July, 2026

Introduction

This month's updates introduce new features that make it easier for Sanas app users to test and troubleshoot their audio experience, dynamically switch inferencing models for the Sanas Chrome Extension, while delivering significant enhancements to Speech Intelligence with more customizable call insights, improved call quality detection, faster dashboard performance, more resilient post-call processing, and a more intuitive live tracking experience, to name a few!

What’s New

Hear Your Voice is a new in-app feature that lets users record and instantly preview how they'll sound through the active Sanas pipeline, with access controlled by a toggle in the Sanas Portal.
Synthesis Indicators add clear visual status indicators in the Sanas app, helping users quickly identify if audio synthesis isn't running and troubleshoot common issues without checking logs or contacting support.
Remote inferencing models for the Sanas Chrome Extension can now be switched dynamically through the Sanas Portal.
Speech Intelligence Enhancements
- Live Tracking roster tiles now display up to five sentiment indicators, showing how in-progress conversations have evolved over time.
- Call Detail now displays clear post-call analysis states, showing processing or failed status while ensuring analytics are only presented after the call has ended.
- Live Tracking now delivers more accurate active call tracking and a cleaner roster experience through automatic orphaned call recovery, improved post-call processing, uninterrupted call timers, and a refined Alerts indicator layout.
- Speech Intelligence platform reliability has been improved through enhanced call lifecycle and transcript processing, with expanded post-call audit logging for easier summary investigation.
- Speech Intelligence now supports Portal-managed Call Summary formatting, replacing per-POC prompt configuration with centralized output settings.
- The Call History is providing more accurate duration-based sorting within the selected date and time range and supports CSV export for easier reporting.
- Live Tracking agent tiles have been refined to keep the View action and call timer properly displayed, improving readability and preventing layout issues.
- Strengthened security and transcript lifecycle management with improved ownership validation, retention handling, and security updates.
- Transcript management provides flexible storage, retention, and transfer options, including secure and reliable delivery to customer-owned storage destinations.
- Added configurable Industry, Call Type, and Live Tracking conversation duration targets to support more tailored processing and monitoring.
- Live Tracking has been enhanced with recent speech activity indicators, conversation duration targets, improved call handling, and paginated call history.
- Call Details and transcripts have been improved with a cleaner transcript view, faster navigation to alerted utterances, and clearer processing status indicators.
- The download experience for Dashboard and Compliance tables has been improved with client-side CSV exports, enhanced security, and better browser compatibility.
- Live Tracking and Call History now use improved status filtering to more accurately identify active and completed calls.
- Improved the reliability of the post-call fraud pipeline by ensuring exhausted chunk retries correctly advance the checked counter and unblock post-call drain.
- Customizable Call Details now support group-specific prompt configurations for Action Items and Discussion Points, enabling customizable account-level formats in the Sanas Portal without requiring code changes.
- For Call Quality, we introduce five new inbound call-quality alert rules and improved backend block patterns to reduce false positives, delivering more accurate Speech Intelligence alerts throughout.
- Dashboard and Call History have been improved with faster dashboard and call history loading, plus optimized caching for Call Details to deliver quicker access to summaries, action items, and discussion points.
- Improved Post-Call LLM Reliability adds OpenAI as a fallback provider when the primary Anthropic service is unavailable, ensuring more resilient and reliable post-call processing.
- Interactive Live Tracking now offers improved usability through refined alert layouts, smoother call drill-downs, longer agent visibility after calls, and color-coded call duration timers for quicker monitoring.

Hear Your Voice In-App Feature

Sanas Desktop App version: 5.7.4 (Self-Serve) and 5.7.5 (Direct-CX), and later

A new in-app utility that lets app users record a short audio clip and play it back through the active Sanas pipeline. This gives users an immediate preview of how they will sound to others during meetings, without needing to schedule a test call.
How it works?
- Can be enabled by account admins through a toggle in the Sanas Portal.
- Sanas Portal → Settings → Configuration → Desktop Application → Application Settings.
  Note: This release goes together with new Sanas Portal version 1.11.0, and above.

Synthesis Indicators in the Sanas App

Sanas Desktop App version: 5.7.4 (Self-Serve) and 5.7.5 (Direct-CX), and later

The Sanas app now displays clear visual indicators of the audio synthesis state, available to all app users by default. When synthesis is not running, app users can immediately see why, making issues easier to diagnose without checking logs or contacting support.
Currently handled states include:
- Audio device failures
- Model load errors
- Synthesis is being explicitly turned off by the user
- The Sanas device is not configured as the default device

Note: Some of the Sanas models are dynamic i.e they have the capability to figure out speakers gender and also match the output voice to the original speaker voice. When these models are used, Sanas application hides the voice personalization and Gender selection UI as models handles it internally.

