Applied AI·Evaluation and reliability
you record every prompt, tool call and response in one linked timeline, so you can replay exactly what the run did when it went wrong.
Tracing / spans
Also calledspans
Draft summary, pending review
Recording each step of a multi-step run (model calls, tool executions, retrievals) as spans in one trace, exactly as in distributed tracing. For agents, the trace is the trajectory, and it is the primary debugging artefact.