Optional iterationsNumber of internal LLM iterations the node ran (GMI/ReAct loop).
Optional toolSuccessful tool result events observed during execution.
Optional toolFailed tool calls observed during execution.
Optional iterationsTrue when the loop hit maxIterations without a natural termination.
Optional promptCumulative LLM token usage and cost across every round of the node's
ReAct loop. Populated by executors that route through a usage-aware
provider call (runToolCallingTurn in wunderland); omitted when the
provider didn't return usage data. Token counts are integers, cost
is in USD, all summed across rounds.
Optional completionOptional totalOptional costUSD
Optional per-node telemetry attached to
node_endevents. Populated by executors that have meaningful internal activity worth surfacing — today the GMI executor reports the ReAct-loop iteration count, the number of tool calls and errors observed, whether the iteration cap was hit, and cumulative token usage / cost for the node's LLM calls. Other executors omittelemetryentirely.