問題文
A team is about to move a document summarization service into production. Which two observability items must be in place before launch to make later problems diagnosable? (Select two.)
選択肢
- Full request and response bodies stored indefinitely for every call, including customer data.
- Counters for each failure category, separating request errors, rate limiting, overload, timeouts, and output validation failures.
- A dashboard showing total monthly spend, which is enough to detect quality problems early.
- A weekly manual review of a few outputs by the development team, in place of automated measurement.
- A per-request record linking the request identifier to the prompt version, model version, token counts, latency, and outcome.