Variable Declaration and Assignment

Variable Declaration

This is a space for creating and managing variables needed in the workflow. The created variables can be referenced by other nodes or have their values modified through variable nodes.

Variable Name

A unique name used to identify the variable.

Variable Type

Indicates the data type of the value that will be stored in the variable.

Description

A memo describing the meaning and purpose of the variable.

Default Value

Sets an initial value that will be automatically used when no specific value is assigned to the variable.

Variable Assignment Node

A node that assigns a value to a specific variable within the workflow. It stores a value in the variable so that subsequent nodes can reference it to perform tasks.

Variable assignment is used to manage data flow or store dynamic values that will be reused in later steps.

Variable Name

Specifies the target variable to assign a value to.

You must choose one of the predefined custom variables in the workflow.

Comparison Operator

Defines how the value should be assigned.

Value

The actual value that will be assigned to the variable.

Using Variables in Multi-turn Conversations

Multi-turn conversation refers to an interaction where the exchange does not end in a single question-and-answer step, but instead continues as the agent and user exchange information across multiple turns. In this process, variable declaration acts as a “storage space” that retains collected information so it can be reused in subsequent turns.

The variable assignment node is the step where actual values are stored or updated in this storage, enabling the agent to reuse the information during later steps in the multi-turn process.

Context retention: to use previously received information in the next turn

Flow control: to ask only for missing information and keep the conversation smooth

Reusability: to reference collected values repeatedly across multiple nodes

Example Use Case

Let’s imagine designing an agent that searches for flight tickets while conversing with the user.
Flight search requires information such as departure location, arrival location, departure date, and return date, but users often do not provide all of this information at once.

The agent collects the necessary information step by step through multi-turn dialogue and stores extracted values in variables to use them in the next steps.

User: “I’m planning a trip to Tokyo.”
→ Save Tokyo to the arrival_id variable
Agent: “Which airport are you departing from? Incheon or Gimpo?”

User: “Incheon.”
→ Save Incheon to the departure_id variable
Agent: “When are your travel dates?”

User: “From December 25 to December 31.”
→ Save 2025-12-25 to outbound_date
Save 2025-12-31 to return_date

After this, the agent uses the stored variable values to call the flight search API node and provides the most suitable flight options to the user based on the results.

Preparation — Variable Declaration
Create the following four variables in advance:

departure_id: Departure airport

arrival_id: Arrival airport

outbound_date: Departure date

return_date: Return date

LLM Node → Variable Assignment Node
Ask for each information slot in separate turns and immediately assign the received value to the corresponding variable.

LLM (extract departure_id) → Variable Assignment (assign departure_id)

LLM (extract arrival_id) → Variable Assignment (assign arrival_id)

LLM (extract outbound_date) → Variable Assignment (assign outbound_date)

LLM (extract return_date) → Variable Assignment (assign return_date)

Using Variables — Referencing Variables in the API Node
The four variables are passed as input parameters to the API node.

Variable Declaration and Assignment