Send Follow-Up

Description

This module generates and prepares follow-up messages to send to the user when they have not responded. It supports three message generation modes:

1. AI (ai): Generates the message automatically using an LLM. The system prompt instructs the AI to create short, natural, and non-invasive messages adapted to the follow-up number. It can reuse the brain configured in the session’s agent or use its own.
2. Template (template): Uses a predefined message with dynamic variables such as {{name}}, {{last_topic}}, {{follow_up_number}}, etc. Variables are replaced with session data and extracted entities.
3. Dynamic (dynamic): Takes the message directly from the input data (message, text, or content).

The module optionally:

• Adds the follow-up message to the session history.
• Updates session counters and timestamps in the database.

If AI generation fails, a predefined fallback message is used based on the follow-up number.

Configuration

Parameter	Type	Required	Description
message_mode	select	Yes	How the follow-up message is generated. Options: ai, template, dynamic. Default: ai.
template_message	textarea	No	Template message. Supports variables like {{name}}, {{last_topic}}, etc. Visible in template mode.
ai_prompt	textarea	No	Instructions for the AI on how to generate the follow-up message. Visible in ai mode.
use_session_brain	boolean	No	Use the same AI configured in the session’s agent. Default: true. Visible in ai mode.
brain_provider	select	No	AI provider for generating the message. Visible if use_session_brain is false.
brain_model	aiModelSelector	No	AI model to use. Default: gpt-4o.
credentials_key	credentials	No	AI provider credentials. Visible if use_session_brain is false.
max_tokens	number	No	Token limit for the generated message. Min: 50, Max: 1000. Default: 200. Visible in ai mode.
temperature	number	No	Message creativity (0.8 recommended for follow-ups). Default: 0.8. Visible in ai mode.
include_context	boolean	No	Include conversation context to generate a more relevant message. Default: true.
update_session	boolean	No	Add the sent message to the session history. Default: true.

Credentials

Optional. If use_session_brain is true, the agent’s credentials are reused. Otherwise, credentials_key must be configured.

Output

{
  "success": true,
  "message": "Hola! Vi que estabamos conversando. Hay algo en lo que pueda ayudarte?",
  "session_id": "uuid-sesion",
  "channel": "whatsapp",
  "external_id": "5491112345678",
  "follow_up_number": 1,
  "message_mode": "ai",
  "generated_at": "2026-03-23T10:00:00.000Z"
}

Usage Example

Basic case

{
  "message_mode": "template",
  "template_message": "Hola {{nombre}}, queria retomar nuestra conversacion sobre {{ultimo_tema}}. Puedo ayudarte en algo mas?",
  "include_context": true,
  "update_session": true
}

Expected input data

• session_id: Agent session ID.
• last_topic: Last conversation topic.
• last_user_message: Last user message.
• memory: Session memory.
• entities: Extracted entities (name, email, etc.).
• follow_up_number: Follow-up number (1, 2, 3…).
• channel: Communication channel.
• external_id: User’s external ID.

API Used

In AI mode, the configured provider’s API (OpenAI, Anthropic, Google, etc.) is called to generate the follow-up message.

Notes

• Predefined fallback messages are: (1) “Hi! I noticed we were chatting…”, (2) “I just wanted to check if you have any pending questions…”, (3) “I hope everything is going well…”.
• In template mode, unreplaced variables are removed from the final message.
• A temperature of 0.8 is recommended for follow-ups, as it generates more varied and natural messages.
• The module does not send the message directly; it generates the text for a subsequent node (e.g. Telegram, WhatsApp) to send it.

Related Nodes

• Agent - AI Agent
• AgentDecision - Agent Decision