A Deep Dive into Tracing Agentic Workflows (Part 2)
Part 1 dived into what to trace in an agentic system and why. How the traditional tracing and metrics, such as latency, scale, cost, uptime, and throughput, need to be redefined. And how to define the new metrics that are at the core of an agentics system, such as response quality, accuracy, and task completion. This part is about the mechanics: how a trace is structured, how context propagates across agent boundaries, and how to make sense of it all.