Dynamic Remote Inferencing Model Switchability for Sanas Chrome Extension

Sanas Chrome Extension version: 1.7.0 and above

With the latest version, remote inferencing models for the Sanas Chrome Extension can now be dynamically configured in the Sanas Portal instead of being hardcoded, allowing admins to switch models at any time.

How it works?

Refer to the in-app guidance for instructions on configuring Remote Inference in the Sanas Portal.

Call Sentiment Phasing for Speech Intelligence

Sanas Speech Intelligence model version: 1.4.0 and above

Live Tracking roster tiles now display a sentiment timeline of up to five emoji indicators for in-progress calls, enabling supervisors to quickly understand how the conversation has evolved without opening the Call Detail view.
Speech Intelligence analyzes both agent and customer speech in approximately 60-second intervals, generating an ordered sequence of positive, neutral, or negative sentiment. The most recent scored intervals are displayed in a first-in, first-out (FIFO) sequence, with a maximum of five sentiment indicators shown on each Live Tracking tile.

Note: This release goes together with new Sanas Portal version 1.17.1, and above.

Enhanced Analytics Readiness for Speech Intelligence

Sanas Speech Intelligence model version: 1.4.0 and above

Call Detail now provides clearer post-call analysis states instead of displaying incomplete analytics. While post-call analysis is in progress, the page displays a processing status. If analysis cannot be completed, a clear failed state is shown while still allowing transcript review when available.
Call Detail analytics are now generated only after the call has ended, ensuring supervisors are presented with a complete and reliable insights view rather than partial or incomplete analytics during an active call.

Note: This release goes together with new Sanas Portal version 1.17.1, and above.

Polished Live Tracking Roster and Reliability for Speech Intelligence

Sanas Speech Intelligence model version: 1.4.0 and above

Live Tracking now more accurately reflects active calls by automatically closing orphaned sessions caused by unexpected disconnects or application crashes. A one-time cleanup, combined with ongoing lifecycle reconciliation, prevents abandoned calls from remaining active and affecting Live Tracking metrics, including active call and talk-target counts.
Post-call processing has also been improved by automatically resolving analytics for previously stuck calls and recovering calls left indefinitely in processing or pending states, reducing delays in post-call insights generation.
Live Tracking interface has been refined by removing call duration limits that could make active call timers appear frozen. The Alerts indicator is now consistently positioned in the bottom-right corner of each roster tile for a cleaner and more predictable layout.

Note: This release goes together with new Sanas Portal version 1.17.1, and above.

Solidified Speech Intelligence Platform and Operations

Sanas Speech Intelligence model version: 1.4.0 and above

The Speech Intelligence platform and operations have been refined by separating the call lifecycle reconciliation process into its own scheduled job, independent of transcript retention and cleanup tasks. Transcript retention, purge, and assembly workflows have also been strengthened to improve overall reliability and processing consistency.
Post-call LLM audit logs now records the compliance alert names sent into the prompts, making it easier to investigate how detected alerts influence the summary behavior.

Note: This release goes together with new Sanas Portal version 1.17.1, and above.

Configurable Call Summary Output Format for Speech Intelligence

Sanas Speech Intelligence model version: 1.3.1 and above

Portal-Driven Configuration: Call Summary output formatting is now managed through the Speech Intelligence settings i the Sanas Portal instead of requiring per-POC prompt configuration.
Backend (Sanas Portal): Added a prompt examples catalog and setting references for summary, action items, and discussion points, along with configurable options for summary tone, point of view, and output volume. Existing production output is preserved through default settings.
Backend (Speech Intelligence): Updated the call-summary worker to consume Portal-managed format tokens and prompt examples from Speech Intelligence configuration, ensuring the summary directive and few-shot examples remain synchronized.

Note: This release goes together with new Sanas Portal version 1.14.1, and above.

Enhanced Call History Sorting & Export for Speech Intelligence

Sanas Speech Intelligence model version: 1.3.1 and above

Backend (Speech Intelligence): Updated Call History duration sorting to use the ClickHouse duration metric within the active date and time range, ensuring accurate ordering.
Frontend (Sanas Portal): Added CSV export support for the current Call History section.

Note: This release goes together with new Sanas Portal version 1.14.1, and above.

Improved Live Tracking View for Speech Intelligence

Sanas Speech Intelligence model version: 1.3.1 and above

Frontend (Sanas Portal): Updated Live Tracking agent tiles to keep the View action and call timer properly contained within each tile, preventing overflow and clipping issues.

Note: This release goes together with new Sanas Portal version 1.14.1, and above.

