Tracing GenAI Requests Quiz

What you'll be asked

1. In the context of GenAI observability, what is the primary purpose of a 'trace'?
2. When implementing observability for a RAG (Retrieval-Augmented Generation) pipeline, which span metadata is most critical for debugging 'hallucinations'?
3. Which of the following is a key reason to use OpenTelemetry (OTel) for GenAI observability?
4. Why is it important to include 'prompt templates' and 'model parameters' (like temperature) in your trace metadata?
5. You notice a high latency in your GenAI application. You observe that the span for the 'Vector Store Retrieval' is taking 500ms, while the 'LLM Completion' span is taking 4 seconds. What is the most effective observability strategy to optimize the total request time?

Answers and explanations come after you submit.