An agent answer is slow, and you must find which step is responsible. What do you use?
Why C is the answer
Tracing
Tracing captures LLM calls, tool invocations and agent decisions.
In detail
- Tracing shows the time spent in each step.
- Trace the slow path before you optimize.
Use case
An illustrative scenario, not a real incident.
An agent takes 20 seconds. A trace shows 15 of them in one retrieval step.
See it in different ways
flowchart LR A["Request"] --> B["Retrieval span"] B --> C["Model span"] C --> D["Tool span"] D --> E["Response"]
sequenceDiagram participant C as Change participant P as CI/CD pipeline participant R as Production participant T as Traces C->>P: Runs evaluators on a test dataset P->>R: Releases if it passes R->>T: Records real traffic T->>P: New failing cases
flowchart LR S0["Agent run"] --> S1["Trace"] --> S2["Slowest span"] --> S3["Fix it"]
Key point: Tracing finds the slow step.
Why the other options are wrong
- A. A model catalog comparison
- A catalog comparison does not show your agent's steps.
- B. A higher quota in another region
- More quota is a guess.
- D. A content filter
- A filter checks content.