SUVA v23.12 Release Notes
Starting from January 4, all SUVA users will have access to the final release of 2023, v23.12. This release marks the culmination of a remarkable year in which Large Language Models (LLMs) have revolutionized the industry and transformed how businesses operate.
SUVA v23.12 introduces a suite of new features, including the Session Feedback mechanism and LLM Error Logs, to elevate your SUVA experience.
The high points of the v23.12 SUVA release are:
-
New Feedback Mechanism: Session Feedback
Session Feedback captures the overall sentiment of a user interacting with your chatbot. It supplements SUVA's existing story-based (non-LLM) and feedback-per-response (LLM) mechanisms.
While Response Feedback provides insights into the effectiveness of each response generated by LLM-powered chatbots, as well as each story of non-LLM chatbots, Session Feedback offers a broader assessment of user sentiment throughout the entire chatbot session.
Administrators can access Session Feedback Analytics through Analytics > Feedback > Session Feedback.
The analytics for response feedback is also available for administrators in Analytics > Feedback > Response Feedback.
Related Doc: In Feedback: Review User Reactions on Chatbot Conversations
-
Large Language Model: Error Logs
Error Logs maintains a comprehensive record of LLM disruptions, empowering admins to proactively address these issues and minimize their impact on chatbot operations. By reviewing the error logs, admins can identify patterns and take preventive measures to ensure the smooth functioning of their LLM(s).
More Info: Large Language Model > Error Logs
-
New Button in Chat Window: Skip Generating
Response generation by LLM-powered chatbots can sometimes take longer than usual, or users might feel that a response is incorrect while it is being generated. To address this concern, the Skip Generating button enables users to halt the response generation process, allowing them to continue with other aspects of their interaction with the chatbot. It's important to note that pressing the button will still result in token consumption.
-
Enhanced Coverage: Admin Logs
The scope of events recorded in the Admin Logs has been expanded to include the Large Language Model section. Consequently, all actions performed in the Large Language Model section are captured in the Admin Logs.
