This is a simple Python ESL script which initiates a call to a SIP extension and bridges that to a Vosk WebSocket / TCP endpoint using mod_audio_stream (it's my updated version of this module, here's the original module: https://github.com/amigniter/mod_audio_stream).
Default configuration values are ready to use, so you can just start everything using Docker Compose: docker compose up -d --build. Once it's started, you can authorize on the FreeSWITCH server with any SIP client. Login credentials would be:
- SIP server: localhost
- Username: anything in range from 1000 to 1019
- Password:
extensionpassword(could be changed indocker-compose.yml, variableEXTENSION_PASSWORD, servicefreeswitch, sectionenvironment)
When it's done, you need to dial 99999 (number could also be changed in docker-compose.yml, variable NUMBER_TO_DIAL in esl-app and in freeswitch services). The call will be answered and connected to the Websocket endpoint (by default it's Vosk recognition service endpoint, Vosk instance is deployed via Docker Compose along with other services. You can change a WS / TCP endpoint address in docker-compose.yml, variable ENDPOINT in esl-app service).
Once you're connected to Vosk WS endpoint, speech recognition would start working, and you'll see recognition results in esl-app logs (you can check logs this way: docker compose logs -f esl-app).
That's it — now you can run the script: python app.py.