Skip to main content

NodeGraph

Description​

NodeGraph is the in-memory representation of one graph inside a Rivet project. A graph contains node instances, connections between their ports, and optional metadata.

Projects store many graphs in Project.graphs, keyed by graph ID.

Definition​

export type GraphId = Opaque<string, 'GraphId'>;

export interface NodeGraph {
  metadata?: {
    id?: GraphId;
    name?: string;
    description?: string;
    attachedData?: AttachedData;
  };

  nodes: ChartNode[];
  connections: NodeConnection[];
}

Properties​

metadata​

Type:

{
  id?: GraphId;
  name?: string;
  description?: string;
  attachedData?: AttachedData;
}

Optional graph metadata.

id is the graph ID. name is the name shown in the Rivet UI and can also be used by run helpers to select a graph. description is user-facing documentation for the graph. attachedData stores serialized sidecar data for graph content.

nodes​

Type: ChartNode[]

The nodes that make up the graph. Each node has a type, visual position, node-specific data, and shared execution settings such as run mode.

connections​

Type: NodeConnection[]

The directed connections between node ports.

Creating an Empty Graph​

The core package exports emptyNodeGraph():

import { emptyNodeGraph } from '@valerypopoff/rivet2-core';

const graph = emptyNodeGraph();

It creates a graph with no nodes, no connections, a generated ID, the name Untitled Graph, and an empty description.

Notes​

  • Graph IDs are opaque strings at the type level, but they serialize as strings in project YAML.
  • A graph can be run directly, used as the project main graph, or called from another graph through a Subgraph node.
  • Graph Inputs and Graph Outputs define the callable interface when another graph uses this graph as a subgraph.

See Also​