Devices, Orchestrator, results
Devices
Section titled “Devices”One card per paired deck: online or busy, its deck version, its workflows and how many of them are broken, the Cloud task it is running with its progress, what waits for it, its next schedule, and its last result. Devices can be grouped into platforms (Flow rig = pump + collector) for organization. A device’s name can be changed from its card; its id cannot.
Library
Section titled “Library”Every workflow of every device, as the devices published them, plus workflows authored on Cloud’s own editor. A Cloud-authored workflow is sent to its device at save time (or kept on Cloud only, in which case it cannot be run until it is sent, since the device would run its own version).
Orchestrator
Section titled “Orchestrator”A canvas where each node is a step on one device: a method call, or a saved workflow of that device. Edges say what waits for what; a node with no incoming edge starts at once, a join waits for every node feeding it. The same Logic steps as the Designer exist as Cloud logic nodes (Wait, User input, If) that run on Cloud between devices.
Each node has a run mode: single (the step once, with its values), spreadsheet (the device’s workflow once per row, with the same per-sample/per-batch walk as the deck’s Iterate tab), or optimization (the device runs the budget loop with its own optimizer). A node can also repeat N times with an interval. Every node is one task, dispatched on its own, one task per device at a time; a device that is busy at the bench is not sent anything until it is free.
Run validates the graph (cycles, unreachable nodes, a missing device, an offline device) and refuses it with the reasons rather than letting it hang. While it runs, each node shows its progress (sample 3/6 · uv_vis.measure_absorbance); a failed task cancels what depended on it and pauses nothing else.
The same Assistant as the Designer’s can draft a graph from a description, for the devices you choose; “Put on canvas” is the only way a draft reaches the canvas.
Schedules
Section titled “Schedules”A planned run that fires on a recurring trigger, from one minute upward. A new schedule first fires one interval from now, never immediately, and a paused one re-arms from the moment it is resumed.
Results
Section titled “Results”Every dispatched run’s record, as the device recorded it, read with the same table Data History uses, with CSV export. Very large records are trimmed on the way (the full record stays on the device).