# When exceeded
HTTP/1.1 429 Too Many Requests
Retry-After: 30
X-RateLimit-Limit: 100
X-RateLimit-Remaining: 0
X-RateLimit-Reset: 1706792400

Rate Limit Tiers

Rate Limit Scopes

// Per-user limiting
const userKey = `ratelimit:user:${userId}`;

// Per-IP limiting (for anonymous)
const ipKey = `ratelimit:ip:${clientIp}`;

// Per-endpoint limiting
const endpointKey = `ratelimit:${method}:${path}`;

// Combined (most flexible)
const combinedKey = `ratelimit:${userId}:${method}:${path}`;

Client-Side Rate Limit Handling

async function apiCallWithRetry(request: Request): Promise<Response> {
  const response = await fetch(request);

  if (response.status === 429) {
    const retryAfter = parseInt(
      response.headers.get('Retry-After') || '60'
    );

    console.log(`Rate limited. Retrying in ${retryAfter}s`);
    await sleep(retryAfter * 1000);
    return apiCallWithRetry(request);
  }

  return response;
}

// Proactive rate limiting
class RateLimitedClient {
  private remaining: number = Infinity;
  private resetTime: number = 0;

  async request(url: string): Promise<Response> {
    // Wait if we know we're out of quota
    if (this.remaining <= 0 && Date.now() < this.resetTime) {
      await sleep(this.resetTime - Date.now());
    }

    const response = await fetch(url);

    // Update tracking from headers
    this.remaining = parseInt(
      response.headers.get('X-RateLimit-Remaining') || '100'
    );
    this.resetTime = parseInt(
      response.headers.get('X-RateLimit-Reset') || '0'
    ) * 1000;

    return response;
  }
}

API Gateway Rate Limiting

Most API gateways support rate limiting out of the box:

# AWS API Gateway
throttle:
  burstLimit: 100
  rateLimit: 50

# Azure API Management
<rate-limit calls="100" renewal-period="60" />

# Kong
plugins:
  - name: rate-limiting
    config:
      minute: 100
      policy: redis

Enterprise API Integration Patterns

Patterns from FishbowlGovernance multi-platform integrations (Feb 2026)

Microsoft Graph API (Planner)

Key Limitations:

ETag Pattern for Updates:

PATCH /planner/tasks/{task-id}/details
If-Match: {etag}
Content-Type: application/json

{"description": "Updated description"}

Azure DevOps REST API

JSON Patch Pattern:

PATCH /_apis/wit/workitems/{id}?api-version=7.0
Content-Type: application/json-patch+json

[
  {"op": "add", "path": "/fields/System.Title", "value": "New Title"},
  {"op": "replace", "path": "/fields/System.State", "value": "Active"}
]

Key Patterns:

• Use application/json-patch+json content type
• Implement idempotency with WIQL queries before insert
• Link work items with relation URLs, not IDs
• Batch updates sequentially (parallel causes conflicts)

WIQL Query for Duplicate Check:

POST /_apis/wit/wiql?api-version=7.0
Content-Type: application/json

{
  "query": "SELECT [System.Id] FROM WorkItems WHERE [System.Title] = 'Exact Title'"
}

Qualtrics API v3

Rate Limits by Endpoint:

Integration Patterns:

• Use webhook subscriptions for real-time events
• Poll distributions every 5-15 minutes
• Cache survey definitions (TTL: 1-4 hours)
• Flat JSON response structure (no nested objects)

Microsoft Fabric REST API

Async Operation Pattern (202 Accepted):

# 1. POST triggers async operation
$response = Invoke-WebRequest -Uri $uri -Headers $headers -Method Post

# 2. Get Location header for polling
if ($response.StatusCode -eq 202) {
    $operationUrl = $response.Headers["Location"][0]
    
    # 3. Poll until complete
    do {
        Start-Sleep -Seconds 2
        $status = Invoke-RestMethod -Uri $operationUrl -Headers $headers
    } while ($status.status -ne "Succeeded")
    
    # 4. Fetch result
    $result = Invoke-RestMethod -Uri "$operationUrl/result" -Headers $headers
}

Token Acquisition:

# Fabric API token
$fabricToken = az account get-access-token --resource https://api.fabric.microsoft.com --query accessToken -o tsv

# Storage token (for Unity Catalog/schema-enabled lakehouses)
$storageToken = az account get-access-token --resource https://storage.azure.com --query accessToken -o tsv

Multi-Platform Sync Architecture

%%{init: {'theme': 'base', 'themeVariables': {
  'primaryColor': '#cce5ff',
  'primaryTextColor': '#333',
  'primaryBorderColor': '#57606a',
  'lineColor': '#57606a',
  'secondaryColor': '#e6d5f2',
  'tertiaryColor': '#c2f0d8',
  'background': '#ffffff',
  'mainBkg': '#cce5ff',
  'secondBkg': '#e6d5f2',
  'tertiaryBkg': '#c2f0d8',
  'textColor': '#333',
  'border1Color': '#57606a',
  'border2Color': '#57606a',
  'arrowheadColor': '#57606a',
  'fontFamily': 'ui-sans-serif, system-ui, sans-serif',
  'fontSize': '14px',
  'nodeBorder': '1.5px',
  'clusterBkg': '#f6f8fa',
  'clusterBorder': '#d0d7de',
  'edgeLabelBackground': '#ffffff'
}}}%%
flowchart LR
    subgraph Sources["Source Systems"]
        ADO[Azure DevOps]
        QX[Qualtrics]
    end
    
    subgraph Integration["Integration Layer"]
        SB[Service Bus Queue]
        FN[Azure Functions]
        DB[(Cosmos DB<br/>Mappings)]
    end
    
    subgraph Targets["Target Systems"]
    PL[Microsoft Planner]
        FB[Microsoft Fabric]
    end
    
    ADO --> SB
    QX --> SB
    SB --> FN
    FN --> DB
    FN --> PL
    FN --> FB
    
    style Sources fill:#ddf4ff,color:#0550ae,stroke:#80ccff
    style Integration fill:#d8b9ff,color:#6639ba,stroke:#bf8aff
    style Targets fill:#d3f5db,color:#1a7f37,stroke:#6fdd8b

Key Design Principles:

1. Decouple with queues — Service Bus between source and target
2. Persist mappings — Cosmos DB for ID correlation
3. Timer renewals — Subscription expiry handled automatically
4. Accept limitations — Some features (Planner comments) simply not feasible

Security Checklist

□ Authentication on all non-public endpoints
□ Authorization checked for each resource
□ Rate limiting enabled
□ Input validation (size limits, type checking)
□ No sensitive data in URLs (tokens, passwords)
□ CORS configured appropriately
□ Security headers set
□ Audit logging for sensitive operations

Documentation Requirements

Good API docs include:

• Authentication: How to get and use tokens
• Quick Start: Working example in < 5 minutes
• Reference: Every endpoint, parameter, response
• Errors: All error codes and recovery steps
• Changelog: What changed in each version
• SDKs: Client libraries if available

API Review Checklist

Before shipping an API:

□ Resource names are nouns, plural
□ HTTP methods match semantics
□ Status codes are appropriate
□ Error responses are consistent
□ Pagination is implemented for lists
□ Versioning strategy is clear
□ Authentication is documented
□ Rate limits are defined
□ Breaking changes are versioned
□ OpenAPI spec is accurate

Synapses

See synapses.json for connections.