Patch Release CX4.10.5

The platform now assigns unique credentials to each CX service and restricts access so every component (including MongoDB, Redis, ActiveMQ, and PostgreSQL clients) can only access its own database or data store. This enforces least‑privilege access, improves data store isolation, and strengthens overall platform security. For configuration details, see Configure Service-Based Accounts

CX-Core now uses HashiCorp Vault to manage credentials for MongoDB, Redis, ActiveMQ, and PostgreSQL. Credentials no longer rely on plaintext; instead, they retrieve them securely from Vault at runtime.
For deployment configuration steps, see the implementation guides.

Securely manages sensitive information within channel connector settings so that credentials are no longer exposed in API responses or application logs. Administrators can enable masking for specific fields to prevent unauthorized access to confidential data. For configuration steps, see CCM Masking Fields Configuration Guide.

The API for saving third-party activities now supports intent-based activity naming. When specific intents are included in the request body, the API derives and assigns the corresponding activity name automatically. As part of this enhancement, the QM_ACTIVITY intent and activity name have been added to the system.

Screen recording is now available in Quality Management (QM), enabling quality evaluators, supervisors, and compliance officers to review synchronized audio and screen recordings of agent–customer interactions. This provides a fuller context for quality assessments, coaching, and compliance verification—bridging the gap between voice‑only call monitoring and full interaction visibility.

Displays submitted form data as an activity in the QM Conversation View so evaluators can see all form-based inputs in context with the rest of the interaction.

Notifies agents when a conversation evaluation falls below a configurable score threshold, allowing them quickly identify and act on low-scoring interactions.

Conversation List filters now include teams, channels, and customer identifiers for more precise search and discovery.

Allows agents to initiate outbound email messages directly from the customer info panel on the Conversation View page, enabling them to either respond to an existing email thread or start a new outbound email to the same customer as needed.

When an agent changes their state to 'Not Ready – Callback', the system now reroutes all ongoing conversations back to the original queue, ensuring consistent handling and preventing conversations from being assigned to unavailable agents.

The following new enhancements are done to the Customer Widget:

• Configurable sound notifications: Customers can mute or unmute sounds during chat.

• Control widget dimensions: Set widget size manually when embedding in a website or mobile app.

• Screen responsiveness

• Bulgarian language support

• View images in an overlay

• Title and subtitle in the widget header

• Resizable font for button messages

• Timer-based transcript download screen with a dedicated button

• Network interruptions show a reconnecting loader with multiple retry attempts instead of an immediate error

• The redundant bottom close button hides when the widget opens.

Enable your self-service bots to exchange new message types such as Carousel, Clickable List, and Form messages, allowing them to capture customer input using the most suitable format.

Add and manage multiple dynamic custom variables for your web widget to control its behavior. Add as many variables as needed.

Added a search option to the dropdown fields in the Pre-Conversation Form to help users find and select options quickly.

Issue: Previously, the Customer Widget did not store conversation data via Web Channel Manager; it only worked if the value field was a string. When the value field was a number, boolean, or object, the Web Channel Manager threw an error, and the data was not submitted or stored, causing incomplete or missing conversation context for agents and reporting.

Resolution: This is now fixed, and Web Channel Manager correctly accepts and stores conversation data where the value field can be a string, number, boolean, or object. The error is no longer thrown for non‑string values, and the conversation data with these types is successfully submitted via the Customer Widget.

Issue: Previously, agent comments on Facebook posts were truncated if the agent used any special characters (such as #, $, %) in a reply. The full text was visible in the backend, but on the Facebook page, users only saw the content up to the first special character, leading to incomplete and confusing replies.

Resolution: The system now handles encoding correctly, and comments containing special characters are no longer truncated. The full comment text stored in the backend is displayed on the Facebook page for end users.

Issue: Previously, Conversation Manager and the customer widget displayed encoding issues for multilingual content, especially for Arabic, Bulgarian, and Urdu. This affected customer names during ringing, customer info, conversation data, form submissions, and customer, bot, and agent messages, including messages shown in the customer widget after a page refresh.

Resolution: These encoding issues are now resolved. Customer names, profile details, conversation data, form submissions, and messages across customer, bot, and agent views, as well as messages reloaded in the customer widget, now display correctly in Arabic, Bulgarian, Urdu, and other non‑Latin languages.

