This is a module to recognize speech using Vosk server. You can run the server in docker with simple:
docker run -d -p 2700:2700 alphacep/kaldi-en:latest
See for more details https://github.com/alphacep/vosk-server
To use this server with freeswitch:
- Make sure you have libks installed
- Configure and install freeswitch including
mod_vosk.so - Make sure mod_vosk.so is enabled in
modules.conf.xmlandconf/vosk.conf.xmlis placed in autoload_configs
Run the following sample dialplan:
<include>
<context name="default">
<extension name="asr_demo">
<condition field="destination_number" expression="^.*$">
<action application="answer"/>
<action application="play_and_detect_speech" data="ivr/ivr-welcome.wav detect:vosk default"/>
<action application="speak" data="tts_commandline|espeak|You said ${detect_speech_result}!"/>
</condition>
</extension>
</context>
</include>
!!!! ATTENTION, for reliable work this module requires several fixes in libks which are not yet merged, so rebuild libks from this:
git clone --branch vosk-fix --single-branch https://github.com/alphacep/libks
You can create more advanced dialplans with ESL and scripts in various languages. See examples in scripts folder.
!!! ATTENTION In order for ESL to recieve events, make sure that fire_asr_events variable is set to true (false by default). The dialplan can look like this:
<include>
<context name="default">
<extension name="asr_demo">
<condition field="destination_number" expression="^.*$">
<action application="answer"/>
<action application="set" data="fire_asr_events=true"/>
<action application="detect_speech" data="vosk default default"/>
<action application="sleep" data="10000000"/>
</condition>
</extension>
</context>
</include>
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent