Error handling and rate limits in Prompt Engineering / GenAI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding Rate Limits in API Calls

When using a machine learning API that enforces rate limits, what is the best practice to avoid hitting the limit?

A Send all requests as fast as possible and retry only if the server crashes.

B Ignore rate limits because the API will queue requests automatically.

C Implement a delay between requests to stay within the allowed number of calls per minute.

D Use multiple API keys simultaneously without restrictions.

💡 Hint

Think about how to respect the server's capacity to handle requests.

APIs with rate limits require clients to space out requests to avoid errors. Implementing delays or backoff strategies helps stay within limits.

❓ Predict Output

intermediate

Output of Error Handling Code

What will be the output of the following Python code snippet that calls a machine learning API with error handling?

Prompt Engineering / GenAI

import time

class APIError(Exception):
    pass

def call_api():
    raise APIError('Rate limit exceeded')

try:
    call_api()
except APIError as e:
    print(f'Error caught: {e}')
    time.sleep(1)
    print('Retrying...')

A

Error caught: Rate limit exceeded
Retrying...

B Error caught: Rate limit exceeded

C No output, program crashes

D Retrying...

💡 Hint

Look at what happens inside the except block.

The exception is caught and prints the error message, then waits 1 second, and prints 'Retrying...'.

❓ Model Choice

advanced

Choosing a Strategy for Handling API Rate Limits

You have a machine learning model API with a strict rate limit of 5 requests per second. Which strategy best handles this limit while maximizing throughput?

A Queue requests and send them at a steady rate of 5 per second.

B Send requests randomly without delay and ignore errors.

C Send 10 requests at once and retry failed ones after 10 seconds.

D Send 1 request per minute to be safe.

💡 Hint

Consider how to keep requests within the allowed rate without wasting time.

Queuing requests and sending them at the allowed rate maximizes throughput without triggering errors.

❓ Metrics

advanced

Interpreting Error Rate Metrics

You monitor your ML API calls and see the following metrics over 1000 requests: 950 successful, 30 rate limit errors, 20 timeout errors. What is the error rate percentage?

A 50%

B 2%

C 10%

D 5%

💡 Hint

Error rate = (number of errors / total requests) * 100

Total errors = 30 + 20 = 50. Error rate = (50/1000)*100 = 5%.

🔧 Debug

expert

Debugging API Rate Limit Handling Code

Given the following Python code snippet that calls an ML API, which option correctly identifies the bug causing the program to crash?

Prompt Engineering / GenAI

import time

rate_limit = 3
calls = 0
start_time = time.time()

def call_api():
    global calls, start_time
    if calls >= rate_limit:
        elapsed = time.time() - start_time
        time.sleep(1 - elapsed)
        calls = 0
        start_time = time.time()
    calls += 1
    print('API call made')

for _ in range(5):
    call_api()

A The variable 'calls' is not declared global inside call_api(), causing UnboundLocalError.

B The time.sleep() call may receive a negative argument causing ValueError.

C The start_time is reset too early, causing infinite loop.

D The loop runs 5 times but only 3 API calls are made.

💡 Hint

Check the calculation inside time.sleep().

If elapsed is greater than 1, 1 - elapsed is negative, and time.sleep() raises ValueError.