Hardened Security & Transcript Lifecycle Management for Speech Intelligence

Sanas Speech Intelligence model version: 1.3.1 and above

Backend (Speech Intelligence)
- Added call ownership validation to ingestion endpoints before accepting transcript and compliance alert writes (VUL-24), preventing cross-tenant call IDOR vulnerabilities.
- Improved transcript management metadata handling for retention and transfer workflows, and corrected retention-day override behavior so configured retention periods are no longer incorrectly limited by lower-environment caps.
- Updated dependencies to address high-severity CVEs in mako and json-repair.

Note: This release goes together with new Sanas Portal version 1.14.1, and above.

Refined Transcript Management for Speech Intelligence

Sanas Speech Intelligence model version: 1.3.0 and above

The new user interface of the Sanas Portal settings introduces the following transcript management configuration changes for Speech Intelligence:
- Transcript Storage: Configure the percentage of PII-redacted transcripts to retain, set the retention period using preset day options, and choose whether to retain alert utterance snippets after call transcripts are deleted.
- Transcript Transfer: Configure optional transcript transfers to customer-owned Amazon S3 destinations, including creating and managing destinations, defining allowlists, and optionally including call-level metadata with each transfer.
At the backend, the following configurations have been added:
- Speech Intelligence in the App: Added configurable transcript retention with scheduled cleanup, along with a transcript transfer service that reliably delivers retained transcripts to configured destinations using job tracking and automatic retries.
- Sanas Portal: Added new Transcript Management settings and APIs to support group-level storage and transfer configurations, including secure service-to-service access for Speech Intelligence.

Note: This release goes together with new Sanas Portal version 1.14.0, and above.

Expanded Speech Intelligence Settings: Industry, Call Type & Live Tracking Target

Sanas Speech Intelligence model version: 1.3.0 and above

Industry & Call Type: Assign Industry and Call Type values to groups using a predefined catalog, ensuring Speech Intelligence applies the appropriate defaults, processing, and filtering.
Conversation Duration Target: Configure a target Live Tracking supervisor timer (MM:SS); thresholds are calculated automatically at 80% (amber) and 100% (red) of the configured target.
Backend (Sanas Portal): Added Live Tracking / Target Conversation duration settings, support for Industry and Call Type catalogs, and Sanas-root entitlements for Speech Intelligence configuration.

Note: This release goes together with new Sanas Portal version 1.14.0, and above.

Upgraded Live Tracking for Speech Intelligence

Sanas Speech Intelligence model version: 1.3.0 and above

Backend (Speech Intelligence): Call History now exposes lastUtteranceAT, enabling Live Tracking to display how recent each call's speech activity is.
Frontend (Sanas Portal): Agent roster tiles now display last-utterance age with color-coded freshness indicators, honor the configured conversation duration target thresholds (green, amber, and red), improve grace-period handling for recently ended calls, and add pagination to Live Tracking call history.

Note: This release goes together with new Sanas Portal version 1.14.0, and above.

Elevated Call Detail & Transcript Experience for Speech Intelligence

Sanas Speech Intelligence model version: 1.3.0 and above

Compact Transcript View: Added chat-bubble transcript rendering across Call Detail and Live Tracking drill-down views for a cleaner, more readable experience.
Alert Navigation: Added transcript alert navigation to quickly jump between flagged alert utterances.
Clearer Processing States: Improved Call Detail analytics empty states to better indicate when summaries or action items are still processing or unavailable.

Note: This release goes together with new Sanas Portal version 1.14.0, and above.

Optimized Download Experience for Speech Intelligence

Sanas Speech Intelligence model version: 1.2.1 and above

Added client-side CSV export for the Speech Intelligence Dashboard and Compliance tables (Time View, User View, Compliance Alerts, and Compliance Statistics), allowing users to download the current table view without requiring an additional API call.
Enhanced CSV export security by protecting against spreadsheet formula injection and improved download reliability across supported browsers, including Safari and Firefox.

Note: This release goes together with new Sanas Portal version 1.13.1, and above.

Live Tracking and Call History Refined Status Filtering for Speech Intelligence

Sanas Speech Intelligence model version: 1.2.1 and above

Backend: Improved status filtering for In Progress and Completed calls by using call end times (call_end_utc) to determine call status. Active calls are now identified by a null end time, while recently completed calls remain visible during a short grace period. This eliminates reliance on legacy call status values, preventing stale or incorrectly classified calls from appearing in the Live Tracking and Call History views.

Note: This release goes together with new Sanas Portal version 1.13.1, and above.

Reliable Post-Call Fraud Pipeline for Speech Intelligence

Sanas Speech Intelligence model version: 1.2.1 and above

Backend: Fixed fraud worker result publishing so that exhausted chunk retries correctly advance the checked counter, unblocking post-call drain and preventing abandoned post-call runs where arrived ≫ checked.

Note: This release goes together with new Sanas Portal version 1.13.1, and above.