Python for AI: From Zero to Your First AI App

100%

Page 16 of 20 · 22 min left

Chapter 5 · Python for AI Work

Web APIs with requests

Page 16 of 20
3 min read

An API (Application Programming Interface) is a way for one program to ask another for something. A web API does it over the internet with HTTP — the same protocol your browser uses. Every hosted AI model — OpenAI, Anthropic, Google, Hugging Face — is used through a web API. Before using an AI SDK on the next page, it is worth seeing what it does underneath, with the popular requests library (python -m pip install requests).

A request and a response

Your program sends a request: a method, a URL, some headers and sometimes a body. The server sends back a response: a status code, headers and a body, usually JSON.

Method	Meaning	Example
`GET`	Read something	Fetch a list of available models
`POST`	Send something to be processed or created	Send a prompt and get an answer
`PUT` / `PATCH` / `DELETE`	Replace / change / remove	Update or delete a stored file

GET: reading data

The examples use httpbin.org, a free service that echoes back what you send — perfect for learning:

Python

import requests

response = requests.get(
    "https://httpbin.org/get",
    params={"q": "python for ai", "page": 1},     # becomes ?q=python+for+ai&page=1
    timeout=10,                                   # seconds — always set one
)
print(response.status_code)
print(response.headers["Content-Type"])
data = response.json()                            # JSON text → Python dict
print(data["args"])
print(data["url"])

Output

200
application/json
{'page': '1', 'q': 'python for ai'}
https://httpbin.org/get?q=python+for+ai&page=1

POST: sending data

This is the shape of every call to an AI model: a POST with a JSON body and an Authorization header carrying your API key:

Python

import requests

payload = {"model": "demo-model", "input": "Translate 'good morning' into Bangla."}
response = requests.post(
    "https://httpbin.org/post",
    json=payload,                                  # sends JSON and sets Content-Type
    headers={"Authorization": "Bearer my-secret-key"},
    timeout=10,
)
echo = response.json()                             # httpbin sends back what it received
print(response.status_code)
print(echo["json"])
print(echo["headers"]["Content-Type"])
print(echo["headers"]["Authorization"])

Output

200
{'input': "Translate 'good morning' into Bangla.", 'model': 'demo-model'}
application/json
Bearer my-secret-key

json=payload converts the dictionary to JSON for you and sets the Content-Type: application/json header.

Status codes

Code	Meaning	What to do
200–299	Success	Read the body
400	Bad request — something in what you sent is wrong	Fix your code; retrying will not help
401 / 403	Not authenticated / not allowed	Check your API key and permissions
404	Not found	Check the URL or the resource name
429	Too many requests (rate limit)	Wait, then retry with backoff (page 11)
500–599	The server had a problem	Retry later with backoff

Handling failures properly

The network fails in three different ways, and robust code handles each one:

Python

import requests

def fetch_json(url):
    try:
        response = requests.get(url, timeout=2)
        response.raise_for_status()          # turns 4xx/5xx into an exception
        return response.json()
    except requests.Timeout:
        print("Timed out:", url)
    except requests.HTTPError as error:
        print("HTTP error:", error.response.status_code, url)
    except requests.ConnectionError:
        print("Could not connect:", url)
    return None

print(fetch_json("https://httpbin.org/json") is not None)
fetch_json("https://httpbin.org/status/404")
fetch_json("https://httpbin.org/status/503")
fetch_json("https://httpbin.org/delay/5")

Output

True
HTTP error: 404 https://httpbin.org/status/404
HTTP error: 503 https://httpbin.org/status/503
Timed out: https://httpbin.org/delay/5

Always pass timeout. Without one, requests can wait forever for a server that never answers, and your program hangs.
raise_for_status() turns an error status into an exception. Without it, a 404 is a "successful" call whose body is an error page.
Catch the specific exceptions: a timeout and a 503 are worth retrying; a 404 is not.

Never put a real API key in your code. The next page shows the right way to keep it secret.