Issue: Previously, for some conversations and channels, the system recorded multiple question submissions against a single conversation, causing the reported number of submissions to be higher than the actual customer interactions. This led to inconsistent survey and reporting data and made submission metrics unreliable for analysis.

Resolution: The submission logic now enforces a single set of question submissions per conversation and channel. Duplicate submissions for the same conversation are no longer recorded, ensuring that submission counts in reports accurately reflect real customer interactions and remain consistent with the underlying database data.

Issue: Previously, when agents replied to emails via the MS Email Connector, Exchange IDs were sometimes inconsistent due to server load and slow synchronization on the email server. In addition, some incoming emails contained invalid or case‑mismatched Content-Id values, causing embedded images not to render correctly.

Resolution: The connector now generates its own InternetMessageId using the same pattern as MS Exchange, eliminating the need to fetch it after sending. Inline images are also handled correctly when incoming HTML contains invalid or case‑insensitive Content-Id values, ensuring that embedded images render reliably.

Issue: When an agent pasted an image immediately after a <br> HTML tag in the email editor, the image was inserted outside the editable content area. This is a known issue in Summernote, and Summernote is the editor used for email composition.

Resolution: This issue is fixed by wrapping the email content in a container element before initializing it in the editor. Pasted inline images now remain within the editable content area, and the email editor behaves as expected.

Issue: Previously, two-factor authentication (2FA) did not function correctly for the agent on the Agent Desk after it was enabled in the previous release. As a result, agents could not complete 2FA flows when signing in.

Resolution: 2FA has been restored and now works as expected on Agent Desk, ensuring consistent authentication security across both unified‑admin and unified‑agent applications.

Issue: The outbound SMS window remained open after sending a message, even when the “auto-close” option was enabled in settings. This increased the likelihood that agents would resend the same SMS, resulting in duplicate messages being sent to customers.

Resolution: The response from the Conversation Manager conversation-manager/activities API has been corrected. The outbound SMS window in Agent Desk now closes automatically after a successful send, reducing the risk of accidental resends.

Issue: Previously, after transferring an outbound call in CX, an agent’s MRD state remained Not_Ready. Agents had to manually change their MRD back to Ready to become available again, which made them appear unavailable longer than necessary and negatively impacted occupancy and routing.

Resolution: After an outbound call transfer completes, Agent1’s MRD state is now automatically set back to Ready. Agent1 becomes available for new interactions without any manual state change, and the transfer flow now aligns with the expected MRD state handling behavior.

Issue: Previously, when an agent clicked a link inside the active chat window, the link opened in the same browser tab. This navigated away from the Agent Desk and terminated the active chat session, disrupting ongoing conversations and forcing agents to reopen the desk and recover context.

Resolution: Agent Desk now opens all links clicked from within the agent chat window in a new browser tab. The main Agent Desk tab remains open with the active conversation intact, so agents can safely open external resources (web pages, knowledge base articles, etc.) without losing the current chat session or interrupting their workflow.

Issue: Previously, TYPING_INDICATOR events were treated as Message / bot_message. The frontend handled them like standard chat messages. Typing indicators could appear in or affect the message stream instead of being a transient UI hint.

Resolution: TYPING_INDICATOR events are now classified and processed as notifications rather than messages. The frontend can render them as temporary “typing…” indicators, separate from the main message history. Message lists remain clean and focused on actual conversation content, while still giving agents and users visibility that the bot is composing a reply.

Issue: In ongoing chats, when agents scrolled up to see past interactions, some bot messages showed as encoded/unintelligible text. Only certain bot integrations were affected, leading to inconsistent history views and confusion for agents.

Resolution: All bot messages in the chat history are now displayed in clear, readable text, regardless of the bot provider. The system consistently uses the isEncoded flag to encode and decode bot messages, ensuring a reliable and uniform transcript experience for agents.

Issue: Previously, clickableList messages containing emojis and non‑Latin scripts (e.g., Bulgarian) displayed correctly only in the live session. After a browser refresh, the same messages loaded from CCM appeared corrupted (broken emojis, garbled Bulgarian/special characters)

Resolution: clickableList messages with emojis and Bulgarian text now retain their formatting and characters even after page refresh. Messages reloaded from CCM now match the original live rendering, ensuring a consistent multilingual and emoji‑rich experience.