TRACING AND SPANS

ARTIFICIAL INTELLIGENCE ASSISTANT PROMPTS [START OF PROMPT] Analyze this trace (list of spans with timing and status) and identify the latency bottleneck span(s) and any span with an elevated error rate: [PASTE LAB OR PRODUCTION TRACE DATA HERE]. Output format: a ranked list of spans by time contribution, plus a one-sentence root-cause hypothesis for the slowest or most error-prone span. [END OF PROMPT] [START OF PROMPT] Review this tracing/span-instrumentation code for missing fields (span identifier (ID), parent span ID, start/end time, input, output, status, error message) that a production-quality tracing system needs: [PASTE LAB TRACING CODE HERE]. For each missing field, state the specific debugging scenario it would make harder to diagnose without it. Output format: a gap table (Field, Present? Y/N, Debugging Scenario Affected). [END OF PROMPT] CHAPTER SUMMARY A span records one unit of work, and a trace is the full chain of spans for one request. Parent and child identifiers let spans form a tree. Each span should record identifiers, timing, input, output, status, and errors. Traces quickly reveal bottlenecks and failures, while privacy must be protected. KEY TERMS REVIEW Span: a record of one unit of work. Trace: the chain of spans for one request. Parent span: a span that contains other spans. Step-level tracing: recording every step of the agent's work. Bottleneck: the step that limits overall speed.