Evals And Observability easy
A user reports that an agent request took 40 seconds. The request made three LLM calls and two tool calls. What logging design lets you find the slow step quickly?
- A A trace per request with a span for each model and tool call, recording latency and token counts per span.
- B A single log line per request recording the total latency, final output, and HTTP status returned to the user.
- C Separate log streams for each service, correlated afterward by matching timestamps that fall within the same second.
- D Full prompt and output logging for every request at maximum verbosity, retained indefinitely for later analysis.
Answer: A. A trace per request with a span for each model and tool call, recording latency and token counts per span.
A trace that links the user request to every model and tool call costs only kilobytes per request and shows exactly which step was slow or token-heavy. Correlating separate logs by timestamp seems workable but breaks under concurrency, when many requests share the same second. A single summary line is cheaper and suffices only for single-call apps with no tools.
One system design question a day, with spaced repetition so the ones you miss come back.
Get the appRelated questions
- In one release, a RAG assistant got a new embedding model and a new chunking strategy, and the end-to-end judge score dropped 6 points. The team only tracks the end-to-end score and has spent a week tweaking the generation prompt with no recovery. What is the best next step?
- Over two months, your team iterated through 40 prompt versions, each tuned against the same 150-case golden set. The offline score rose from 72 to 91. Online task completion and retry rate have not changed. What is the most likely problem, and what is the fix?
- A prompt change produced an impressive answer in a live demo for leadership. The PM wants to ship it today. What should you do?
- You plan to replace human graders with an LLM judge for your answer-quality eval. Before trusting its scores in a CI gate, what should you do first?
- After you swap to a smaller model, p95 latency falls from 900 ms to 350 ms and the ops dashboard is all green. Leadership asks whether answer quality is still fine. What is the right response?
- Your CI gate fails any prompt change that drops the score 3 or more points on a 100-point rubric over 60 cases. Production generates at temperature 0.7. Running the eval twice on the same commit gives scores that differ by up to 4 points, and engineers have started rerunning failed gates until they pass. How do you fix the gate?