Changelog
AppService status
Start typing to search…
Back to All

v257 Enhanced Analytics & Voice Features 📊

about 2 hours ago by Michał Brzezicki

'"More human than human" is our motto.' - Tyrell, Blade Runner 🤖
SentiOne Automate version 257 was released on 15.01.2026. This release brings significant improvements to analytics export capabilities, voice channel enhancements, and various security and performance updates.

Export Analytics Data to Excel and CSV

Users can now export any table in the Analytics tab directly to XLSX and CSV formats. This highly requested feature eliminates the need for manual data copying and streamlines reporting workflows. Simply click the new Export button available on analytics tables to download your data in your preferred format.

Enhanced Voice Channel Support

New TTS Providers:

  • Deepgram TTS - Added comprehensive form support for Deepgram's text-to-speech with all Aura-2 English and Spanish voices
  • OpenAI TTS - Form support for OpenAI text-to-speech with voices including alloy, ash, coral, echo, fable, onyx, nova, sage, and shimmer

Voice Performance Improvements:

  • Custom timeout configuration for NLU calls during voice conversations (default 10s)
  • Support for external recordings in the new Conversations module
  • Fixed Speech key field validation in ASR Provider Settings

API Enhancements

New admin panel APIs for project and knowledge base management:

  • Export entire projects to ZIP files
  • Import/upsert projects from ZIP files
  • Export classic Knowledge Base to CSV
  • Import Knowledge Base from CSV with overwrite support

NLU & Language Detection

New NLU Capabilities:

  • AEL function to perform custom queries in NLU beyond standard user input
  • Language identifier integrated into NLU pipeline with support for 176+ languages, with detected language exposed as new AEL value
  • Cross-validation support for LLM intentizer

LLM Integration Improvements:

  • Temperature parameter support in LLM Say block for creativity control
  • Azure OpenAI credentials now support endpoints with cognitiveservices subdomain
  • Google ASR model selection cleaned up - removed problematic "short" and "telephony_short" options

Integration Status Code Handling

Integration blocks now expose HTTP status codes as part of their output, providing better control over error handling. When an integration request times out, the block returns status code 408 along with success=false, instead of throwing a bot error. This allows bot designers to handle timeouts gracefully through conditional logic rather than error contexts. The status code is available alongside existing output values (response, success) and is also displayed in the integration tester for easier debugging.

💊 IMPROVEMENTS

  • [Analytics] Export functionality for any table to XLSX and CSV formats
  • [Voice] Deepgram TTS integration with full voice model support
  • [Voice] OpenAI TTS integration with 9 voice options
  • [NLU] Configurable timeout for NLU calls with reduced default (10s for voice) to improve error handling
  • [Voice] External recording support in new Conversations module
  • [API] Project export to ZIP via admin API
  • [API] Project import/upsert from ZIP via admin API
  • [API] Knowledge Base export and import endpoints
  • [NLU] Custom query function in AEL for flexible NLU testing
  • [NLU] Language detection integrated into pipeline with AEL exposure (supports 176+ languages)
  • [LLM] Temperature parameter in LLM Say block
  • [LLM] Azure OpenAI support for cognitiveservices subdomain endpoints
  • [Integration] Integration block now returns HTTP status codes (including 408 for timeout) in response, allowing graceful error handling without throwing bot errors
  • [Security] SSO-enabled label added to user list in Organization settings
  • [Security] 2FA enforcement automatically disabled for SSO-forced users
  • [Security] Fixed permissions bypass vulnerability in Documents tab
  • [Conversations] Commands now visible in Conversations view
  • [WebChat] Confirm popup before closing webchat

🐛 FIXES

  • [UI] Multi-select form styling corrections (AI-generated phrases display)

  • [UI] Markdown formatting now works correctly in Knowledge Base search

  • [NLU] Blocked unsupported AI credentials (Azure SPN, Gemini, OpenAI compatible) from LLM intentizer

  • [NLU] Cross-validation now works properly for LLM intentizer

  • [NLU] Multi-intent phrase form can now create intents correctly

  • [NLU] Training seed automatically set to "Random" type by default in settings

  • [Flow] Fixed starting flow changes affecting active sessions

  • [Flow] Extended processing time message errors now properly handled

  • [Voice] API key field now properly marked as required in ASR Provider Settings

Follow the changelog for more exciting new features! 👋