Quiz: Module 9 - LLM APIs & Integration

Question 1

What does temperature=0 produce?

1. Most deterministic output
2. Most random output
3. No output
4. Error

Question 2

What is function calling used for?

1. Enabling LLM to request external tool execution
2. Calling Python functions faster
3. Debugging API calls
4. Testing endpoints

Question 3

What is streaming useful for?

1. Showing partial results to users immediately
2. Reducing costs
3. Improving accuracy
4. Storing responses

Question 4

What library abstracts multiple LLM providers?

1. LiteLLM or LangChain
2. NumPy
3. Pandas
4. Requests

Question 5

How to handle rate limits?

1. Exponential backoff with retries
2. Ignore and continue
3. Restart application
4. Use different API key

Question 6

What is RAG pattern?

1. Retrieve context, Augment prompt, Generate response
2. Random Answer Generation
3. Recursive API Grouping
4. Response Accuracy Grading

Question 7

Why cache LLM responses?

1. Reduce costs and latency for repeated queries
2. Improve accuracy
3. Required by API
4. Debugging only

Question 8

What does max_tokens control?

1. Maximum length of generated response
2. Input length
3. Cost limit
4. Request timeout

Question 9

What is the system role in chat API?

1. Set overall behavior and context for assistant
2. Operating system information
3. Error messages
4. API configuration

Question 10

How to estimate API costs?

1. Track token usage × price per token
2. Count API calls only
3. Time spent
4. Character count

Questions 11-30

(Covering: top_p, frequency_penalty, stop sequences, assistants API, vision API, embeddings API, batch processing, error codes, SDK versions, async calls, webhooks, response formats, JSON mode, seed for reproducibility, logprobs, token counting, budget alerts, multi-turn conversations, context management)

Question 11

What is top_p (nucleus sampling)?

1. Sampling from top P probability mass
2. Top priority
3. Temperature parameter
4. Penalty term

Question 12

What does frequency_penalty reduce?

1. Token repetition based on frequency
2. API call frequency
3. Cost
4. Latency

Question 13

What are stop sequences?

1. Strings that end generation when encountered
2. Error indicators
3. API timeouts
4. Model warnings

Question 14

What is the Assistants API for?

1. Building persistent, stateful AI assistants
2. Human assistant hiring
3. Documentation help
4. Code assistance only

Question 15

What does JSON mode ensure?

1. Response is valid JSON
2. Faster response
3. Lower cost
4. Better accuracy

Question 16

What is seed parameter for?

1. Reproducible outputs (same seed = same output)
2. Random initialization
3. Database seeding
4. Training initialization

Question 17

What are logprobs?

1. Log probabilities of generated tokens
2. Error logs
3. API logs
4. Debug information

Question 18

How to count tokens before API call?

1. Use tiktoken library
2. Count words
3. Count characters
4. Estimate from length

Question 19

What is async API usage benefit?

1. Non-blocking, concurrent requests
2. Lower costs
3. Better accuracy
4. Simpler code

Question 20

How to manage multi-turn conversations?

1. Include conversation history in messages array
2. API handles automatically
3. Session cookies
4. Database storage only

Question 21

What is presence_penalty?

1. Penalty for tokens that have appeared, encouraging new topics
2. User presence detection
3. API availability penalty
4. Connection penalty

Question 22

What HTTP status indicates rate limit?

1. 429 Too Many Requests
2. 500 Internal Error
3. 404 Not Found
4. 401 Unauthorized

Question 23

What is batch processing API for?

1. Processing many requests at lower cost with longer wait
2. Real-time processing
3. Streaming
4. Testing

Question 24

What is context window management?

1. Handling input that exceeds model's context limit
2. Window GUI management
3. Browser control
4. Session handling

Question 25

What's a fallback strategy?

1. Using alternative model/provider if primary fails
2. Backup data
3. Error messages
4. Logging

Question 26

What does embeddings API return?

1. Vector representations of text
2. Generated text
3. Classifications
4. Summaries

Question 27

What is the benefit of response_format parameter?

1. Guarantee structured output format (JSON, etc.)
2. Faster responses
3. Lower cost
4. Better accuracy

Question 28

How to handle context overflow?

1. Summarize or truncate earlier messages
2. Just send everything
3. API handles automatically
4. Restart conversation

Question 29

What is tool_choice parameter?

1. Control whether/which function to call
2. Select IDE
3. Choose model
4. Pick provider

Question 30

What is usage.prompt_tokens?

1. Number of tokens in input/prompt
2. Output tokens
3. Total tokens
4. Maximum tokens

---

Passing Score: 24/30 (80%)

← Chapter 3Module 10 →