Build an OpenAI Chat Completion Request

The Chat Completion API is the primary interface for all GPT models and the foundation of every OpenAI-powered application. A well-formed request includes a model identifier, a messages array with at least one user role message, and optional parameters to control output behavior. This example shows a complete request for a code explanation task with a system prompt, user message, temperature, and max_tokens configured for concise technical responses. The messages array follows a strict role protocol: system sets the persona and instructions, user carries the human turn, and assistant carries previous AI responses for multi-turn conversations. The order of messages matters — the system message should always come first, followed by alternating user and assistant messages. Inserting a second system message mid-conversation is not standard and may behave differently across models. Key parameters: temperature (0.0–2.0) controls randomness — use 0.0–0.3 for factual tasks and code, 0.7–1.0 for creative writing. max_tokens caps the response length in tokens (not words). top_p is an alternative randomness control; use either temperature or top_p, not both. The model field must exactly match an available model ID including version suffix (e.g., "gpt-4o-2024-08-06").

{
  "model": "gpt-4o",
  "messages": [
    {
      "role": "system",
      "content": "You are a senior software engineer. Explain code clearly and concisely."
    },
    {
      "role": "user",
      "content": "Explain what this function does:\n\nconst debounce = (fn, delay) => {\n  let timer;\n  return (...args) => {\n    clearTimeout(timer);\n    timer = setTimeout(() => fn(...args), delay);\n  };\n};"
    }
  ],
  "temperature": 0.2,
  "max_tokens": 300
}