Working with Nodes

Nodes are the building blocks of chatflows. Each node represents a specific component like an LLM, tool, memory system, or data processor.

Node Anatomy

Every node has a consistent structure:

┌─────────────────────────────┐
│  [Icon] Node Name      [⚠️] │  ← Header with icon and status
├─────────────────────────────┤
│         INPUTS              │  ← Input section
├─────────────────────────────┤
│  ○ Input Anchor 1           │  ← Connection points (left)
│  📝 Parameter 1: [value]    │  ← Configurable parameters
│  📝 Parameter 2: [value]    │
│  [Additional Parameters]    │  ← Advanced settings button
├─────────────────────────────┤
│         OUTPUT              │  ← Output section
├─────────────────────────────┤
│           Output Anchor ○   │  ← Output connection (right)
└─────────────────────────────┘

Node Categories

Language models that power conversations:

ChatOpenAI: OpenAI’s GPT models (3.5, 4, etc.)
ChatAnthropic: Anthropic’s Claude models
ChatOllama: Local models via Ollama
AzureChatOpenAI: Azure OpenAI Service
ChatGoogleGenerativeAI: Google’s Gemini models

Common Parameters:

Model Name: Which model version to use
Temperature: Creativity (0 = deterministic, 1 = creative)
Max Tokens: Maximum response length
Top P: Nucleus sampling parameter

Enable chatflows to remember conversation history:

Buffer Memory: Stores recent messages in-memory
Buffer Window Memory: Last N messages only
Summary Memory: Summarized conversation history
Vector Store Memory: Semantic memory retrieval
Agent Memory: Persistent memory for agents

Configuration:

Session ID: Group conversations by user/session
Memory Key: Variable name in prompts (default: chat_history)
Return Messages: Return as message objects vs strings

Adding and Configuring Nodes

Open Node Palette

Click the + Add Node button on the canvas. This opens the node palette with all available components.

Use the search bar to quickly find nodes by name or category.

Select a Node

Browse by category or search for a specific node. Click to add it to the canvas.The node appears at your cursor position. You can also drag-and-drop from the palette.

Configure Parameters

Click on the node to view its configuration panel:

Required parameters are marked and must be filled
Optional parameters have default values
Additional Parameters button reveals advanced settings
Input anchors accept connections from other nodes

Connect to Other Nodes

Drag from an output anchor to an input anchor:

Click and hold on an output anchor (right side)
Drag to a compatible input anchor (left side)
Release to create the connection
Flowise validates type compatibility automatically

Node Input Types

Nodes accept different types of inputs:

Input Anchors (Connections)

These are connection points that accept output from other nodes:

Displayed as circles on the left side of nodes
Type-checked (e.g., BaseChatModel, Tool, VectorStore)
Can be single or list (accepting multiple connections)
Shows visual feedback on hover (compatible = green, incompatible = red)

Example: A Conversational Agent has input anchors for:

model (BaseChatModel) - single connection
tools (Tool[]) - list, accepts multiple tools
memory (BaseChatMemory) - single connection

Input Parameters (Direct Values)

These are form fields for direct configuration:Text Input: Simple string values

System Message: "You are a helpful assistant"

Number Input: Numeric values with validation

Temperature: 0.7
Max Tokens: 2000

Select/Dropdown: Choose from predefined options

Model Name: [gpt-4, gpt-3.5-turbo, gpt-4-turbo]

JSON Input: Structured data

{
  "key": "value",
  "nested": { "data": true }
}

Code Editor: Write custom functions

// Custom tool implementation
const result = await someOperation()
return result

File Upload: Upload documents, images, or data filesCredential: Securely stored API keys and secrets

Variable Inputs

Parameters that support the acceptVariable flag can use dynamic values:

Reference other node outputs: {{nodeId.data.instance}}
Use global variables: {{$vars.variableName}}
Runtime expressions: {{question}} (user input)

See Variables & Expressions for details.

Node Actions

Hover over a node to reveal action buttons:

Action	Icon	Description
Duplicate	📋	Create a copy of the node with same settings
Delete	🗑️	Remove node and all its connections
Info	ℹ️	View node documentation and details

You can also select nodes and press Delete or Backspace to remove them.

Advanced Node Features

Additional Parameters

Many nodes hide advanced settings behind the “Additional Parameters” button:

Open Additional Parameters

Click the Additional Parameters button at the bottom of the node.

Configure Advanced Settings

A dialog opens with optional parameters:

Timeout settings
Custom base URLs
Advanced model parameters
Retry logic
Headers and metadata

Save Changes

Click Save to apply the settings. They’re stored with the node configuration.

Credentials

Nodes that require API keys use the credential system:

Create Credential

Click Connect Credential on the node
Select Create New from the dropdown
Enter your API key and give it a name
Click Add

Reuse Credentials

Once created, credentials are available across all chatflows:

Select from the dropdown on any compatible node
Update credentials in one place, applies everywhere
Credentials are encrypted at rest

Never hardcode API keys in custom tools or parameters. Always use the credential system.

