Skip to main content

Condition Agent Node

The Condition Agent Node is a specialized node that utilizes an LLM Agent to make nuanced, context-aware decisions about the conversation flow. Unlike the Condition Node, which uses static JavaScript expressions, the Condition Agent Node leverages the reasoning capabilities of a language model to dynamically analyze the conversation, extract insights, and determine the appropriate branch for the workflow.

Understanding the Condition Agent Node​

The Condition Agent Node combines aspects of the Agent Node and the Condition Node, offering a higher level of cognitive processing for branching decisions. It's particularly valuable when:

  • The decision-making criteria are complex or subjective, requiring nuanced understanding of language and context
  • The rules for branching cannot be easily expressed as simple conditional statements
  • The decision should take into account the full conversation history and subtle user cues
  • Human-like reasoning is needed to determine the best path forward

For example, a Condition Agent Node might assess whether a customer's issue requires escalation to a supervisor, determine which product category a user is most interested in, or decide if a user's query is related to sales or technical support.

Inputs​

RequiredDescription
System PromptYesA text prompt that defines the Agent's role in making the decision. It should clearly outline the decision-making criteria and the types of outputs expected.
ToolsNoThe Tool Nodes that the Condition Agent Node can access and execute, each offering specific functionality to retrieve information or perform actions that might aid in the decision.
Chat ModelNoA custom Chat Model (LLM) to use instead of the default one defined in the Start Node.
JSON SchemaNoA schema defining the structure of the Condition Agent Node's response when JSON Structured Output is enabled.
Enable JSON Structured OutputNoA configuration to force the Condition Agent Node's response to follow a predefined JSON schema.
Update StateNoA JSON object that defines how the Condition Agent Node should update the custom State before passing it to the next node.

Outputs​

The Condition Agent Node has two possible output connections:

  • True: This path is followed when the agent determines the condition is true.
  • False: This path is followed when the agent determines the condition is false.

Features​

LLM-Powered Decision Making​

The Condition Agent Node uses a language model to make decisions based on:

  • Natural language understanding of the conversation context
  • Reasoning and inference about user intentions, needs, or preferences
  • Pattern recognition in user behavior or conversation flow
  • Decision criteria outlined in the System Prompt

This allows for much more sophisticated branching logic than would be possible with simple conditional expressions.

JSON Structured Output for Clear Decisions​

To ensure reliable and consistent decision-making, the Condition Agent Node typically uses JSON Structured Output with a schema that explicitly captures the decision outcome. A simple schema might look like:

{
    "type": "object",
    "properties": {
        "decision": {
            "type": "boolean",
            "description": "The decision outcome (true or false)"
        },
        "reasoning": {
            "type": "string",
            "description": "Explanation for why this decision was made"
        }
    },
    "required": ["decision", "reasoning"]
}

State Updates for Contextual Information​

In addition to making a decision, the Condition Agent Node can update the custom State to include its reasoning, extracted information, or other contextual details that might be useful for downstream nodes. This allows the decision-making process to contribute to the overall workflow, not just determine its path.

Best Practices​

Craft a clear and focused system prompt

Provide a well-defined persona and clear instructions to the agent in the System Prompt. This will guide its reasoning and help it generate relevant output for the conditional logic.

Structure output for reliable conditions

Use the JSON Structured Output feature to define a schema for the Condition Agent's output. This will ensure that the output is consistent and easily parsable, making it more reliable for use in conditional evaluations.