API Design Principles

Source: https://github.com/aipoch/medical-research-skills

API Design Principles

When to Use

• Designing a new REST API contract for CRUD-style resources and you need consistent resource modeling, naming, and HTTP semantics.
• Designing a new GraphQL schema for multiple clients with different data shapes and you need clear type/field ownership and safe evolution.
• Reviewing an existing API (REST or GraphQL) to identify inconsistencies in naming, error handling, pagination/filtering, or versioning/deprecation.
• Deciding between REST vs. GraphQL (or defining boundaries when mixing both) and documenting trade-offs and constraints.
• Standardizing cross-cutting concerns (authn/authz, rate limiting, observability, long-running operations, idempotency) across multiple services.

Key Features

• End-to-end workflow for API design/review: requirements → style choice → domain modeling → operations → cross-cutting concerns → deliverables.
• REST guidance: resource-oriented modeling, stable identifiers, relationship patterns, and correct HTTP verb usage.
• GraphQL guidance: schema/type modeling, Query vs. Mutation separation, input types for writes, and explicit side-effect handling.
• Cross-cutting design patterns: consistent error model, pagination/filtering/sorting, versioning and deprecation strategy, and operational concerns.
• Review checklist to validate completeness, highlight risks/gaps, and produce actionable follow-ups.

Dependencies

• None (documentation-only skill).
• Reference documents:
  • references/rest.md
  • references/graphql.md
  • references/review-checklist.md

Example Usage

Goal

Design (or review) an API for managing Projects and Tasks, and produce a contract with examples, error model, pagination, and a checklist summary.

Step 1: Clarify requirements and constraints

• Consumers: Web app + mobile app + internal admin.
• Constraints: p95 latency < 200ms for list endpoints; PII present; audit logging required.
• Core use cases: list projects, view project, create task, update task status, search tasks by status/assignee.

Step 2: Choose API style and boundaries

• Choose REST for resource-oriented CRUD with cacheable reads and straightforward endpoints.
• If GraphQL is later introduced for client-specific views, define boundaries (e.g., GraphQL for read aggregation; REST remains source-of-truth for writes).

Step 3: Produce a REST contract skeleton (runnable examples)

Base URL

• https://api.example.com/v1

Resources

• projects
• tasks (scoped under a project)

Endpoints

• GET /v1/projects
• POST /v1/projects
• GET /v1/projects/{projectId}
• GET /v1/projects/{projectId}/tasks
• POST /v1/projects/{projectId}/tasks
• PATCH /v1/projects/{projectId}/tasks/{taskId}

List projects (pagination + filtering)

Request

curl -sS -X GET "https://api.example.com/v1/projects?limit=20&cursor=eyJpZCI6IjEwMCJ9&sort=createdAt:desc" \
  -H "Authorization: Bearer $TOKEN" \
  -H "Accept: application/json"

Response (200)

{
  "data": [
    {
      "id": "proj_123",
      "name": "Roadmap 2026",
      "createdAt": "2026-01-10T12:00:00Z"
    }
  ],
  "page": {
    "limit": 20,
    "nextCursor": "eyJpZCI6InByb2pfMTIzIn0="
  }
}

Create a task (idempotency)

Request

curl -sS -X POST "https://api.example.com/v1/projects/proj_123/tasks" \
  -H "Authorization: Bearer $TOKEN" \
  -H "Idempotency-Key: 2b7b1a2e-7f2b-4c2a-9c2b-0b3b7c9d1a11" \
  -H "Content-Type: application/json" \
  -d '{
    "title": "Draft API spec",
    "assigneeId": "user_42",
    "dueAt": "2026-03-01T00:00:00Z"
  }'

Response (201)

{
  "data": {
    "id": "task_999",
    "projectId": "proj_123",
    "title": "Draft API spec",
    "status": "OPEN",
    "assigneeId": "user_42",
    "dueAt": "2026-03-01T00:00:00Z",
    "createdAt": "2026-02-25T09:00:00Z"
  }
}

Error model example

Response (409)

{
  "error": {
    "code": "CONFLICT",
    "message": "A task with the same title already exists in this project.",
    "details": {
      "field": "title",
      "reason": "DUPLICATE"
    },
    "requestId": "req_01HTZQ8K7Y9M2A3B4C5D6E7F8G"
  }
}

Step 4: Run the review checklist

Use references/review-checklist.md to validate:

• Naming consistency (resources, fields, enums)
• HTTP semantics and status codes
• Pagination/filtering/sorting rules
• Error model completeness and stability
• Versioning/deprecation plan
• Security and observability requirements

Expected deliverable format (save to outputs/)

• API style choice + trade-offs
• Contract skeleton (endpoints or schema)
• Request/response (or query/mutation) examples
• Error model + pagination strategy
• Checklist results + risks/gaps

Implementation Details

Recommended workflow (design/review)

1. Clarify requirements and constraints

   • Identify domain, core use cases, and consumer types (web/mobile/partners/internal).
   • Capture constraints: latency, throughput, consistency, compliance, data sensitivity.

2. Choose API style and boundaries

   • REST: best for resource-oriented APIs, cacheable reads, and simple CRUD.
   • GraphQL: best for multiple clients with varying data shapes and frequent iteration.
   • If mixing, define boundaries to avoid overlapping responsibilities.

3. Domain modeling

   • REST: model stable resources (nouns), stable identifiers, and relationships.
   • GraphQL: define types and field ownership; use input types for writes.

4. Operation and behavior design

   • REST: map operations to HTTP verbs; represent actions via sub-resources or noun-based endpoints when needed.
   • GraphQL: separate Query vs. Mutation; document side effects explicitly.
   • Define idempotency (especially for creates) and patterns for long-running tasks when applicable.

5. Cross-cutting concerns

   • Authentication/authorization
   • Error model (stable codes, actionable messages, request correlation IDs)
   • Pagination, filtering, sorting (document defaults and limits)
   • Versioning and deprecation strategy
   • Observability (logging/metrics/tracing), rate limiting

Reference guides

• REST Principles and Patterns: references/rest.md
• GraphQL Principles and Patterns: references/graphql.md
• Review Checklist: references/review-checklist.md