Skip to content

4.80

Upcoming Feature Removals

  • The Chatbase analytics integration will be removed in version 4.81. The Chatbase platform has not been supported since 2021.
  • The Any Slots feature will be removed in version 4.83. Instead, utilize the LLM Entity Extraction Node to identify Slots in the user input via LLMs.
  • The facebook-bot-messenger module within a Code Node is deprecated and will be removed in version 4.83.

Security Changes and their Impact

  • While addressing security vulnerabilities, changes in third-party vocabularies may cause some exact matching cases to fail, but conversations should continue uninterrupted. Re-training the model will resolve exact matching issues. If persistent errors occur, contact Cognigy technical support.
  • While addressing security vulnerabilities, changes in third-party vocabularies may cause slight variations in tokenization for certain locales. For example, in French, hyphens may now be included as part of suffixes in tokens. Variations in token outputs are expected. If you encounter any significant issues, contact Cognigy technical support.

Resolving Connection Errors in Microsoft Azure OpenAI LLMs

In recent releases, we have updated the connection settings to Azure OpenAI LLMs and added new parameters, such as the Resource Name. If you have an older connection (for example, created in the 4.53 release) to Azure OpenAI LLMs, especially Azure OpenAI text-embedding-ada-002 for knowledge search features, you might encounter the following error when an LLM is triggered: Error while performing knowledge search. Remote returned error: Search failed: Could not fetch embeddings due to missing API resource name for Azure OpenAI. To resolve this issue, recreate the LLM and the connection so that both have the latest format.

4.80.1

Infrastructure

Released July 10th, 2024

Cognigy.AI + Insights + AI Copilot

A typo was fixed in the globalAzureDocumentIntelligenceConfig attribute in the Cognigy.AI Helm chart. With the typo corrected, this release requires the system-wide Azure AI Document Intelligence configuration to be set up as follows:

knowledgeSearch:
  enabled: true

  globalAzureDocumentIntelligenceConfig:
    enabled: true

4.80.0

Cognigy.AI

Released July 9th, 2024

Features

Improvements

  • Added German localization to the product
  • Redesigned the Project list in the Cognigy.AI UI for a more convenient way to switch between Projects
  • Included the Max knowledge chunks field in the Change Organisation Quotas section of the management-ui interface
  • Added the authorization method OAuth 2.0 for Microsoft Azure OpenAI
  • Added the Embedding type of model for custom LLMs
  • Implemented the Stream Reprompt to Output option within the LLM Prompt method of the Question Node

Bugfixes

  • Fixed the issue in the Cognigy.AI OpenAPI documentation where the description of the entrypoint parameter was incomplete for Endpoints
  • Fixed the issue of disconnect events not being sent when the Async Mode setting was enabled in the AudioCodes Endpoint
  • Fixed the issue where fetching audit events would result in a detailed error when retrieval fails
  • Fixed the issue where skipped or ignored trainer records from Snapshots were not displayed when filtering using Snapshots
  • Fixed the issue where conversations could end up in an unresponsive state
  • Fixed the issue where textCleaner removes the leading zeros from the phrase

Cognigy Voice Gateway

Released July 9th, 2024

Features

  • Added the Silence Overlay feature. You can configure this feature in the Voice Gateway Set Session Config Node. This feature allows you to play audio to the caller while the speech input is being transcribed

Improvements

  • Changed the behavior of the Barge In setting in the Voice Gateway Set Session Config Node for calls transferred to the contact center. Previously, Barge In was always active regardless of its setting. Now, if Barge In is turned off, callers cannot interrupt conversations. If it's turned on, callers can interrupt conversations
  • Added support for REGISTER with SIP over TLS using the SIP scheme. Previously, only the SIPS scheme was supported. Now, SIP is the default for all new Carriers; existing ones remain the SIPS scheme. If you want to add encryption to make outbound calls more secure, activate the use of SIPS in the Voice Gateway Self-Service Portal by navigating to Carriers > SIP Gateways > Use sips scheme

Bugfixes

  • Fixed the issue where the TRANSFER_REFER_SUCCESS event was missing after a successful REFER with final_referred_call_status: 410 (Gone). This event is now properly handled, including a correct value for handover escalations (Analytics)
  • Fixed the issue where an error log was generated when REFER with final_referred_call_status: 410 (Gone) occurred
  • Fixed the issue where a timeout is triggered when a message is sent to Cognigy.AI. The issue happens in scenarios where no response was expected. The solution now ensures the timeout is only triggered when ENDPOINT_VG2_TIMEOUT_FOR_AI_RESPONSES_IN_SECONDS is set
  • Fixed the issue where the code was missing to support testing STT for on-premises Microsoft speech service with a private endpoint
  • Fixed the issue where the default model for speech-to-text using Google was incorrect. Leading to possible unsupported languages
  • Fixed the issue of streaming failures with the Microsoft Speech Services provider

Cognigy Live Agent

Released July 9th, 2024

Bugfixes

  • Fixed the issue where removing a human agent from an inbox caused conversations outside of that inbox to be unassigned

Cognigy Insights

Released July 9th, 2024

Improvements

  • Added information about Privacy Policy acceptance to the detailed view of the Transcript Explorer
  • Improved by including the agentName field in the Live Agent Escalations model, allowing the OData LiveAgentEscalations endpoint to display this information

Bugfixes

  • Fixed the issue where attachments from the Live Agent were shown incorrectly in the Transcript Explorer

Infrastructure

Cognigy.AI + Insights + AI Copilot

New Azure AI Document Intelligence Configuration

The mechanism for the system-wide Azure AI Document Intelligence configuration has been updated. The following environment variables are no longer used:

  • FEATURE_ENABLE_AZURE_DOCUMENT_INTELLIGENCE_ORG_WHITELIST
  • AZURE_DOCUMENT_INTELLIGENCE_ENDPOINT
  • AZURE_DOCUMENT_INTELLIGENCE_APIKEY

These environment variables, which were configured for task-process-knowledge-source-file, should be removed. Instead, the following section should be added under the knowledgeSearch section:

  gloabalAzureDocumentIntelligenceConfig:
    enabled: true

For example, on a system where the Knowledge AI feature is enabled, the configuration would be the following:

knowledgeSearch:
  enabled: true

  gloabalAzureDocumentIntelligenceConfig:
    enabled: true

The cognigy-azure-ai-document-intelligence-credentials parameter should be configured to contain the required API key and Endpoint URL.

Rollback Instructions

In case you need to revert to a version below 4.80.0, run the following script as we've added new fields to the Project Settings in Cognigy.AI.

  1. Establish a connection to the MongoDB instance within your cluster and execute the database script:

    kubectl exec -it -n mongodb mongodb-0 -- bash
    mongo -u $MONGODB_ROOT_USER -p $MONGODB_ROOT_PASSWORD --authenticationDatabase admin
    
    use service-resources;
    db.agentsettings.updateMany({},{$unset:{"generativeAISettings.useCasesSettings.answerExtraction":""}});
    
  2. Execute the Helm Chart rollback.

Cognigy Voice Gateway

  • Upgrade postgres-exporter to 0.15.0-debian-12-r17
  • Upgrade postgres to 14.12.0-debian-12-r15
  • Update mysql to 8.0.37-debian-12-r5

  • Increase the resources of postgresql:

    • CPU from 0.5 to 1.
    • Memory from 3Gi to 4Gi.
    postgresql:
      primary:  
        resources:
          requests:
            cpu: "1"
            memory: 4Gi
          limits:
            cpu: "1"
            memory: 4Gi
    

Version Compatibility Matrix

The version compatibility matrix was updated for the following Cognigy products: