Deepgram Speech Recognition (SR) Plugin 1.3.0 to the UniMRCP Server (UMS) has been released.
The plugin is based on the following components:
- Deepgram WebSocket API
- Libevent 2.1.9
- Rapidjson 1.1.0
The plugin is available for the following UMS versions:
The binaries are currently available for the following Linux distributions:
- Red Hat 7
- unimrcp-deepgram-sr-1.7.6-1.el7.x86_64.rpm
- unimrcp-deepgram-sr-1.8.4-1.el7.x86_64.rpm
- Red Hat 8
- unimrcp-deepgram-sr-1.7.6-1.el8.x86_64.rpm
- unimrcp-deepgram-sr-1.8.4-1.el8.x86_64.rpm
- Red Hat 9
- unimrcp-deepgram-sr-1.7.6-1.el9.x86_64.rpm
- unimrcp-deepgram-sr-1.8.4-1.el9.x86_64.rpm
- Ubuntu 20.04 LTS
- unimrcp-deepgram-sr_1.7.6-focal_amd64.deb
- unimrcp-deepgram-sr_1.8.4-focal_amd64.deb
- Ubuntu 22.04 LTS
- unimrcp-deepgram-sr_1.7.6-jammy_amd64.deb
- unimrcp-deepgram-sr_1.8.4-jammy_amd64.deb
This release allows to specify multiple 'deepgram.tag' parameters. The release also provides a few fixes and improvements.
The detailed list of changes introduced in this release follows.
New Features- Allow to specify multiple 'deepgram.tag' parameters.
- When 'vad_events' is set to 'true', derive START-OF-INPUT from 'SpeechStarted' event.
- Use 'speech_final' flag as an indication of input completion when 'single-utterance' is set to 'false'.
Fixed Problems- Skip processing of new messages received from Deepgram when connection is closing.
- Make
sure the keep-alive timer is properly stopped when start-of-input is set to
internal. This issue could possibly cause a segfault.
Configuration ParametersMiscellaneous Visit the Deepgram plugin page for more information.
Thank you for using UniMRCP.
--