State Node

​

Node Details

• Name: seqState

• Type: State

• Category: Sequential Agents

• Version: 2.0

​

Parameters

The State node has one main input parameter:

​

Custom State

Users can define the structure of the state object using either a table-based UI or custom code.

​

Option 1: Custom State (Table)

• Type: Data Grid

• Description: Allows users to define the state structure using a table interface.

• Columns:

  • Key: The name of the state variable

  • Operation: The type of operation to perform (Replace or Append)

  • Default Value: The initial value for the state variable

​

Option 2: Custom State (Code)

• Type: Code Input

• Description: Allows users to define the state structure using a JSON object

• Default:

  {
    aggregate: {
      value: (x, y) => x.concat(y), // append new message to existing messages
      default: () => []
    }
  }
  

​

Functionality

1. The node initializes a state object based on the user’s input.

2. For each key in the state:

   • If the operation is “Replace”, it creates a function that replaces the existing value with a new one (or keeps the existing if the new is null).

   • If the operation is “Append”, it creates a function that appends the new value to the existing array.

3. The node can handle both UI-based input and code-based input, allowing for flexibility in defining the state structure.

​

Output

The node outputs an ISeqAgentNode object containing:

• id: The node’s unique identifier

• node: The state object with defined keys and their corresponding update functions

• name: ‘state’

• label: ‘state’

• type: ‘state’

• output: START (a constant from @langchain/langgraph)

​

Use Cases

• Maintaining conversation history in a chatbot

• Tracking progress through a multi-step process

• Aggregating information from multiple API calls or data sources

• Storing intermediate results in complex workflows

​

Notes

• The node uses a sandbox environment (NodeVM) to safely execute user-provided code when using the code-based input option.

• It supports both built-in and external dependencies, which can be configured through environment variables.

• The state object is highly flexible and can be customized to fit various use cases in sequential agent flows.