APIs and HTTP

Almost every AI service is reached through an HTTP API: you send a request (method + URL + JSON if needed) and get a response (status code + JSON).

Code	Means	What you do
200	OK	read the response
400	your request is wrong	read the error message
401/403	no key, or the wrong one	check the key
404	wrong address	check the URL
429	too many calls	wait and try again
500	the server fell over	try again later

Keys are passwords. Never in the code, never in git — read them from an environment variable.

import os, requests

KEY = os.environ["LLM_API_KEY"]           # set outside the code
r = requests.post(
    "https://api.example.com/v1/chat/completions",
    headers={"Authorization": f"Bearer {KEY}"},
    json={"model": "small", "messages": [{"role": "user", "content": "Hi"}]},
    timeout=30,
)
if r.status_code == 429:
    print("quota exhausted — wait", r.headers.get("Retry-After"))
r.raise_for_status()                       # raises on 4xx/5xx
response = r.json()
print(response["choices"][0]["message"]["content"])

Always set a timeout — otherwise the program can hang forever when the server does not answer.