Module 03 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. Previous assistant messages are…
- A. ignored on the next call
- B. input on the next call, including anything wrong
- C. stored separately
- D. automatically fact-checked
2. Reading only response.text discards…
- A. nothing important
- B. finish_reason, usage and the model version
- C. the request id
- D. the system message
3. A response truncated by the token limit…
- A. raises an exception
- B. is valid text that stops mid-thought
- C. returns an empty string
- D. retries automatically
4. The latency number users actually experience is…
- A. total time to complete
- B. time to first chunk
- C. server processing time
- D. the p50 average
5. A log line should always include…
- A. the full prompt
- B. the prompt version and the model version
- C. the API key hash
- D. the user's message
---
Answer Key and Explanations
Check these only after attempting every question.
1. B - input on the next call, including anything wrong
There is nothing marking them as unverified, which is how a guess becomes a premise.
2. B - finish_reason, usage and the model version
Those are exactly what you need during an incident, and none is recoverable afterwards.
3. B - is valid text that stops mid-thought
It is also the most common cause of intermittent JSON parse failures, usually misdiagnosed as model quality.
4. B - time to first chunk
A change that halves total time while doubling time-to-first-chunk feels slower.
5. B - the prompt version and the model version
The prompt changes weekly and is rarely thought of as a deploy; without the version you cannot explain older runs.
Practical Check
Apply this module to your own work: complete the module activity for *Messages, Instructions and Response Handling*, 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.