aillom-vox-client

v1.0.2

Published

23 days ago

Enterprise-Grade Voice AI SDK for Speech-to-Speech, Audio-to-Audio, and Realtime Multimodal applications.

🎙️ AillomVox Public Client

The Enterprise-Grade Voice AI SDK.

Build Speech-to-Speech, Audio-to-Audio, and Realtime Multimodal applications with a single, unified protocol. Connect effortlessly to OpenAI Realtime, Gemini Multimodal, AWS Nova, Qwen, Grok, UltraVox, and AillomVox native models.

📦 Installation

npm install aillom-vox-client

📚 Documentation

Quick Start (SDK) - The modern way
Quick Start (WebSocket) - The low-level way
Examples - Ready-to-use client implementations
N8N Integration - Official AillomVox node for n8n
Client Tools - Add custom UI controls to your AI
Voice Catalog - All voices across all providers
Asterisk Integration - Complete guide for Asterisk 23/SIP
Protocol Specification - WebSocket messages and binary formats
Supported Providers - AI models and configuration options
Troubleshooting - Common errors and solutions

🚀 Key Features

Unified API: Switch between OpenAI, Gemini, and others by changing one string
Real-Time Streaming: Full-duplex WebSocket for sub-500ms latency
65 Voices: Full Inworld TTS 1.5 catalog across 15 languages
Robust Audio: Native PCM 16-bit at 8kHz, 16kHz, or 24kHz
Client Tools: Add custom UI controls (hangup, alerts, navigation) to your AI
15 Languages: en, pt, es, fr, de, it, ja, zh, ko, hi, ar, ru, pl, nl, he
Event Driven: Simple event emitter (audio, transcript, interrupt)
Enterprise Security: Automatic key redaction and sanitized error messages

💰 Pricing & Performance

Choose the tier that fits your budget. AillomVox is optimized for telephony and high-volume use cases.

Why AillomVox?
Native optimized pipeline delivers sub-500ms latency and 8kHz support at less than half the cost. Choose from 65 voices with dynamic mid-conversation switching.

📱 Examples

This repository contains multiple examples ranging from a minimal connection script to full-featured dashboards and creative use cases.

⚡ Quick Start (SDK)

The easiest way to connect to AillomVox.

import { AillomVox } from 'aillom-vox-client';

const client = new AillomVox({
  apiKey: 'av_YOUR_KEY',
  voice: 'Edward',
  debug: true
});

client.on('transcript', (msg) => {
  console.log(`[${msg.role}] ${msg.text}`);
});

client.on('audio', (chunk) => {
  // Play chunk (ArrayBuffer)
});

await client.connect();

🔌 Quick Start (WebSocket)

If you prefer raw WebSockets (e.g. for Python, Go, or minimal JS):

const ws = new WebSocket("wss://vox.aillom.com/ws");

ws.onopen = () => {
  ws.send(JSON.stringify({
    type: "config",
    apikey: "YOUR_API_KEY",
    provider: "aillomvox",
    voice: "Edward"
  }));
};

ws.onmessage = (event) => {
  if (event.data instanceof ArrayBuffer) {
    playAudio(event.data);
  }
};

🎤 Voices

AillomVox uses Inworld TTS 1.5 with 65 voices across 15 languages. See the full Voice Catalog.

Top Picks

🛠️ Client Tools

Client Tools allow the AI to control your application's UI directly. When the AI decides to execute a tool, your app receives a callback and can respond.

Registering Tools

{
    "provider": "aillomvox",
    "voice": "Edward",
    "tools": [
        {
            "name": "hangup",
            "description": "End the call when user says goodbye.",
            "parameters": { "type": "object", "properties": {} }
        },
        {
            "name": "show_alert",
            "description": "Show alert to user",
            "parameters": {
                "type": "object",
                "properties": {
                    "message": { "type": "string", "description": "Alert message" }
                },
                "required": ["message"]
            }
        }
    ]
}

Handling Tool Calls (Client-Side)

socket.onmessage = (event) => {
    if (typeof event.data !== 'string') return;
    const msg = JSON.parse(event.data);
    
    if (msg.type === 'tool_call') {
        console.log(`Tool requested: ${msg.name}`, msg.args);
        
        let result = 'OK';
        if (msg.name === 'hangup') {
            disconnect();
            result = 'Call ended';
        } else if (msg.name === 'show_alert') {
            alert(msg.args.message);
            result = 'Alert displayed';
        }

        // Always respond — AI waits for this (15s timeout)
        socket.send(JSON.stringify({
            type: 'tool_result',
            call_id: msg.call_id,
            result: result
        }));
    }
};

🔧 Advanced Configuration

Audio Formats

{
    sample_rate: 16000,  // 8000 (telephony), 16000 (standard), 24000 (high-quality)
    // Audio is PCM 16-bit little-endian Mono
}

Session Limits

{
    max_duration: 300,  // 1-3600 seconds (default: 300 = 5 minutes)
}

At 15 seconds remaining, the AI will say the farewell_message. At 0 seconds, the connection closes.

Multi-Language Support

{
    language: 'pt-BR',
    voice: 'Heitor',  // or 'Maitê' for female
    system_prompt: 'Você é um assistente da Aillom. Seja conciso.',
    first_message: 'Olá! Como posso ajudar?',
    farewell_message: 'Obrigado por ligar. Até logo!'
}

Supported: en-US, pt-BR, es-ES, fr-FR, de-DE, it-IT, ja-JP, ko-KR, zh-CN, hi-IN, ar-SA, ru-RU, pl-PL, nl-NL, he-IL

Published

Vulnerabilities

Links

Maintainers

Keywords

Readme