Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.spn.wtf/llms.txt

Use this file to discover all available pages before exploring further.

​
Why LiveKit Is Used

LiveKit is the realtime audio layer. It provides:
  • Rooms for each call
  • Audio transport between the phone participant and the AI agent
  • Agent job dispatch through livekit-agents
  • SIP APIs for connecting phone calls
  • JWT-based access tokens for browser/local testing

​
Room Naming

Outbound lead calls use: 
ignitech-lead-{lead_id}
The agent reads the room name to understand which lead is being called.

​
SIP Participant Creation

The scheduler calls LiveKit’s SIP API: 
CreateSIPParticipantRequest(
    sip_trunk_id=LIVEKIT_SIP_TRUNK_ID,
    sip_call_to="+91XXXXXXXXXX",
    room_name="ignitech-lead-123",
    participant_identity="sip-lead-123",
    participant_name="Lead",
    krisp_enabled=True,
    play_ringtone=True,
)
If VOBIZ_PHONE_NUMBER is configured, it is also passed as sip_number.

​
Agent Runtime

Both agents are LiveKit agents:
  • agent.py: Sarvam STT + LLM + TTS pipeline
  • agent_realtime.py: OpenAI Realtime session
Each agent joins the LiveKit room, listens for the SIP participant, conducts the conversation, and saves the result to PostgreSQL.

​
Token Server

token_server.py exposes: 
GET /api/token?identity=<id>&room=<room>&lang=<lang>
It returns: 
{
  "token": "jwt",
  "url": "wss://your-project.livekit.cloud",
  "identity": "user",
  "room": "ignitech-lead-123"
}
This is mainly for local browser testing, not the production outbound calling path.

​
Required Environment

LIVEKIT_URL=wss://your-project.livekit.cloud
LIVEKIT_API_KEY=your_livekit_api_key
LIVEKIT_API_SECRET=your_livekit_api_secret
LIVEKIT_SIP_TRUNK_ID=ST_xxxxxxxxxxxx