Skip to main content

​
Overview

The Direct Reply Node sends a response to the user without invoking a language model. It is designed for cases where the response content is known in advance — greetings, error messages, confirmation notices, static instructions, or templated messages that incorporate workflow variables. Because it bypasses the LLM entirely, the Direct Reply Node executes instantly with zero token cost, making it ideal for high-frequency, low-complexity responses.

​
Configuration

{
  "type": "direct-reply-node",
  "config": {
    "reply_mode": "text",
    "reply_type": "static",
    "reply_content": "Thank you for your message. A support agent will follow up within 24 hours.",
    "append_to_history": true
  }
}
ParameterTypeDefaultDescription
reply_modestring"text"Response format: "text" or "markdown"
reply_typestring"static"Content source: "static" (literal text) or "variable" (from context)
reply_contentstring""The response content or variable reference. Allows empty string.
append_to_historybooleantrueWhether to add this reply to the conversation history

​
Reply Types

​
Static Reply

Returns a fixed string. Supports variable interpolation with {{variable_name}} syntax.
{
  "reply_type": "static",
  "reply_content": "Hello {{user_name}}, welcome to our support portal!"
}
Variables in the template are resolved from the workflow context at execution time.

​
Reply Modes

ModeDescriptionRendering
textPlain text responseDisplayed as-is, no formatting applied
markdownMarkdown-formatted responseRendered with headings, lists, code blocks, links, etc.

​
Empty Reply Content

The reply_content field explicitly allows empty strings. This is useful in two scenarios:
  1. Silent acknowledgment — The node executes and appends to history without displaying visible output, useful for internal workflow tracking.
  2. Conditional paths — An empty reply on one branch paired with a substantive reply on another branch.
{
  "reply_type": "static",
  "reply_content": "",
  "append_to_history": false
}

​
Conversation History

When append_to_history is true, the reply is added to the conversation’s message history as an assistant message. This means downstream AI Agent nodes will see this reply as part of the conversation context. Set append_to_history to false when:
  • The reply is a transient status message (e.g., “Processing your request…”)
  • The reply contains internal metadata not relevant to the LLM
  • You want to avoid polluting the conversation context

​
Use Cases

Greeting Messages

Welcome users with a consistent, branded message before routing to an AI agent.

Error Handling

Return friendly error messages on failed branches without consuming LLM tokens.

Confirmation Notices

Acknowledge user actions (“Your file has been uploaded successfully”) instantly.

Fallback Responses

Provide a default response when no conditions match or the knowledge base returns no results.

​
Example: Conditional Greeting

A workflow that greets the user differently based on the time of day:

​
Example: Error Fallback

Use a Direct Reply Node as the default output of a Condition Node to handle missing search results: 
{
  "type": "direct-reply-node",
  "config": {
    "reply_mode": "markdown",
    "reply_type": "static",
    "reply_content": "I'm sorry, I couldn't find any relevant information for your question. Please try rephrasing or contact our support team at **support@example.com**.",
    "append_to_history": true
  }
}

​
Best Practices

If the response text is known at design time, always use a Direct Reply Node instead of an AI Agent Node. It is faster, cheaper, and deterministic.
When the reply includes links, lists, code snippets, or formatted text, set reply_mode to "markdown" for proper rendering in the chat interface.
Only append to history when the reply adds meaningful context for future LLM calls. Transient messages like “Please wait…” should not be appended.

​
Next Steps

Start & End Nodes

Configure workflow entry and exit points

Condition Node

Branch logic to decide which reply to send

Form Node

Collect structured input from users

AI Agent Node

Use LLM-powered responses for dynamic content