A model view can use either stream, or both together:
| Need | Use |
|---|---|
| Model outputs (screens, beam parameters) evaluated at the machine’s current inputs | useLiveStream (lume-model-api SSE) |
| Raw machine readbacks or setpoints, e.g. to compare with the model’s prediction | usePvStream / useHareStream (Canopy Hare WebSocket) |
| A model’s inputs as the machine sees them, without evaluating | useHareStream(config.inputs.map((i) => i.id)), or useMachineSnapshot for a one-off read |
The two run independently: each has its own connection, status and error.
The streaming pages read either source through useModelStream(source, model, config)
(src/hooks/useModelStream.ts), which converts both into one snapshot of inputs and outputs. It
updates that snapshot at most once per second (modelStreamTiming.displayIntervalMs, or the
intervalMs option), so the tables don’t re-render on every frame. The connection still receives
every update, and the latest values are shown at the end of each interval.