A new callBOT command for seamless GPT calls from the command state has been introduced.

It can be used as follows:

• Simple: /control:runRoutingApplication,callBOT → the bot will receive a /callBot message.
• Advanced: /control:runRoutingApplication,callBOT#<message with placeholders $> → the bot will receive the message after the # with the placeholders replaced from pdata.

GPT/RAGBOT now supports sending interactive buttons and images for a richer user experience.

Version 4 of the portal can now embed pages from version 3 of the portal without the necessasity of logging in again. This functionality is available to beta testers of the new portal only.

The chat CSS has been updated to a new style conforming with the design of the new portal.

Some PBXs cannot detect an anonymous number correctly, and send the dial prefix along with an empty number (i.e. the telephony server receives a call from 0). For the incoming number converter “Telephone 0, 00” the behaviour has been changed, so that if only a 0 is received, it is assumed to be anonymous.

It is now possible to control the rendering of live agent services such as AI transcription, AI summary or AI suggestions. Each service has its own resource.

• portal.Acd.AgentHome.LiveAgent.AI.Transcription - render the AI Transcription section
• portal.AgentHome.LiveAgent.AI.Assistant - render the AI Assistant services like Suggestion, Summary and Sentiment.
• portal.Acd.AgentHome.LiveAgent.AI.Assistant.Caller - render the caller’s sentiment.
• portal.Acd.AgentHome.LiveAgent.AI.Assistant.Agent - render the agent’s sentiment.
• portal.Acd.AgentHome.LiveAgent.AI.Assistant.Sugesstion - render the AI suggestion section.
• portal.Acd.AgentHome.LiveAgent.AI.Assistant.Summary - render the AI Summary section

A new API endpoint has been added to the REST server, to retrieve dialler results in a new aggregated form for a list of contact IDs or a list of UserData entries.

See this page for more details:

https://wiki.jtel.de/display/JPW/Dialler+REST+API+-+Dialler+Contacts+-+Extended+Retrieval

The chat server and the chatbot have been updated to integrate RAG functionality. This enables the use of a RAG knowledge base for querying LLMs such as GPT or Mistral.

Note: the preparation of the RAG database involves the integration of custom processes to integrate the data. Please contact jtel if you require the integration of RAG functionality.

Two new settings are available in the configuration:

• UnknownMapsToTelStatus
  • This defines what status should be set, when “Unknown” is received from the AMI
• UnavailableMapsToTelStatus
  • This defines what status should be set, when “Unavailable” is received from the AMI

Allowed values for both settings are:

• NotPresent (which means grey / not registered)
• Hangup (which means green / idle)
• Ringing (which means yellow / ringing)
• Up (which means red / busy )

If not configured, NotPresent is used.

In all cases, the default behaviour has been changed, so that NotPresent maps to Not Registered (grey) in the jtel portal.

A new variable $calltime has been added to the IVR engine. This contains the number of seconds since the call went offhook, or 0 if the call is not yet offhook.

A new configuration option has been added:

timestampsSourceOfTruth: systemClock

The available options are:

• systemClock
• hardwareClock

The standard value is systemClock. This changes the behaviour from the previous behaviour, which was hardwareClock.

Notes

• Some VM installations do not correctly synchrronise the hardware clock, causing timestamps to vary from the systemclock. The default behaviour was therefore changed to use the system clock, which in all installations should be setup to synchronise from an NTP server. This behaviour should be acceptable for most installations.
• In installations which require certification for billing purposes, this could cause certification problems. If this is a requirement, it is recommended to ensure that the hardware clock is synchronised to a high quality source (such as a GPS source) and use this setting instead.

A bug was identified when choosing the destination variable in the IVR object connect variable, which caused the call to fail and throw an error. This problem has been fixed.

In the dialer, the maximum tries of call attempts per contact was not being enforced, allowing contacts to be called even after reaching the limit. This problem has been fixed.

Previously, when an agent reserved a contact and initiated a call, if the contact did not answer, the contact was released immediately. With this fix, the contact remains reserved until a result code is set or the maximum number of busy call attempts is reached.

Events generated by the chatbot / chat server were not always closed if the user refreshed the page. This problem has been fixed.

If a caller hung up the phone during the call recording announcements, the call could “stick” in the wallboard and still be visible. This was due to the incorrect execution of the end call routing, even though the call was correctly disconnected, the statistics and status of the call were not updated. This problem has been fixed.

The IMAP mail retrtieval process does not stop processing if an error is received. Instead, it will now repeat the process (possibly indefinately).

Error messages are now sent by the process to the configured alerting email account in the client master data settings.

If a temporal error occurs, the system will try the process one more time, before generating an error email.

An “OK” email is generated, once the problem is resolved.

Copying a user did not copy all parameteres when selecting the ACD Option. This problem has been fixed.

An exception was produced in the portal when no summary was produced for a call. This usually was a short call, where no relevant utterances were made by the caller or the agent. This problem has been fixed.

The Availability column in the PBX tab of MiniClient was previously displaying only the Telephony status instead of the correct combined status from Telephony and Teams.

This issue has now been fixed.

Previously, chat sessions did not expire if the user kept their browser open, leading to sessions remaining active indefinitely.

This issue has now been resolved by introducing an idle expiration mechanism. A configurable idle timeout parameter (ACD.Chat.Expiration.IdleTime) has been implemented, ensuring that inactive chat sessions expire after the defined period. When a session expires, the system sends a "bye" message (configured in the ChatConnector) to inform the user, and the session is properly closed, preventing further input.

The blazar server would not detect an empty music on hold setting when using the connect to external destination if no music on hold was specified, and would go into an infinite loop filling up the disk. This problem has been fixed.

As a workaround if your blazar server has not been updated, always ensure that the music on hold setting specifies a valid file. If you do not require music on hold, then set to an empty “silence” file.

While the Threshold-Based Tiles Coloring feature worked seamlessly within the supervisor view, it did not function correctly in the independent Wallboard URL.

This has been fixed.

The task scheduler remains in the RUNNING state if max tries is set to -1. This problem has been fixed.

The Recordings Cleaner daemon on Blazar was not functioning correctly, causing recordings to remain undeleted. This issue has now been resolved.

The telephone number was not displayed correctly as "anonymous" for anonymous callers on the agent side of calls. This problem has been fixed.

The IVR - Script Object had no Name parameter. This has been added.

In the reports ACD-A-003 - ACD Agent Performance Report and ACD-A-004 - ACD Agent Performance Report Current Agent, the Ready Time calculation was producing incorrect values for the following use case:

• The agent logs in and receives a call.
• During the post-call interval, the agent switches to manual post-call status.
• The agent then logs out from the ACD Group while remaining in manual post-call status.

This issue occurred because the manual post-call time was not included in the Ready Time calculation, resulting in an over-subtraction of Ready Time.

The calculation has been updated to include ManualPostCallTime, ensuring that Ready Time is accurately computed.

This problem has been fixed.