Try it yourself

Send a GET to https://httpbin.org/get with your name as a parameter and print it from the response.
Write get_with_retries(url, attempts=3) that retries only on timeouts and 5xx errors, waiting longer each time.
Use https://httpbin.org/headers to see which headers requests sends by default.

Python for AI: From Zero to Your First AI App

100%

Page 16 of 20 · 22 min left

Chapter 5 · Python for AI Work

Web APIs with requests

Page 16 of 20
3 min read

A request and a response

Your program sends a request: a method, a URL, some headers and sometimes a body. The server sends back a response: a status code, headers and a body, usually JSON.

Method	Meaning	Example
`GET`	Read something	Fetch a list of available models
`POST`	Send something to be processed or created	Send a prompt and get an answer
`PUT` / `PATCH` / `DELETE`	Replace / change / remove	Update or delete a stored file

GET: reading data

The examples use httpbin.org, a free service that echoes back what you send — perfect for learning:

Python

import requests

response = requests.get(
    "https://httpbin.org/get",
    params={"q": "python for ai", "page": 1},     # becomes ?q=python+for+ai&page=1
    timeout=10,                                   # seconds — always set one
)
print(response.status_code)
print(response.headers["Content-Type"])
data = response.json()                            # JSON text → Python dict
print(data["args"])
print(data["url"])

Output

200
application/json
{'page': '1', 'q': 'python for ai'}
https://httpbin.org/get?q=python+for+ai&page=1

POST: sending data

This is the shape of every call to an AI model: a POST with a JSON body and an Authorization header carrying your API key:

Python

import requests

payload = {"model": "demo-model", "input": "Translate 'good morning' into Bangla."}
response = requests.post(
    "https://httpbin.org/post",
    json=payload,                                  # sends JSON and sets Content-Type
    headers={"Authorization": "Bearer my-secret-key"},
    timeout=10,
)
echo = response.json()                             # httpbin sends back what it received
print(response.status_code)
print(echo["json"])
print(echo["headers"]["Content-Type"])
print(echo["headers"]["Authorization"])

Output

200
{'input': "Translate 'good morning' into Bangla.", 'model': 'demo-model'}
application/json
Bearer my-secret-key

json=payload converts the dictionary to JSON for you and sets the Content-Type: application/json header.

Status codes

Code	Meaning	What to do
200–299	Success	Read the body
400	Bad request — something in what you sent is wrong	Fix your code; retrying will not help
401 / 403	Not authenticated / not allowed	Check your API key and permissions
404	Not found	Check the URL or the resource name
429	Too many requests (rate limit)	Wait, then retry with backoff (page 11)
500–599	The server had a problem	Retry later with backoff

Handling failures properly

The network fails in three different ways, and robust code handles each one:

Python

import requests

def fetch_json(url):
    try:
        response = requests.get(url, timeout=2)
        response.raise_for_status()          # turns 4xx/5xx into an exception
        return response.json()
    except requests.Timeout:
        print("Timed out:", url)
    except requests.HTTPError as error:
        print("HTTP error:", error.response.status_code, url)
    except requests.ConnectionError:
        print("Could not connect:", url)
    return None

print(fetch_json("https://httpbin.org/json") is not None)
fetch_json("https://httpbin.org/status/404")
fetch_json("https://httpbin.org/status/503")
fetch_json("https://httpbin.org/delay/5")

Output

True
HTTP error: 404 https://httpbin.org/status/404
HTTP error: 503 https://httpbin.org/status/503
Timed out: https://httpbin.org/delay/5

Always pass timeout. Without one, requests can wait forever for a server that never answers, and your program hangs.
raise_for_status() turns an error status into an exception. Without it, a 404 is a "successful" call whose body is an error page.
Catch the specific exceptions: a timeout and a 503 are worth retrying; a 404 is not.

Never put a real API key in your code. The next page shows the right way to keep it secret.

Try it yourself

Send a GET to https://httpbin.org/get with your name as a parameter and print it from the response.
Write get_with_retries(url, attempts=3) that retries only on timeouts and 5xx errors, waiting longer each time.
Use https://httpbin.org/headers to see which headers requests sends by default.