Voice

GNETiX supports outbound voice calls, allowing users to transition from a chat conversation to a phone call while preserving full conversation context. The voice system is powered by LiveKit for real-time audio, ElevenLabs for speech-to-text and text-to-speech, and SIP trunking via Twilio for PSTN connectivity.

How It Works

A user types "call me" (or similar) in any chat platform
The Director recognizes the intent and initiates a voice call
GNETiX creates a LiveKit room with an AI agent participant
The call is placed to the user's registered phone number via SIP/Twilio
The user picks up and speaks naturally with the AI agent
The full chat conversation context carries over into the voice session

Conversation context flows seamlessly from chat to voice. If a user was troubleshooting a network issue in Webex and says "call me," the voice agent already knows the full context of the conversation.

Architecture

User's Phone <--SIP/PSTN--> Twilio SIP Trunk <--SIP--> LiveKit Server
                                                            |
                                                     LiveKit Agent
                                                       |        |
                                                ElevenLabs    Director
                                                (STT + TTS)   (LLM)

Component	Role
LiveKit Cloud	Real-time audio infrastructure, room management, SIP bridge
ElevenLabs Scribe v2	Speech-to-text (STT) -- transcribes user speech
ElevenLabs TTS	Text-to-speech -- generates natural voice responses
Twilio SIP Trunk	PSTN connectivity -- routes calls to/from phone numbers
Director	LLM orchestration -- same Director that handles chat messages

Requirements

To enable voice calling, you need:

LiveKit Cloud account with a project and API key/secret
ElevenLabs API key with access to the Conversational AI or standard TTS/STT APIs
Twilio account with a SIP trunk and phone number configured
User phone numbers registered in GNETiX (configured per user in the portal)

Configuration

Voice settings are configured through environment variables on the backend:

# LiveKit
LIVEKIT_URL=wss://your-project.livekit.cloud
LIVEKIT_API_KEY=APIxxxxxxx
LIVEKIT_API_SECRET=xxxxxxxxxxxxxxxx

# ElevenLabs
ELEVENLABS_API_KEY=sk_xxxxxxxxxxxxxxxx

# Voice API authentication
VOICE_API_KEY=<random-secret-for-voice-endpoints>

SIP trunk configuration (Twilio dispatch rules, inbound/outbound trunk setup) is managed in the LiveKit dashboard.

User Phone Numbers

Each user who wants to receive voice calls must have a phone number registered in GNETiX. Admins can set this in the user profile, or users can configure it themselves (if permitted).

Phone numbers should be in E.164 format (e.g., +15551234567).

Limitations

Voice calls are outbound only -- the system calls the user, not the other way around
Requires PSTN connectivity through a SIP trunk provider
Voice quality depends on the user's phone connection and the SIP trunk provider
Tool calls during voice sessions use the same MCP infrastructure as chat