Replace default passwords and secrets.
Confirm the softphone source network.
Create the recording directory:
- /var/spool/asterisk/monitor/mvpcc
Point ASTERISK_QUEUE_MAP_JSON in the platform to the matching queue code:
- voice_lab
- voice_lab_ai for the dedicated AI-first staging entrypoint
- voice_lab_ivr for the dedicated IVR staging entrypoint
Configure operator extension mapping in bridge env for Track 11 call control:
- ASTERISK_OPERATOR_EXTENSION_MAP_JSON={"operator_a":"2001","operator_b":"2002"}
Optional transfer map for queue-code based blind transfer:
- ASTERISK_TRANSFER_TARGET_MAP_JSON={"voice_lab":"2001","voice_lab_ai":"2001","voice_lab_ivr":"7200"}
For IVR-over-Asterisk staging:
- enable ASTERISK_IVR_FASTAGI_ENABLED=1
- set ASTERISK_IVR_FASTAGI_HOST=<bridge-host>
- set ASTERISK_IVR_FASTAGI_PORT=4573
- set IVR_RUNTIME_TRUSTED_SERVICE_SUBJECTS=svc:asterisk-bridge
- provision the referenced prompt_audio_key files under the Asterisk sounds directory
- if you want a higher-quality prompt from OpenAI TTS, generate it once and keep it on the server:
  - python scripts/materialize_tts_prompt.py --text "Здравствуйте! Добро пожаловать в контакт-центр. Для русского языка нажмите 1. Қазақ тілі үшін 2 басыңыз." --output /usr/share/asterisk/sounds/en/mvpcc_ivr_language.wav --language ru
  - python scripts/materialize_tts_prompt.py --text "Для отдела продаж нажмите 1. Для службы поддержки нажмите 2." --output /usr/share/asterisk/sounds/en/mvpcc_ivr_menu_ru.wav --language ru
  - python scripts/materialize_tts_prompt.py --text "Сату бөлімі үшін 1 басыңыз. Қолдау қызметі үшін 2 басыңыз." --output /usr/share/asterisk/sounds/en/mvpcc_ivr_menu_kz.wav --language kk
  - repeated runs reuse the server cache from AI_VOICE_TTS_CACHE_DIR instead of calling OpenAI again for the same text
For Track 14 browser softphone QA:
- enable ASTERISK_WEBRTC_ENABLED=1
- set ASTERISK_WEBRTC_WS_URL=wss://<asterisk-host>:8089/ws
- configure ASTERISK_BROWSER_SIP_MAP_JSON
- replace TLS placeholders in http.conf with a trusted cert/key pair
For Voice AI V1 lab wiring:
- enable AI_VOICE_ENABLED=1
- set AI_VOICE_QUEUE_CONFIG_JSON={"voice_lab_ai":{"mode":"ai_first","agent_profile":"voice_support","handoff_queue_code":"voice_lab","language":"ru"}}
- set AI_VOICE_RUNTIME_SERVICE_URL=http://ai-voice-runtime-service:8000
- keep AI_VOICE_RUNTIME_TRUSTED_SERVICE_SUBJECTS=svc:ai-voice-runtime
- load app_audiosocket.so on the Asterisk VM before enabling the AI media path
- set MVPCC_AI_AUDIOSOCKET_SERVICE=<media-host>:<port> in extensions.conf only after the AudioSocket listener is reachable

Entrypoints in extensions.conf:

7000: human-only baseline. This path preserves the original Track 9/11 behavior.
7100: AI-first staging path. It emits the normal MVPCCCallStarted event with QueueCode=voice_lab_ai, then:
- falls back to human dialing immediately if MVPCC_AI_AUDIOSOCKET_SERVICE is blank
- otherwise starts AudioSocket(<uuid>,<host>:<port>) for the media bridge
7200: IVR staging path. It emits the normal MVPCCCallStarted event with QueueCode=voice_lab_ivr, then:
- hands control to FastAGI at MVPCC_IVR_FASTAGI_HOSTPORT
- routes completed IVR sessions through mvpcc-transfer
- falls back to the human baseline if FastAGI does not return a transfer target

Important Voice AI note:

AI_VOICE_RUNTIME_SERVICE_URL is the HTTP control-plane endpoint for platform services.
MVPCC_AI_AUDIOSOCKET_SERVICE is a separate raw AudioSocket address for Asterisk media.
Do not point MVPCC_AI_AUDIOSOCKET_SERVICE at the runtime HTTP port.

Default lab SIP accounts from pjsip.conf:

1001 (caller)
2001 (operator A)
2002 (operator B)

Track 14 note:

2001 and 2002 are allowed to keep multiple contacts so browser registration can coexist with MicroSIP/Zoiper in QA.