Release Date: May 18, 2022

• Add support for opentelemetry tracing and Jaeger.
• rtpengine performance improvements when running on EC2 by building/installing/running the rtpengine kernel module; this also enables additional recording features (e.g. recording all calls directly from rtpengine).
• Update to Microsoft Azure speech version 1.21, with support for new languages and other features.
• Add support for auto-detecting language when using Azure STT.
• Add support for model and singleUtterance properties when using Google STT.
• Dial: transcribe and listen should be based on the caller (A leg) endpoint.
• Add ability to edit a speech credential by changing the region (aws or azure).
• Jambonz portal: add ability to create service provider apikeys.
• Jambonz portal: add ability to enter phone numbers with wildcards so large blocks of numbers can be handled more easily.
• Kubernetes: health check improvements
• Send application defaults for speech in initial webhook
• CloudFormation: add optional parameter for the DNS name for the deployed jambonz-mini system and configure nginx accordingly.

• Implement confirmHook for dial
• Fix 481 Transaction Does Not Exist error on BYE after successful REFER.
• Gather: resolve with reason killed prevented task from ending.
• Gather: don't restart transcribing if task has been killed.
• Azure stt - if we get no speech detected, listen again.
• Send error notification over websocket if tts fails.
• Gather: catch errors when webhook fails.
• CreateCall needs to work with wss url.
• ws-requestor: queue outgoing messages if we are in the process of connecting to the remote wss server.
• Rest dial should exit upon call failure, not after call timeout is reached.
• Allow call status to be sent before killing rest dial on failure.
• Outbound rest dial over websocket api needs to send session:new.
• WsRequestor: reconnect if socket dropped from far end.
• Emit vad event on partial transcript.
• Background gather listen only once for vad and other interrupt events.
• Transcribe: Azure interim transcripts were missing.
• Transcribe of a dialed call can now occur on both legs.
• On gather timeout, if at least minDigits are collected, resolve with reason dtmf detected.
• Retrieve aws_region when getting an AWS speech credential.
• CreateCall accepts call_hook that is a ws(s) url.
• Sbc-inbound: memory leak fix.

• Available shortly on AWS Marketplace
• Deploy to Kubernetes using this Helm chart

Questions? Contact us at support@jambonz.org