atelier-typescript-api-design
API Design Patterns
Best practices for designing REST APIs with consistent structure, error handling, and resource patterns.
Additional References
- references/error-responses.md - Detailed error handling examples
Resource Naming
Use consistent, predictable URL patterns:
# Collection resources (plural nouns)
GET /api/v1/users # List users
POST /api/v1/users # Create user
GET /api/v1/users/:id # Get user
PUT /api/v1/users/:id # Update user (full)
PATCH /api/v1/users/:id # Update user (partial)
DELETE /api/v1/users/:id # Delete user
# Nested resources
GET /api/v1/users/:userId/posts # List user's posts
POST /api/v1/users/:userId/posts # Create post for user
GET /api/v1/users/:userId/posts/:postId # Get specific post
# Actions (use verbs sparingly)
POST /api/v1/users/:id/activate # Activate user
POST /api/v1/posts/:id/publish # Publish post
POST /api/v1/invoices/:id/send # Send invoice
Guidelines
- Use plural nouns for collections (
/users, not/user) - Use lowercase with hyphens for multi-word resources (
/ledger-accounts) - Avoid deep nesting (max 2 levels:
/users/:id/posts/:id) - Use query parameters for filtering, sorting, pagination
- Use verbs only for actions that don't fit CRUD (activate, publish, send)
API Versioning
Version APIs in the URL path:
/api/v1/users
/api/v2/users
# Not in headers (harder to test/debug)
# Not in query params (breaks caching)
Version Strategy
// v1/routes.ts
export async function v1Routes(app: FastifyInstance) {
app.get('/users', getUsersV1)
app.post('/users', createUserV1)
}
// v2/routes.ts
export async function v2Routes(app: FastifyInstance) {
app.get('/users', getUsersV2) // Breaking change in response structure
app.post('/users', createUserV2)
}
// server.ts
app.register(v1Routes, { prefix: '/api/v1' })
app.register(v2Routes, { prefix: '/api/v2' })
RFC 7807 Problem Details
Standardized error response format:
interface ProblemDetail {
type: string // Error type identifier
status: number // HTTP status code
title: string // Short, human-readable summary
detail: string // Specific explanation for this occurrence
instance: string // URI reference to specific occurrence
traceId: string // Request trace ID for debugging
}
// Example error response
{
"type": "NOT_FOUND",
"status": 404,
"title": "Not Found",
"detail": "User with ID usr_01h455vb4pex5vsknk084sn02q not found",
"instance": "/api/v1/users/usr_01h455vb4pex5vsknk084sn02q",
"traceId": "req_abc123xyz"
}
Error Types
// Domain error base class
abstract class AppError extends Error {
abstract readonly status: number
abstract readonly type: string
constructor(message: string, public readonly context?: ErrorContext) {
super(message)
}
toResponse(instance: string, traceId: string): ProblemDetail {
return {
type: this.type,
status: this.status,
title: this.name,
detail: this.message,
instance,
traceId,
...this.context,
}
}
}
// Specific error types
class NotFoundError extends AppError {
readonly status = 404
readonly type = 'NOT_FOUND'
}
class ConflictError extends AppError {
readonly status = 409
readonly type = 'CONFLICT'
constructor(
message: string,
public readonly retryable: boolean = false,
context?: ErrorContext
) {
super(message, context)
}
}
class ServiceUnavailableError extends AppError {
readonly status = 503
readonly type = 'SERVICE_UNAVAILABLE'
constructor(
message: string,
public readonly retryable: boolean = true,
context?: ErrorContext
) {
super(message, context)
}
}
See references/error-responses.md for complete examples.
Pagination (Cursor-Based)
Use cursor-based pagination for large datasets:
// Request
GET /api/v1/posts?limit=20&cursor=pst_01h455vb4pex5vsknk084sn02q
// Response
{
"items": [
{ "id": "pst_01h455w3x8k5z9y7q1m0n2b3c4", ... },
{ "id": "pst_01h455x2y9l6a0z8r2n1o3c5d6", ... }
],
"nextCursor": "pst_01h455z1a0m7b8y9s3o2p4d6e7",
"hasMore": true
}
Implementation
interface PaginatedRequest {
limit?: number // Max items to return (default 20, max 100)
cursor?: string // Cursor for next page (opaque to client)
}
interface PaginatedResponse<T> {
items: T[]
nextCursor?: string
hasMore: boolean
}
async function listPosts(req: PaginatedRequest): Promise<PaginatedResponse<Post>> {
const limit = Math.min(req.limit ?? 20, 100)
const queryLimit = limit + 1 // Fetch one extra to check hasMore
const posts = await db.query.posts.findMany({
where: req.cursor ? gt(posts.id, req.cursor) : undefined,
orderBy: desc(posts.createdAt),
limit: queryLimit,
})
const hasMore = posts.length > limit
const items = posts.slice(0, limit)
const nextCursor = hasMore ? items[items.length - 1].id : undefined
return { items, nextCursor, hasMore }
}
Why Cursor Over Offset
❌ Offset-based (/posts?offset=40&limit=20)
- Unstable: Items can shift if new records inserted
- Performance: DB must scan all previous rows
- Inaccurate: Can miss or duplicate items
✅ Cursor-based (/posts?cursor=pst_xyz&limit=20)
- Stable: Cursor points to specific item
- Performant: DB uses index seek
- Accurate: No gaps or duplicates
Filtering & Sorting
Use query parameters for filtering and sorting:
# Filtering
GET /api/v1/users?status=active&role=admin
GET /api/v1/posts?author=usr_abc&published=true
