Skip to main content

​
Introduction

Beyond the prompt and tools, you can customize multiple technical aspects of how your agent handles calls. This configuration allows you to adapt the agent’s behavior to different contexts: adjust the voice volume, configure reminders to keep users engaged, and set time limits to protect your budget.
Customization ScopeAll these settings apply globally to the agent, regardless of whether you use Single Prompt or Conversational Path mode. You don’t need to adjust them in each node individually.

​
Voice Parameters

​
Language

The language setting controls which languages your agent understands and speaks during a call. It affects both speech recognition and voice synthesis, so it should match the language your callers are likely to use. You can configure it in two ways:
  • Multilingual (default): the agent automatically detects the language each caller speaks and replies in that same language, switching between languages mid-call if the caller does. Use this when you can’t predict your callers’ language or you serve a multilingual audience. In this mode the agent supports every language in the list below.
  • A single language: the agent always understands and speaks in the one language you choose, regardless of what the caller uses. Use this when all your calls happen in the same language and you want the most consistent results.
Screenshot showing language configuration in the agent panel
If the language you need isn’t listed as a single-language option in the selector, choose Multilingual — the agent still understands and speaks it. For example, Basque (euskara) isn’t available as a fixed single language, but it works in multilingual mode.
Your agent can understand and speak the following languages:

​
Volume

Controls how loud the agent’s voice is during the call.
  • Value range: 0 to 100
  • Default value: 50

​
Background sound

Adds an ambient sound while the agent performs actions or waits for a response, helping to avoid long silences during the call.

​
Inactivity Detection System

The inactivity detection system helps keep the conversation active when the user stops responding. This mechanism is fundamental to avoid long silences that end the call or create a poor experience.

​
How it Works

When the user stops speaking for a certain time, the agent sends a reminder message to verify they’re still on the line. If after several reminders there’s no response, the call is automatically terminated.
Flow diagram showing the inactivity detection system with progressive reminders

​
Configurable Parameters

​
Reminder Messages

Define what phrases the agent will use to check if the user is still on the line. You can define:
  • Fixed messages: A list of specific phrases the agent will use randomly
  • Generated messages: The agent will generate natural messages based on context. This is the recommended approach for a more human experience.

​
Reminder Frequency

Defines how often (in milliseconds) the agent should send a reminder if it detects no activity.
  • Default value: 7500 (7.5 seconds)
  • Recommended range: 5000 to 15000 ms
Avoid very low values: Frequencies below 5 seconds can cause the agent to interrupt the user while they’re thinking of their response, creating a frustrating experience.

​
Maximum Number of Reminders

Specifies how many reminders the agent will send before ending the call due to inactivity.
  • Default value: 3
  • Recommended range: 2 to 5