Implementation:FlowiseAI Flowise InitNode

Attribute	Value
Source Repository	FlowiseAI/Flowise
Source File	packages/ui/src/utils/genericHelper.js
Domain	Chatflow_Creation
Workflow	Chatflow_Creation
Last Updated	2026-02-12

Overview

Concrete utility function for initializing a ReactFlow node from a server-provided component definition. The initNode function transforms a raw NodeDefinition into a fully configured node data object with classified inputs, initialized outputs, default values, and visibility states. A companion function getUniqueNodeId generates collision-free node identifiers.

Code Reference

Source Location

• File: packages/ui/src/utils/genericHelper.js
• Lines: L118-227 (initNode), L4-17 (getUniqueNodeId)

Signature

export const initNode = (nodeData, newNodeId, isAgentflow) => { ... }

Where:

• nodeData: NodeDefinition from server (with inputs[], outputs[], credential?, baseClasses[])
• newNodeId: string -- unique ID generated by getUniqueNodeId()
• isAgentflow: boolean -- toggles output anchor format (agentflow vs standard)

Returns: Modified nodeData with id, inputAnchors[], inputParams[], inputs (object with defaults), outputAnchors[]

The companion ID generator:

export const getUniqueNodeId = (nodeData, nodes) => {
    let suffix = 0
    let baseId = `${nodeData.name}_${suffix}`
    while (nodes.some((node) => node.id === baseId)) {
        suffix += 1
        baseId = `${nodeData.name}_${suffix}`
    }
    return baseId
}

Import

import { initNode, getUniqueNodeId } from '@/utils/genericHelper'

I/O Contract

Inputs

Parameter	Type	Required	Description
nodeData	NodeDefinition	Yes	Raw component definition from the server registry, containing inputs[], outputs[], credential?, baseClasses[], name, label, type
newNodeId	string	Yes	Unique node identifier generated by getUniqueNodeId()
isAgentflow	boolean	No	When true, output anchors use agentflow format (simple indexed IDs); when false or omitted, standard format with base class encoding is used

Outputs

Field	Type	Description
id	string	The assigned unique node ID
inputAnchors	InputDef[]	Inputs classified as connection handles (non-whitelist types), each with a generated id and display flag
inputParams	InputDef[]	Inputs classified as form fields (whitelist types), each with a generated id and display flag. Credential input is prepended if present.
inputs	object	Key-value map of input names to their default values (or empty string)
outputAnchors	OutputAnchor[]	Output anchors with IDs encoding base classes
outputs	object	Key-value map of output names to their default values
credential	string	Reset to empty string (credential ID is stored in inputs at runtime)

Implementation Details

Whitelist Types

The following input types are classified as form-configurable parameters (inputParams):

const whitelistTypes = [
    'asyncOptions',
    'asyncMultiOptions',
    'options',
    'multiOptions',
    'array',
    'datagrid',
    'string',
    'number',
    'boolean',
    'password',
    'json',
    'code',
    'date',
    'file',
    'folder',
    'tabs',
    'conditionFunction'
]

Any input with a type not in this list becomes an inputAnchor (connection handle).

Input ID Format

Each input is assigned an ID following the pattern:

{nodeId}-input-{inputName}-{inputType}

For example: chatOpenAI_0-input-model-asyncOptions

Output Anchor Formats

Standard chatflow format (when isAgentflow is false):

If the definition has explicit outputs, they are wrapped in an options structure:

{
    name: 'output',
    label: 'Output',
    type: 'options',
    options: [
        {
            id: `${newNodeId}-output-${output.name}-${baseClasses}`,
            name: output.name,
            label: output.label,
            type: baseClasses.join(' | ')
        }
    ],
    default: outputs[0].name
}

Agentflow format (when isAgentflow is true):

{
    id: `${newNodeId}-output-${index}`,
    label: nodeData.label,
    name: nodeData.name
}

Usage Examples

Drag-and-drop node creation (from Canvas component)

import { getUniqueNodeId, initNode } from '@/utils/genericHelper'

const onDrop = (event) => {
    let nodeData = JSON.parse(event.dataTransfer.getData('application/reactflow'))

    const position = reactFlowInstance.project({
        x: event.clientX - reactFlowBounds.left - 100,
        y: event.clientY - reactFlowBounds.top - 50
    })

    const newNodeId = getUniqueNodeId(nodeData, reactFlowInstance.getNodes())

    const newNode = {
        id: newNodeId,
        position,
        type: nodeData.type !== 'StickyNote' ? 'customNode' : 'stickyNote',
        data: initNode(nodeData, newNodeId)
    }

    setNodes((nds) => nds.concat(newNode))
}

Programmatic node initialization

import { getUniqueNodeId, initNode } from '@/utils/genericHelper'

// Given a node definition from the registry
const nodeDef = {
    name: 'chatOpenAI',
    label: 'ChatOpenAI',
    type: 'ChatOpenAI',
    baseClasses: ['BaseChatModel', 'BaseLanguageModel'],
    inputs: [
        { name: 'model', type: 'asyncOptions', label: 'Model Name' },
        { name: 'temperature', type: 'number', label: 'Temperature', default: 0.9 },
        { name: 'cache', type: 'BaseCache', label: 'Cache' }
    ],
    outputs: [],
    credential: { name: 'credential', type: 'credential', label: 'OpenAI API Key' }
}

const existingNodes = []
const newNodeId = getUniqueNodeId(nodeDef, existingNodes) // "chatOpenAI_0"
const initialized = initNode(nodeDef, newNodeId, false)

// initialized.inputParams includes: credential selector, model (asyncOptions), temperature (number)
// initialized.inputAnchors includes: cache (BaseCache)
// initialized.inputs = { model: '', temperature: 0.9, cache: '' }

Related Pages

• Principle:FlowiseAI_Flowise_Node_Initialization
• Heuristic:FlowiseAI_Flowise_Tool_Ordering_Convention