RAG answer latency increases.
Find whether retrieval, model, or a tool is responsible.
Choose two observability practices.
Correlate intake, retrieval, reranking, inference, and output with a request ID and record latency, tokens, errors, and evidence count while minimizing sensitive data.
Detailed explanation
Performance and cost bottlenecks become visible.
Stage latency and quality are hidden.
One response's path can be traced.
Sensitive exposure and retention increase.
Check official documentation
Compare the following points with the official documentation.
AWS公式AIF-C01 Domain 3.4の監視、トレース、ログ、費用を確認する。
Official references
What to confirm
RAGの処理経路を工程別に可視化し、機密情報を過剰に保存しない設計ができる。
Key points
- Trace
- Request ID
- Minimization
Notes
- Environment: AWS公式AIF-C01試験ガイドとAWS公式ドキュメントの確認
- Compare the definition with the AIF-C01 objective and the service documentation before applying it.
- Keep the concept separate from a particular answer choice or implementation step.
Foundation review
Distributed trace
A record that follows one request across services and their individual processing spans.
Request ID
An identifier used to correlate logs and events belonging to one request.