Variable node

Variable

The Variable node sets a run-scoped value that downstream nodes can read. Use it for anything that needs to be computed once and referenced many times: a timestamp, a batch id, a threshold value read from a config source.

When to use

• A downstream substitution or expression needs a single-value input that isn't on the row ({{ exec.runId }}, {{ exec.batchTimestamp }}).
• You want to compute a value once (at run start) and reference it in many later nodes.
• The value is shared across all rows for the run.

Ports

Port	Direction	Kind	Notes
in	in	data	Optional. When rows flow in, the expression can reference row columns.
out	out	data	Pass-through of the input rows.
error	out	control	Fires on expression eval errors.
notify	out	control	Notification policy port.

Configuration

• Name. The variable name (kebab / camel / snake — whatever fits your reference style downstream). No dots.
• Value or expression. Literal, an expression via evalTransformExpr, or a row-derived value (row.customerId when input rows are present).
• Scope.
  • once (default) — evaluate at run start; the value persists for the whole run.
  • per_row — evaluate per row; downstream sees the row-specific value.

Runtime behaviour

• once scope: the executor evaluates the expression on the first pass and stores it in ctx.execVars. Downstream VAR("name") calls read from that map.
• per_row scope: the value is stored in ctx.execVarsPerRow[nodeId] (or similar) so downstream can pick up per-row.
• Multi-source guard: if a Variable node is reachable from multiple sources, the runtime rejects parallel dispatch and falls back to serial to avoid race conditions.

Failure modes

• Duplicate name. The later Variable overwrites the earlier one; use unique names.
• Multi-source with parallel dispatch. Rejected upfront; documented in the multi-source workflows runbook.

Related nodes

Cache

Persist values across runs

Compute

Derive a row-level value instead of a run-level one

Transform

Reference the variable in a mapping expression