Q2 2026 (Apr–Jun 2026)
Enhancements
-
Enhanced Projects List with Ownership & Version Details
The Data Flow projects list now shows who created each project, who last modified it, when, and clear version information (default version and total version count) — replacing less useful columns and giving teams better visibility at a glance. -
Downloadable Call Recordings from Reports
A download button has been added to recordings on the Reports page, letting users save call audio directly without extra steps. -
Built-in Phone Number Grammar
A ready-to-use built-in grammar for recognizing phone numbers is now available, reducing setup effort for common recognition scenarios. -
Audit Logging for Core Backend
Core backend actions are now captured in audit logs, supporting compliance and troubleshooting needs. -
Listen to TTS Output During Project Testing
Users testing a project can now listen to TTS audio output directly on the test screen, including streamed audio, without needing to test elsewhere. -
Manual TTS Voice Entry
Users can now manually type a voice name directly into the TTS voice field, rather than being limited to selecting from the predefined dropdown list.
Improvements
-
Smarter Analytics Exclusion Logic
Calls are now automatically excluded from or included in analytics processing based on whether Analytics is actually enabled for that conversation, avoiding unnecessary processing of calls that were never meant to be analyzed. -
Clearer Version History Navigation
The project version button has been renamed from "Current Version" to "Version History" and restyled to match other primary actions, making it more intuitive to find past versions. -
Improved Scrolling Experience
Scrolling behavior has been refined across SR Recognitions, Models, Grammars, Pronunciations, and Data Flow Projects and Data Sources pages for a smoother browsing experience. -
Better Barge-In Handling with AI Agents
When a caller interrupts (barges in), Data Flow now tells the AI Agent exactly which message was interrupted, so the agent can correctly trim its conversation history and avoid referencing things the caller never actually heard. -
Refined Voice Activity Detection Sensitivity
The default noise gate threshold in VAD Elastic has been tuned based on field feedback, improving out-of-the-box speech detection accuracy. -
Faster Database Response Times
Connection pooling has been enabled at the ODBC level across products, cutting database-dependent request times dramatically.
Bug Fixes
-
Fixed an Incorrect Type Value for Duration-Type Notifications
Resolved an issue where duration-type notifications carried an incorrect type value. -
Fixed Unicode Conversion Errors in TTS
Resolved Unicode conversion errors affecting TTS output. -
Fixed Event Numbers Restarting on the Run Page
Resolved an issue where event numbers incorrectly restarted from the beginning for consecutive recordings on the Run page. -
Fixed a Missing Credential Field for Cloud Node Configuration
Resolved an issue where the credential field did not appear in node configuration when set to cloud. -
Fixed a Recognition Error with Uppercase Text
Resolved an error occurring in recognition requests containing uppercase text. -
Fixed an Error Removing a Language from SR Configuration
Resolved an error that occurred when removing a language from the selected language list in SR configuration. -
Fixed Issues with the Reports Page Download Button
Resolved issues affecting the download button on the Reports page. -
Fixed Duplicate-Looking Results for Different Operations in the Same Session
Resolved an issue where results from different operations within the same session were incorrectly displayed as identical on the Reports page. -
Fixed a Language Code Error for RTG Calls Sent to Conversational Analytics
Resolved a language code error in information sent to Conversational Analytics for RTG calls. -
Improved Noise Filter Detection Accuracy
Fixed incorrect audio classification indices affecting the Noise Filter node's detection accuracy. -
Fixed Incorrect Audio Header Formatting
Resolved an issue where audio files could receive an incorrect header format depending on the triggering event. -
Fixed Turkish Character Input in the Custom Node Editor
Resolved an issue where the custom node code editor did not accept Turkish characters. -
Improved Memory Stability
Fixed a memory allocation issue that could cause instability under certain conditions. -
Fixed a Null-Language Error in the Masking Process
Resolved a null-language error occurring during the masking/redaction process.