Break Down a Run's Latency by Component
Beginner
The problem, from first principles
A slow agent run has one visible symptom -- it took too long -- but usually many possible causes: a slow retrieval step, a slow LLM call, a slow external tool. Fixing the slowest part first requires knowing which part it actually is, not guessing.
The task
Write latency_breakdown(events) that takes a list of (component, duration) pairs and returns a dict mapping each component to the sum of its durations across the whole run.
Keywords
agentic-systemsagent-observabilitytracing
Python
Run your code to see results
No results yet
Run your code to check it against the examples, or submit to run every test.