ConvyAI

Release notes


Release notes

In this section, you will find all the release notes with details of what each one contains.

 

6.25.0

New features

  • Starting today, you have the option to enable the file search functionality during conversations between your users and our generative AI. This new feature allows you to upload documents directly into the chat, enabling our AI to interpret and analyze them in real-time.

 

Enhancements

  • Optimizations have been applied to make the customer interface lighter when included in web pages, increasing insertion speed and minimizing generated traffic.
  • Optimizations have been applied to prevent the disconnection of web consoles, both admin and agent desktop, in case of user inactivity.

Bug fixing

  • Resolved an issue during the routing phase of conversations coming from social media channels, which caused slowdowns or conversations not being queued correctly.
  • Resolved an issue with the use of System Jumps applied during communication with the Generative node.
  • Resolved an issue with intent reporting in cases where the intent was not recognized by the semantic engine.

 

6.24.0

New features

  • Drag & Drop and Multiple File Sending. The drag-and-drop feature allows agents to easily upload one or more documents into the chat window. Operators can then choose which files to send, offering flexibility and control over the uploaded content. It is also possible to use copy and paste tools to attach files or screenshots taken directly from the desktop.

 

Enhancements

  • Generative long memory. The long memory functionality of generative models allows them to retain context and information over extended interactions, enhancing their ability to provide coherent and contextually relevant responses throughout the conversation.

  • New Generative AI Model - OpenAI Gpt-4o. Integrating the GPT-4o model into ConvyAI's omnichannel process automation platform enhances customer interactions with advanced natural language understanding, enabling more accurate, efficient, and personalized responses across multiple communication channels.

Bug fixing

  • Fixed incorrect handling of email attachments in drafts within AgentDesktop

  • Resolved anomaly during email sending

  • Fixed anomalous data during data extraction in the Contact Center Statistics section

  • Fixed anomaly during the viewing of the Survey section on AgentDesktop

  • Resolved anomaly when viewing WhatsApp queued conversations on AgentDesktop

  • Improved rendering of received links on the client interface

  • Fixed failure in sending scheduled reports when special characters were present in the extracted conversation

  • Resolved anomaly causing duplication of outbound conversations with multiple WhatsApp connectors

  • Resolved anomaly when indexing PDF documents in the knowledge base

  • Fixed anomaly in handling special characters during semantic engine reporting extraction

  • Fixed anomaly related to incorrect retention of the previous context when arriving at a new generative node

  • Resolved anomaly causing incorrect retention of the previous context when arriving at a new generative node after performing a SYSTEM_JUMP

6.23.0

New features

  • Suggestions and Private Notes. Starting today, supervisors can instantly access ongoing conversations between operators and customers and, when necessary, intervene by providing messages or suggestions directly to the operators. This feature allows supervisors to offer real-time coaching and support, ensuring that agents are equipped with the information and guidance needed to deliver exceptional customer experiences.

  • Suggestion Approval. This new feature allows operators to receive suggestions via private notes from supervisors during conversations with customers. Operators can approve the suggestions and send them to the customer after careful review. This functionality enhances customer support and promotes collaboration between supervisors and operators.

  • New embedding model - Cohere. From now on, you can leverage the indexing capabilities offered by the embedding model developed by Cohere, the cohere.embed-multilingual-v3, made available in ConvyAI thanks to the native integration of the AWS Bedrock stack.

Enhancements

  • Hazelcast Client Libraries Update

  • Critical Vulnerability Update for npm Package xlsx

  • We have clarified the data presented for outbound conversations within ConvyAdmin reports.

  • We have implemented improvements in the routing of outbound conversations, providing more accurate management during the routing design phase.

Bug fixing

  • Correction of Outbound Message Sending Failure with Multiple WhatsApp Connectors

  • Correction of an Exception During User Logout