Node Versioning

Nodes are versioned to support updates and improvements:

Version badge: Shows current node version (e.g., v2, v3)
Outdated warning: Orange warning icon if newer version available
Sync nodes: Button appears to upgrade all outdated nodes
Breaking changes: Review changelog before upgrading

Upgrading Nodes

When a sync button appears:

Click the Sync Nodes button (orange refresh icon)
Flowise automatically updates all outdated nodes
Incompatible connections are removed (review carefully!)
Test your flow after upgrading
Save the updated flow

Note: This operation cannot be undone. Export your flow first as backup.

Custom Nodes

Flowise allows creating custom nodes for specialized functionality:

Custom Tool
Custom Loader
Custom Function

Write JavaScript/Python functions that agents can call:

// Custom Tool Example
// Tool Name: Get Weather
// Description: Get current weather for a city

const axios = require('axios')

const city = $input // From tool parameter
const apiKey = $credentials.weatherApiKey

const response = await axios.get(
  `https://api.weather.com/v1/current?city=${city}&key=${apiKey}`
)

return response.data.temperature

Create specialized document loaders:

// Custom Loader Example
// Load data from custom API

const fetchData = async () => {
  const response = await fetch('https://api.example.com/data')
  const data = await response.json()
  
  return data.items.map(item => ({
    pageContent: item.text,
    metadata: { id: item.id, date: item.date }
  }))
}

return await fetchData()

Transform data between nodes:

// Custom Function Example
// Parse and format LLM output

const llmOutput = $input

const parsed = JSON.parse(llmOutput)
const formatted = {
  response: parsed.text,
  confidence: parsed.score,
  timestamp: new Date().toISOString()
}

return formatted

Connection Rules

Flowise enforces type safety when connecting nodes:

Compatible Connections

✅ Allowed:

ChatOpenAI → Conversation Chain (model input)
Calculator Tool → Agent (tools input)
Vector Store → Retrieval QA Chain (vectorstore input)
Document → Text Splitter → Embeddings → Vector Store

❌ Not Allowed:

Agent → Chat Model (wrong direction)
Text Splitter → Agent (incompatible types)
Multiple outputs to single input (unless input accepts lists)

List Inputs

Some inputs accept multiple connections: The tools input accepts a list of tools, so multiple connections are valid.

Sticky Notes

Add documentation and comments to your canvas:

Add Sticky Note

Open node palette
Go to Utilities category
Select Sticky Note
Place on canvas

Add Content

Click the note and type your documentation:

Explain complex logic
Add TODOs and reminders
Document configuration decisions
Add usage examples

Organize

Resize notes by dragging corners
Color-code by type (coming soon)
Position near relevant nodes

Best Practices

Node Configuration

Use descriptive names for custom tools and functions
Keep tool descriptions clear - LLMs use them to decide when to call tools
Set reasonable token limits to control costs
Use lower temperatures (0.1-0.3) for factual tasks, higher (0.7-0.9) for creative tasks
Test with different model versions to find the right balance of cost and performance

Connection Patterns

Keep data flow left-to-right when possible (easier to read)
Group related nodes together visually
Avoid crossing connections when possible
Use consistent positioning (e.g., all LLMs on the right, all data sources on left)

Performance Tips

Enable caching on LLM nodes for repeated queries
Use appropriate chunk sizes in text splitters (500-1000 tokens)
Configure vector store batch sizes for bulk uploads
Set timeouts on external API calls
Use streaming for better user experience

Get Started

Core Concepts

Building Flows

Deployment

Configuration

Working with Nodes

Node Anatomy

Node Categories

Adding and Configuring Nodes

Node Input Types

Node Actions

Advanced Node Features

Additional Parameters

Credentials

Node Versioning

Custom Nodes

Connection Rules

Compatible Connections

List Inputs

Sticky Notes

Best Practices

Next Steps

Testing & Debugging

Variables & Expressions

Build docs developers (and LLMs) love

Get Started

Core Concepts

Building Flows

Deployment

Configuration

​Node Anatomy

​Node Categories

​Adding and Configuring Nodes

​Node Input Types

​Node Actions

​Advanced Node Features

​Additional Parameters

​Credentials

​Node Versioning

​Custom Nodes

​Connection Rules

​Compatible Connections

​List Inputs

​Sticky Notes

​Best Practices

​Next Steps

Testing & Debugging

Variables & Expressions

Build docs developers (and LLMs) love

Node Anatomy

Node Categories

Adding and Configuring Nodes

Node Input Types

Node Actions

Advanced Node Features

Additional Parameters

Credentials

Node Versioning

Custom Nodes

Connection Rules

Compatible Connections

List Inputs

Sticky Notes

Best Practices

Next Steps