Module 08 Knowledge Check
5 questions. Pass mark 4 out of 5. Answer every question before checking the answer key below, then retry after reading the feedback.
1. The safest place to cut cost first is usually…
- A. A smaller model everywhere
- B. Removing unnecessary context and redundant calls
- C. Fewer evals
- D. Shorter answers
2. Caching identical requests is straightforward, but caching in RAG requires care because…
- A. It is slow
- B. The corpus can change, so a cached answer may become stale and wrong
- C. Vectors cannot cache
- D. Costs rise
3. Rate limits should be handled with…
- A. Immediate retry
- B. Backoff with jitter and a bounded retry count
- C. Failing instantly
- D. More parallelism
4. Streaming improves…
- A. Total latency
- B. Perceived latency - time to first token, not total time
- C. Cost
- D. Accuracy
5. A cost spike with stable traffic most likely indicates…
- A. Vendor pricing
- B. Growing context - longer histories, more retrieved chunks, or retry loops
- C. Better answers
- D. Cache hits
---
Answer Key and Explanations
Check these only after attempting every question.
1. B - Removing unnecessary context and redundant calls
Trimming waste rarely costs quality; downgrading the model often does.
2. B - The corpus can change, so a cached answer may become stale and wrong
Cache keys must incorporate corpus version or the cache silently serves outdated facts.
3. B - Backoff with jitter and a bounded retry count
Tight retry loops amplify the incident that caused the limit in the first place.
4. B - Perceived latency - time to first token, not total time
Useful and real, but do not confuse it with making the work finish sooner.
5. B - Growing context - longer histories, more retrieved chunks, or retry loops
Context length and retries are the usual culprits and are directly measurable.
Practical Check
Apply this module to your own work: complete the module activity for *Cost, Latency, Caching, and Rate Limits*, then write one sentence naming what your result shows and one naming what it does not.
Strong Answer Pattern
A strong answer names the task, the evidence used, the check performed, and the remaining limitation. It avoids "proved", "guaranteed", or "always" unless the evidence genuinely supports it.