Version 5.2
23 June, 2026
AI Agents
-
Sync API Timeout Configuration:
A new tenant-level feature has been introduced that allows the Sync API timeout to be configured per tenant. When enabled, administrators can set a timeout value in seconds within tenant settings, ensuring that projects with longer processing times are not interrupted by premature timeouts. The setting is disabled by default and supports values between 30 and 600 seconds. -
Transfer Between Sub-Agents:
A new "Sub-Agent Collaborations" capability has been added to supervisor agents, enabling direct transfers between sub-agents without routing back through the supervisor. Each collaboration is configured between two sub-agents with dedicated transfer instructions, and bi-directional transfers can be enabled when required. This reduces unnecessary requests and lowers token consumption in multi-agent scenarios. -
AI Agent Tool Call Retention:
A new Result Retention option has been introduced for AI Agent tool calls, allowing tool outputs to be removed from conversation history after a defined point. Retention can be triggered after a configurable number of human turns or on agent transfer, with an optional replacement message displayed in place of the removed output. This helps reduce unnecessary token consumption in long-running conversations. -
Proxy LLM Integration:
A new "SESTEK Proxy" provider type is now available in the LLM Integrations page, enabling connection to OpenAI-compatible gateways such as LiteLLM Proxy. By defining a custom proxy endpoint, users can centrally manage provider key handling, rate limiting, model routing, and cost tracking from a single point. -
Insights Conversations Tool and AI Agent Filters:
New "AI Agent" and "Tool" filters have been added to the Insights Conversations screen, allowing conversations to be filtered by the AI Agent that handled them or the tool that was invoked. Tools are grouped under their respective AI Agent, and deleted agents and tools are excluded from the filter lists. This makes it easier to locate specific conversations involving a particular agent or tool.
Improvements
AI Agents
-
Knowledge Base Token Per Minute Guidance:
The documentation link beneath the Token Per Minute field has been replaced with inline guidance text, clarifying that the default value is 30,000 and that token limits are model- and provider-specific. Users are now directed to consult the relevant provider documentation for the correct limits. -
CAI-to-CA Session Correlation:
The relatedGroupId parameter sent to the CA Post Chats endpoint is now populated with the conversation Session ID, enabling messages from the same session that are routed to both the AI Agent and Live Chat to be displayed in a correlated manner within the CA interface. -
Rating Comment and Conversation Link in CA Post Chats:
The CA Post Chats method now includes the rating comment provided by the customer in the AI Agent along with a link to the corresponding conversation. This enables customers using AI Agent and Analytics together to view the rating comment and access the related conversation directly from the CA interface. -
Irrelevant AI Test Criteria Excluded from Calculations:
AI Test evaluations now respect the relevance indicator returned for each success criterion. Criteria flagged as irrelevant are marked as "Irrelevant" in the Simulated Conversations table and excluded from all numerical calculations, preventing success scores from being lowered by criteria unrelated to the conversation. -
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.