Interactive block example — the Data Router

DataRouter (data_router.py) is SciStudio's built-in Data Router block: the workflow runs until it reaches this block, pauses, opens a drag-and-drop window, and waits. The user drags each input item onto the output port it should leave by; the block resumes, routes the items exactly as assigned, and the decision is part of the run's record. No code is written for any specific routing — the same block merges, splits, or re-sorts items across branches.

The four declarations of an interactive block

An interactive block is declared completely or not at all — the registry rejects a block that has only some of these (see the comments on the class attributes in data_router.py):

Declaration	What it does
InteractiveMixin	Adds the pause/resume contract to the block.
execution_mode = ExecutionMode.INTERACTIVE	The run pauses at this block until a decision arrives.
interactive_panel = PanelManifest(panel_id=...)	Names the panel the frontend opens while paused — here the built-in core.interactive.data_router panel (panels example).
prepare_prompt(...)	Builds the plain-JSON view the panel renders: here the input ports, every item with a human label, and the output ports.

The two halves of the run

1. prepare_prompt runs before the pause, in an isolated worker. It sees the full inputs and returns plain JSON only — no data objects, no references — describing what the user will decide over.
2. run runs after the user confirms. The panel's decision arrives as config.get("interactive_response", {}); here it is assignments, a mapping of output port name to item refs. The outputs follow from the inputs, the config, and that decision alone — nothing the user did in the window is needed beyond the JSON.

Notice also the variadic ports: variadic_inputs/variadic_outputs let the user add as many ports as the routing needs (min_input_ports/ min_output_ports bound them), and get_effective_output_ports() reads the ports the user actually configured.

The paired panel

The window the user drags items in is panels/core.interactive.data_router/ — the interactive-panel example. Read the two together: this file decides what the panel is offered (prepare_prompt) and what happens with the answer (run); the panel folder decides how the decision looks and is submitted.

Try it

Add a Data Router node between two branches, run, and the run pauses on it. Drag every item to an output, press Confirm, and each output port carries the Collection you built.