OpenAI API Examples: Chat Completions in Practice

What you will build

Assuming you finished OpenAI API Application, you will:

• Authenticate with an API key
• Send a chat completion
• Parse the reply and handle common errors
• See a minimal streaming pattern

Docs: https://platform.openai.com/docs/ · Console: https://platform.openai.com/

Model names and paths change—treat samples as structural templates and confirm against current Docs.

1. curl smoke test

curl https://api.openai.com/v1/chat/completions \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o-mini",
    "messages": [
      {"role": "system", "content": "You are a concise assistant."},
      {"role": "user", "content": "Explain what an API is in three sentences."}
    ],
    "temperature": 0.4
  }'

Look for choices[0].message.content in the JSON response.

2. Minimal Python

import os
from openai import OpenAI

client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])

resp = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[
        {"role": "system", "content": "You are a careful technical writer."},
        {"role": "user", "content": "List five security checks before calling the OpenAI API."},
    ],
    temperature=0.3,
)

print(resp.choices[0].message.content)

Install the official SDK per Docs; watch for breaking changes across majors.

3. Message roles

role	Use
system	Global style, safety rails, output contract
user	Current request
assistant	Prior replies when continuing a thread

Trim long histories; summarize when context and cost grow.

4. Retry skeleton

import time
from openai import RateLimitError, APIError

def chat_with_retry(client, **kwargs):
    for attempt in range(5):
        try:
            return client.chat.completions.create(**kwargs)
        except RateLimitError:
            time.sleep(2 ** attempt)
        except APIError:
            raise
    raise RuntimeError("retries exhausted")

5. Streaming starter

stream = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Write a short poem about debugging"}],
    stream=True,
)

for chunk in stream:
    delta = chunk.choices[0].delta.content or ""
    print(delta, end="", flush=True)

6. Production habits

1. Keys only in server-side secret stores
2. Bound user input length and apply content policies
3. Log request id, model, latency, token usage
4. Centralize prompt configs—no scattered magic strings
5. Budget alerts against runaway loops

Explore interactively on ChatGPT; product home: OpenAI.

7. Error cheatsheet

Symptom	Likely cause	Fix
401	Bad key	Regenerate; fix env var
429	Rate/quota	Backoff; lower concurrency
Timeouts	Network/proxy	Retry with timeouts; stable path
Truncation	Low max tokens	Raise limit or ask for shorter output

A mirror site entry is a placeholder only—do not route production traffic without review.

Takeaway

Prove auth with curl, wrap a client, then add retries, logging, and streaming. Combine key hygiene from the application guide with these templates, and keep Docs as the living reference.