Speed:
1.0x
The support triage fixture contains seven node records and six connecting edges. Bars show these counts at thirty pixels per record. Planning, tool execution, reflection, memory, human approval and synthesis are descriptive node types. The execution function advances through the nodes array in order, rather than scheduling arbitrary directed graph dependencies. Its approved true flag and refund amount seventy nine point nine nine are literal input payloads, not authenticated approval or an actual refund. There is no model or external tool request in this simulator. In this offline page Cytoscape is missing at initialization. Stepping then fails on a null graph instance before advancing tokens, while the earlier JSON export listener remains available. The intended step arithmetic draws an integer from one hundred fifty through three hundred forty nine for each node. The mean is two hundred forty nine point five, so seven completed nodes would average one thousand seven hundred forty six point five assigned tokens. The source expected result instead stores one thousand two hundred forty. Bars compare those values at point one two pixels per token. This is a counterfactual calculation from unreachable offline step code, not a successfully completed simulation or actual model billing. Missing Cytoscape prevents that execution here. The JSON download exports the workflow schema and its stored expectations, not an independently measured execution trace. The intended telemetry assigns current step count times two hundred forty milliseconds. Seven nodes therefore produce one thousand six hundred eighty milliseconds, whereas the fixture expected elapsed value is one thousand four hundred twenty. Bars use point one two pixels per millisecond. Neither measures real elapsed latency. Native speed edits and actual schema exports can be compared without supplying the missing dependency or changing the source. The node builder mutates its local workflow before attempting graph initialization, so a graph error does not prove the schema stayed unchanged. A local preview preserves this distinction, original tool dimensions and existing bottom links; it is not exhaustive functional proof or public publication.