# Sorting
GET /api/v1/posts?sort=-createdAt # Descending (- prefix)
GET /api/v1/users?sort=name # Ascending
# Combined
GET /api/v1/posts?author=usr_abc&status=published&sort=-createdAt&limit=20
Implementation
interface ListPostsQuery {
author?: string
status?: 'draft' | 'published'
sort?: 'createdAt' | '-createdAt' | 'title' | '-title'
limit?: number
cursor?: string
}
async function listPosts(query: ListPostsQuery): Promise<PaginatedResponse<Post>> {
const conditions = []
if (query.author) {
conditions.push(eq(posts.authorId, query.author))
}
if (query.status) {
conditions.push(eq(posts.status, query.status))
}
const orderByColumn = query.sort?.startsWith('-')
? query.sort.slice(1)
: query.sort ?? 'createdAt'
const orderByDirection = query.sort?.startsWith('-') ? desc : asc
return await db.query.posts.findMany({
where: conditions.length > 0 ? and(...conditions) : undefined,
orderBy: orderByDirection(posts[orderByColumn]),
limit: query.limit ?? 20,
})
}
HTTP Status Codes
Use status codes consistently:
# Success
200 OK # Successful GET, PUT, PATCH
201 Created # Successful POST (include Location header)
204 No Content # Successful DELETE, PUT with no response body
# Client Errors
400 Bad Request # Invalid request body/parameters
401 Unauthorized # Missing or invalid authentication
403 Forbidden # Valid auth, but lacks permission
404 Not Found # Resource doesn't exist
409 Conflict # Resource already exists, optimistic lock failure
422 Unprocessable # Validation error (semantic)
429 Too Many Requests # Rate limit exceeded
# Server Errors
500 Internal Server Error # Unexpected error
503 Service Unavailable # Temporary unavailability, retry later
Response Envelope (When to Use)
Don't use envelopes for simple CRUD:
// ❌ Unnecessary wrapping
GET /api/v1/users/123
{
"success": true,
"data": { "id": "123", "name": "Alice" }
}
// ✅ Return resource directly
GET /api/v1/users/123
{
"id": "123",
"name": "Alice"
}
Use envelopes for pagination:
// ✅ Envelope needed for metadata
GET /api/v1/users?limit=20
{
"items": [...],
"nextCursor": "usr_xyz",
"hasMore": true
}
Timestamps
Use ISO 8601 format for all timestamps:
{
"createdAt": "2024-01-15T14:30:00.000Z", // ISO 8601 UTC
"updatedAt": "2024-01-16T09:15:30.123Z"
}
// In entities
toResponse(): UserResponse {
return {
...
createdAt: this.createdAt.toISOString(), // Date → ISO string
updatedAt: this.updatedAt.toISOString(),
}
}
Idempotency
Use idempotency keys for safe retries:
// Request
POST /api/v1/transactions
Headers:
Idempotency-Key: txn_abc123xyz
Body:
{ "amount": 100, "from": "usr_123", "to": "usr_456" }
// Implementation
async function createTransaction(rq: CreateTransactionRequest, idempotencyKey: string) {
// Check if transaction with this key already exists
const existing = await db.query.transactions.findFirst({
where: eq(transactions.idempotencyKey, idempotencyKey),
})
if (existing) {
return TransactionEntity.fromRecord(existing) // Return existing
}
// Create new transaction
const transaction = TransactionEntity.fromRequest(rq, idempotencyKey)
return await transactionRepo.create(transaction)
}
Guidelines
- Plural nouns - Collections use plural resource names
- Lowercase with hyphens - Multi-word resources like
ledger-accounts - Version in URL -
/api/v1/,/api/v2/for breaking changes - RFC 7807 errors - Standardized error response format
- Cursor pagination - For large datasets (more stable than offset)
- Query params - For filtering, sorting, pagination (not in path)
- HTTP status codes - Use correct codes (200, 201, 204, 400, 404, 409, 500, 503)
- ISO 8601 timestamps - Always use
.toISOString()for dates - Idempotency keys - For non-idempotent operations (POST, PATCH)
- No unnecessary envelopes - Return resources directly unless pagination needed
More from martinffx/claude-code-atelier
python:architecture
Python application architecture with functional core, effectful shell, DDD, and data modeling. Use when designing application layers, separating pure business logic from IO, defining domain models, implementing validation, or structuring bounded contexts.
14python:sqlalchemy
SQLAlchemy ORM patterns for Python database access. Use when defining models, writing queries, implementing upserts, working with JSON columns, or managing database sessions.
12python:fastapi
Building REST APIs with FastAPI, Pydantic validation, and OpenAPI. Use when creating routes, handling requests, designing endpoints, implementing validation, error responses, pagination, or generating API documentation.
11atelier-typescript-drizzle-orm
Type-safe SQL with Drizzle ORM in TypeScript. Use when defining database schemas, writing queries, setting up relations, running migrations, or working with PostgreSQL/MySQL/SQLite/Cloudflare D1/Durable Objects data layers.
10python:temporal
Temporal workflow orchestration in Python. Use when designing workflows, implementing activities, handling retries, managing workflow state, or building durable distributed systems.
9atelier-typescript-functional-patterns
Functional programming patterns for reliable TypeScript. Use when modeling state machines, discriminated unions, Result/Option types, branded types, or building type-safe domain models.
8