‎Conversation Settings: Advanced Controls

Conversation Settings: Advanced Controls

Updated

5 days ago

For Voice Bot applications, configure the following settings:

Global Repeat Intent: Select an intent that enables the AI Agent to repeat its last response when triggered.
TTS Fallback IVR: Select a workflow to execute if the Text-to-Speech (TTS) service is unavailable or fails.
Speech Profile: Select the speech profile for the Voice Bot. This field is mandatory.
Barge-in Functionality
Enable users to interrupt an ongoing voice prompt with their response. You can configure:
- Minimum Response Completion (%): The minimum percentage of the AI Agent's response that must be played before user input is accepted.
- Minimum Response Duration: The minimum time the response must play before interruption is allowed.
When both values are configured, the higher threshold is applied.

Note: Barge-in settings configured at the AI Agent Reply node level override the global barge-in configuration.

Under Feature Suite, enable the following messaging capabilities as required:

Typing Indicator: Displays a typing indicator while the AI Agent is generating a response.
Typing Indicator Text: Customize the message displayed in the typing indicator.
Stream Response: Streams LLM-generated responses in real time within Live Chat. This capability is supported for LLM-powered features such as Dynamic Workflows and Smart FAQ.
Multiple Fan Messages: Controls whether users can send multiple messages while the AI Agent is processing an earlier request. When disabled, users must wait for a response before sending another message.

Note: When Stream Response is enabled for Smart FAQ:

Responses are streamed and displayed in real time as they are generated.
Output guardrails and post-processing for generative AI responses are not applied.
Configuration of the AI Agent Response variable is not required because the response is automatically published by the node.

Configure the following settings to manage model behavior and compliance:

Select Guardrail: Apply guardrails to user inputs and AI Agent responses to help enforce compliance and safe interactions.
PII Masking Template: Select the template used to identify and mask personally identifiable information (PII) within conversations.
Select LLM Provider: Choose the large language model (LLM) provider used by the AI Agent.
Model: Select the model that generates AI Agent responses.

Enable Additional Settings to configure the following model parameters:

Max Tokens: Specifies the maximum number of tokens allowed across the request and generated response.
Temperature: Controls response variability. Lower values produce more predictable responses, while higher values generate more diverse outputs.
Frequency Penalty: Reduces repetitive content by penalizing frequently used tokens.
Presence Penalty: Encourages the model to introduce new concepts by penalizing tokens that have already appeared in the response.