The agent ID for this flow execution
Set of completed step IDs
OptionalcurrentThe current step being executed
What this run's Decision steps have answered, keyed by each step's key and then by question.
Answers holds the answers a path condition may act on, and is what the decisions root
reads. Unresolved says why each of the rest is not usable (the call failed, the answer fell
below its question's minConfidence, or the step was skipped), and a condition that reads one
of them holds. Never written into the payload.
Where this run's steps execute. Decided once by DetermineInitialStep and read by every
later hook, so a run never switches engines part-way through.
OptionalexecutionWhy executionMode was chosen. Unset until DetermineInitialStep has decided.
Ordered list of step IDs in execution order
The keys of the Decision steps whose call failed in this run.
A walker cannot retry, so such a decision never answers in this run, and a path reading it can
never be taken: it is passed over, and a recovery path ranked below it can be. An answer below
its minConfidence is different — the path might have been the one to take — and holds.
OptionalpendingRemaining starting step IDs queued for execution in inRun mode when a flow has multiple starting steps
OptionalspecialSpecial fields from action output mappings (message, reasoning, confidence)
Map of step results by step ID
Flow execution state that tracks the progress through the workflow. This is maintained by the Flow Agent Type during execution.
Note: Loop iteration tracking is now handled by BaseAgent._iterationContext. FlowExecutionState only tracks flow-specific navigation state.