Version 3.11.2

New Features

New Chunk Visualization Feature: We've introduced a new chunk visualization feature that offers clear, direct insights into the content of each chunk, enhancing content management and oversight. This tool allows users to preview the sources that will be queried, similar to how your RAG assistant operates once deployed.

Search and Browse all the available Chunks of your AI Agent!

Enhanced Hybrid NLU

We have improved the LLM based NLU by adding an intent filtering stage that selects the N best candidates to send to the LLM. This reduces dramatically the inference cost and latency. The figure below show you how it works.

Other Enhancements

API Streamlining: Improved the /dialog API to return a 200 status directly when streaming mode is enabled, ensuring a smoother interaction experience.
Token anonymization: All tokens are now anonymize in the debug logs

Generative AI Components Pagination: Implemented pagination across all Generative AI components, including Assistants, Knowledge Bases, and AI Models, enhancing usability and accessibility.
Webchat SDK: We have added new functions to manage the PopUp. More details in the Webchat SDK page.
New GPT-4 Turbo model: In line with OpenAI's recent announcement, we have upgraded to the latest stable version of the GPT Turbo (aka gpt-4-turbo-2024-04-09).
Model Transparency: We are now surfacing the name of the Generative AI model used to generate a response. This information can be accessed in the Inbox module and directly through the API, promoting transparency and insight into our response generation process.

Fixes

Webchat History Bug Fix: Addressed a bug where the expiration duration of a conversation was not honored, resulting in conversations remaining active beyond their intended lifespan. This fix ensures conversations are correctly archived following the specified duration.
New API Route for Retrieval: Launched a new /retrieve API route, enabling the simulation of document retrieval requests by users on an assistant's vector store, further broadening our platform's functionality.