Version 3.0

Version 3.0

28 April, 2025

VA

Enhancements

• Agentic AI:
  We are excited to announce the introduction of Agentic AI in Knovvu Virtual Agent! This new capability empowers users to create intelligent AI Agents that can operate autonomously, perform tasks, and escalate to sub-agents internally when needed.
  Key highlights include:

  • Tenant Feature Activation: Agentic AI feature are enabled per tenant.
  • AI Agents Page: A new "AI Agent" tab has been added within the Projects, enabling users to view and create agents.
  • AI Agent Creation Wizard: A step-by-step wizard that guides users through the agent creation process. It includes:
    • General Settings: The agent’s avatar, name, and description can be entered.
    • Interaction Settings: The agent’s persona, creativity level, response detail level, and instructions can be selected.
    • LLM Settings: The LLM provider and model can be configured.
  • Agent Details Page: A dedicated page where existing agents can be edited. Users can update general settings, interaction settings, and LLM configurations. Additionally, new tools can be added to extend the agent’s capabilities, and sub-agents can be created and linked for more complex, multi-agent workflows.
  • Tool Parameter Context: A new context type has been introduced to define parameters that can be used within tools. Tool parameters can now be utilized in both HTTP Request and Action Flow tool types.
  • AI Agent Node: A new node that allows triggering AI Agents directly within conversation flows. This node enables the Virtual Agent to delegate specific tasks to an AI Agent during a conversation. Users can select which AI Agent to invoke, pass inputs if needed, and handle the AI Agent's response to continue the conversation seamlessly.
  • End with Result Node: A new node has been added to the VA node palette to terminate the tool action, allowing the AI agent to be informed that the tool flow has ended.
  • TestChat Debug Mode: TestChat Debug Mode: AI Agent messages are now visible in TestChat to enhance testing and debugging. Two options are available to display tool call details and the responder AI Agent name.
  • Versioning with Project Publish: AI Agent versions are now managed together with project publishing, ensuring consistency between project updates and agent changes.

• Daily Conversation Report API:
  We’ve introduced the Daily Conversation Report API that enables automated generation, storage, and retrieval of hourly conversation data per tenant. Every night, the system automatically generates reports for tenants who have this feature enabled, storing them securely for on-demand access. The reports can be fetched in hourly chunks, giving you flexibility to access the data in smaller increments. You can retrieve these reports in either a compressed ZIP file or an easy-to-read JSON format, depending on your preference. This feature ensures that only authorized users from your tenant can access your data, with role-based security in place to protect sensitive information. Whether you're tracking conversations for analysis or support, this API provides a simple, reliable way to access the data you need. For more information, please check the architecture overview. For the technical details, please check the API Reference. The feature can be enabled through tenant features, so if you'd like to utilize this functionality, please contact SESTEK.

• Channel Testing in Draft Mode and Separate Channel Management:
  With this update, VA Designer users can now test their projects on channels directly in draft mode, without the need to publish. Changes made in draft can immediately be tested on the relevant channels, allowing users to verify integrations and flows without initiating a full publish cycle after each update.
  Additionally, channel and integration management for draft and published versions has been fully separated. Users can independently add, edit, or remove channels and integrations for draft projects without impacting the published version. Any modifications in draft mode will remain isolated and will not affect the published project.
  If a user modifies an integration setting that could impact the published version, a warning pop-up will be displayed. These warnings are dynamically shown in English, Turkish, or Arabic, based on the user’s language preferences.
  Furthermore, during the publish process, a new interface will display both the draft and published channel configurations. This allows users to clearly review and manage which channels are active in each version before confirming the publish action.
  This enhancement is designed to accelerate development processes while maintaining the stability and reliability of live projects.

Improvements

VA

• Separation of VA Pods for Draft and Published Versions:
  We have implemented the separation of VA pods for published projects and drafts, utilizing two distinct inference pods: one for the draft and one for the published version. Changes made in the draft pod do not affect the published version. Once changes are made in the draft and published, any further changes made in the draft pod will only impact the draft version, while the published version remains unaffected and isolated.

• Concurrent Login Management:
  The Concurrent Login Management feature is now available, enabling system administrators to restrict or disable simultaneous logins to enhance account security. The feature can be configured in three ways:
  • Disabled: Users can log in simultaneously from multiple devices without any restrictions. No session management is enforced, and all sessions remain active across devices.
  • Logout from Same Type Devices: When a user logs into one device (e.g., desktop or mobile), active sessions from the same device type will be terminated. Sessions on different device types (e.g., desktop and mobile) can remain open simultaneously.

• TestChat & WebChatComponent Change:
  TestChat and WebChat components are enhanced to properly display markdown-formatted messages received from LLMs. This update ensures that content is presented in a more organized, readable, and user-friendly manner, providing a cleaner and more intuitive experience for users.
• TTS Integration UI Changes for Elevenlabs & Azure:
  The TTS Integration UI for ElevenLabs and Azure has been enhanced to allow users to enter and select custom voices from these providers, offering greater flexibility and personalization.
• Azeri NLU Language Support:
  This development introduces support for Azeri language in the VA system, including enhancements for natural language understanding (NLU), text normalization, and sentence embedding, enabling more accurate and effective language processing for Azeri.

• LLM Model Updates:
  OpenAI Models: The following models have been added:

  • chatgpt-4o-latest
  • gpt-4.1
  • gpt-4.1-mini
  • gpt-4.1-nano
  • o3
  • o4-mini

• Gemini Models: The following models have been added:

  • gemini-2.5-pro-preview-03-25
  • gemini-2.5-flash-preview-04-17

• Gemini Models: The following models have been removed:

  • gemini-1.5-flash-8b
  • gemini-1.5-flash
  • gemini-1.5-flash-pro

• 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.