Version 3.5

Version 3.5

8 July, 2025

Enhancements

VA

• Adding SSML Support:
  SSML (Speech Synthesis Markup Language) support has been added to the Virtual Agent to enable richer TTS experiences. An SSML toggle is now available in the Project Settings and when enabled, a new SSML toggle appears within the text response sections of Say, Ask, Ask Missing Slot, Ask Attachment, and Confirmation nodes. Designers can add either TTS text or upload voice recordings with fallback text. SSML input is used in TTS requests, while the main text field is shown in the Chat UI and reporting.

• Static Masking (Text):
  The feature previously referred to as redaction in the UI is now renamed to masking for clarity and consistency. This static masking functionality applies to text-based interactions in the Virtual Agent. After a conversation ends, sensitive information (such as personal data or ID numbers) is automatically detected and masked at the database level, enhancing data privacy and ensuring compliance with security standards.

Improvements

VA

• Agentic AI:

  • Streaming Support:
    Streaming capability has been added to the AI Agent node for a more interactive and low-latency chat experience. Two new toggles are available under the node’s advanced settings: “Stream Response” and “Show Tool Call Response During Execution.” While the first toggle is used for enabling streaming feature, the second toggle is used for displaying tool_call_output during internal tool executions like agent transfer or end conversation.
  • Preparing All AI Agents:
    To improve efficiency in AI Agent management, a new “Prepare All Agents” option has been added under the Prepare button on the AI Agent details page. The button has been updated with a dropdown (chevron) menu. When clicked, it reveals the Prepare All Agents option, allowing scenario designers to prepare all sub-agents linked to a supervisor agent in one action.
  • Displaying Status on AI Agents Page:
    To help designers easily identify the preparation status of agents, visual status indicators have been added to the AI Agents and Agent Details pages. Instead of icons, the status is now displayed as a colored ring around each agent’s avatar: green for Prepared, grey for Preparation Required, and orange for Preparation Failed. Hovering over the avatar displays a tooltip with the status description. On the Agent Details page, the preparation statuses of all linked sub-agents are also shown with the same visual indicators.
  • Vllm Support for On-Prem Deployments:
    To meet regulatory requirements and ensure full data privacy, customers can now use Sestek's VLLM provider for on-premise deployments of Agentic AI. The prepare endpoint has been extended to support VLLM-specific parameters under provider_settings. Additionally, the LLM Configurations page now includes fields for Sestek’s VLLM provider; API Key and API Base parameters. This allows enterprises to run AI Agents securely without sharing end-user data with third parties.

• Minor UI Improvements:
  Several minor improvements have been made to the user interface to enhance user experience and usability.

• Minor Bug Fixes:
  Various minor bugs have been addressed to improve overall system stability and performance.