OpenAI API

Introduction

The OpenAI API is the most-used LLM API in the world. Two endpoints matter most: chat.completions (and the newer responses) for text, and embeddings for vectors. Both are dead-simple to call.

Beginner analogy: Like calling Stripe — one HTTPS POST, get a JSON response.

Understanding the topic

Core concepts to understand:

• Authenticate with a Bearer API key.
• Models: gpt-4o, gpt-4o-mini, o1, o3.
• Stream with Server-Sent Events.
• Structured outputs via response_format.
• Tool calling for actions / function calls.

Syntax reference

Visual workflow / architecture:

POST https://api.openai.com/v1/chat/completions
Authorization: Bearer sk-...

{
  "model": "gpt-4o",
  "messages": [
    { "role": "system", "content": "You are helpful." },
    { "role": "user",   "content": "Hello!" }
  ],
  "stream": true
}

Real-world use

Used by Notion, Stripe, Shopify, Khan Academy, Duolingo and basically every consumer AI product launched since 2023.

Best practices

• Never put API keys in frontend code — always proxy through your backend.
• Use gpt-4o-mini for 80% of tasks; escalate when needed.
• Set per-user rate limits and budget caps.

Hands-on exercise

Interview preparation — practice these questions:

• Q1. Walk through a basic chat.completions call.
• Q2. How do you stream responses?
• Q3. How do you guarantee JSON output?