Version 4.9

  • Published on May 13, 2026
  • 2 minute(s) read
  • eA
 Prev Next

Version 4.9

14 May, 2026

Enhancements

AI Agents

  • Conversation Hold Message:
    A new Conversation Hold Message section has been introduced under the Prompts tab, allowing platform administrators to display a message to users when response generation takes longer than expected. The configuration includes a displayed message, a message delay threshold, and a maximum waiting time, giving teams fine-grained control over how the system communicates response delays.

  • GenQ&A Node Deprecation:
    The legacy GenQ&A node has been deprecated in favor of the RAG Agent node, guiding users toward the new architecture for knowledge-based responses. In existing projects, the GenQ&A node is now displayed in a greyed-out "Obsolete Node" state in both the palette and the flow canvas, and a deprecation warning is shown within the node settings indicating the planned removal date. In newly created projects, the GenQ&A node and its settings page are no longer available. Import, export, and copy/paste behaviors have been updated to handle GenQ&A nodes gracefully across project versions.

  • Knowledge Base Reranking:
    A new Reranking capability has been introduced to the Knowledge Base retrieval pipeline, enabling retrieved documents to be resorted and filtered before being sent to the LLM, resulting in more accurate and relevant responses. When enabled, the system retrieves twice the configured document count, reranks them using the selected model, and forwards the top results to the LLM. Reranking is configured under a new "Retrieval Strategy" step in the Knowledge Base edit screen and supports Azure and vLLM as LLM providers.

  • Draft/Published Version of UI:
    A new Draft/Published toggle has been introduced, allowing AI Agents designers to clearly distinguish between the live published version and the editable draft version of a project. When viewing the published version, all project areas, including Project Settings, Flows, Virtual Agent, Intent & Utterances, Responses, System Messages, Guardrails, AI Agents, and Knowledge Bases, are rendered in read-only mode, with editing, creation, and deletion controls removed or disabled.

  • Reveal Characters for Guardrails Entities:
    Guardrails prebuilt entities can now be configured to leave a specified number of characters at the start or end of the detected value unmasked, enabling use cases such as displaying the last four digits of a credit card. The Reveal Characters control is available as an expandable panel on each prebuilt entity row, supports values between 0 and 10, and applies consistently across input and output masking.

Improvements

AI Agents

  • Redis-Based Migration for Remaining Classic Queues:
    The project-published-event and redaction-settings-updated-event queues have been migrated from RabbitMQ classic queues to a Redis-based structure, ensuring uninterrupted message delivery in multi-node high-availability deployments. This change eliminates the single-node dependency of the previous classic queues and reduces the risk of message loss when a node becomes unavailable.

  • Product Branding Update:
    The product interface has been updated to reflect the new Sestek corporate identity across the login screen, About page, and footer.

  • 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.