VoiceAgent Builder — No-Code AI Phone Agent Platform

Give any business a 24/7 AI phone agent in 10 minutes. No code required.

VoiceAgent Builder lets dentists, restaurants, law firms, and any small business configure an AI phone agent that answers calls, books appointments, handles FAQs, and escalates to humans — using their own cloned voice.

Features

• Zero-Shot Voice Cloning — Magpie TTS clones owner's voice from 10-second sample
• Real-Time Transcription — Parakeet-CTC processes caller speech in real time
• Smart Topic Control — NemoGuard keeps agent on-topic for that business
• Multilingual Support — RIVA handles callers in any language automatically
• Knowledge Base RAG — Agent retrieves answers from business FAQ/documents
• Jailbreak Protection — Prevents callers from manipulating the agent
• Human Escalation — Detects when to hand off and does so gracefully
• Call Analytics — Dashboard with call logs, resolution rates, common queries

Tech Stack

• Frontend: React 18 + TypeScript + Tailwind CSS + Vite
• Backend: Node.js + Express + TypeScript
• Telephony: Twilio (BYOK) / WebRTC for browser calls
• Database: SQLite for agent config + call logs
• AI: 8 NVIDIA NIM endpoints

Quick Start

git clone https://github.com/HayreBuilds/voiceagent.git
cd voiceagent
cp .env.example .env
# Add NVIDIA_API_KEY (+ TWILIO_* keys for phone calls)

cd backend && npm install
cd ../frontend && npm install

cd backend && npm run dev
cd ../frontend && npm run dev

Environment Variables

NVIDIA_API_KEY=your_key_here
TWILIO_ACCOUNT_SID=optional_for_real_calls
TWILIO_AUTH_TOKEN=optional_for_real_calls
TWILIO_PHONE_NUMBER=optional
PORT=3001

API Endpoints

Method	Route	Description
POST	/api/agents	Create a new voice agent
PUT	/api/agents/:id	Update agent configuration
GET	/api/agents	List all agents
POST	/api/agents/:id/clone-voice	Upload voice sample for cloning
POST	/api/agents/:id/knowledge	Add FAQ/document to knowledge base
POST	/api/calls/inbound	Handle inbound call webhook
GET	/api/calls	Get call history
GET	/api/analytics	Get agent performance analytics

NVIDIA Endpoints Used

• parakeet-ctc-1.1b-asr — Real-time caller transcription
• magpie-tts-zeroshot — Owner voice cloning
• Background Noise Removal — Call audio preprocessing
• nemotron-3-nano-30b-a3b — Conversation understanding
• llama-3.1-nemoguard-8b-topic-control — Business topic enforcement
• nemojail-jailbreak-detect — Caller manipulation prevention
• riva-translate-4b-instruct — Automatic language handling
• nv-embedqa-e5-v5 — FAQ knowledge retrieval

License

MIT