AI Voice Agent
Build intelligent, production-ready voice agents that handle phone calls with human-like conversations.
What You Can Build
- Customer Support Bots - 24/7 automated support with human handoff
- Order Status IVR - Self-service order tracking
- Appointment Booking - Schedule and confirm appointments
- Lead Qualification - Qualify and route inbound leads
- Outbound Sales - Automated sales campaigns
Key Features
| Feature | Description |
|---|---|
| Sub-500ms Latency | Real-time conversations with minimal delay |
| 50+ Languages | Hindi, Tamil, Spanish, and more |
| Multiple Providers | Twilio, Exotel, Plivo, Telnyx, WhatsApp |
| Function Calling | Integrate with your APIs and databases |
| Native Audio | Gemini Live for audio-to-audio processing |
Quick Start
{
"agent": {
"name": "My Voice Agent",
"language": "en-US",
"llmProvider": "gemini-2.5",
"sttProvider": "deepgram",
"ttsProvider": "cartesia",
"greetingMessage": "Hello! How can I help you today?",
"prompt": "You are a helpful customer support agent..."
}
}
Architecture Overview
Incoming Call → Telephony Provider → WebSocket
↓
┌───────────────────┐
│ Voice Pipeline │
│ │
│ Audio → STT │
│ ↓ │
│ Text → LLM │
│ ↓ │
│ Response → TTS │
│ ↓ │
│ Audio → Caller │
└───────────────────┘
Provider Options
Speech-to-Text (STT)
- Deepgram - Fastest, best for English
- Google Chirp - Best for Indian languages
- Azure Speech - Enterprise reliability
Text-to-Speech (TTS)
- Cartesia - Ultra-low latency
- ElevenLabs - Premium voice quality
- Azure Neural - Best Indian language voices
LLM
- Gemini 2.5 Flash - Fastest, lowest cost
- Gemini Live - Native audio processing
- OpenAI GPT-4o - Best reasoning
Telephony
Documentation Sections
Getting Started
- Introduction - Overview and concepts
- Quick Start - Build your first agent
- Architecture - System design
- Configuration - Setup options
Core Features
- Function Calling - Tool integration
- Call Transfer - Human handoff
- DTMF Handling - Keypad input
- Webhooks - Event notifications
Pipeline
- VAD - Voice activity detection
- Interruptions - Barge-in handling
- Audio Processing - Signal processing
Languages
- Multi-Language - 50+ language support
- Hindi - Hindi configuration
- Tamil - Tamil configuration
- Indian Languages - All Indian languages
Examples
| Example | Description |
|---|---|
| Customer Support | FAQ handling, tickets, human handoff |
| Order Status | Simple order tracking IVR |
| Appointment Booking | Schedule and confirm appointments |
| Lead Qualification | BANT-based lead scoring |
| Outbound Sales | Sales campaign automation |
Next Steps
→ Quick Start Guide - Build your first voice agent in 5 minutes