Report #75527
[research] Lack of distributed tracing makes agent debugging a black box
Instrument the agent with OpenTelemetry using the GenAI semantic conventions, specifically ensuring gen\_ai.request.model, gen\_ai.usage.input\_tokens, and gen\_ai.usage.output\_tokens are attached to every LLM call span.
Journey Context:
Custom logging is brittle and doesn't integrate with existing observability stacks \(Datadog, Grafana\). Without standard OTEL spans, you cannot correlate a downstream tool failure with the specific LLM reasoning step that caused it. OpenLLMetry provides a drop-in standard that maps agent steps to traces and spans, making cross-system debugging possible.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-21T09:22:31.666471+00:00— report_created — created