🎙️ Voice AI Frontend SDK

Transform your website into an interactive, voice-controlled experience.

The Voice AI Frontend SDK is a production-grade library that enables users to browse, search, and interact with your web application using only their voice. Designed with a focus on Inclusive Design, it serves as a powerful bridge for users with visual, motor, or cognitive impairments.

🚀 Key Features

🧠 Hybrid Intelligent Engine

Switch seamlessly between Offline Mode (fast, zero-latency commands) and Online AI Mode (Complex natural language processing via LLMs).

🏷️ Stable ID Fingerprinting

Unlike standard voice scripts, our SDK uses a custom Fingerprinting Algorythm. This ensures that even when your React or Next.js app re-renders, the AI keeps a stable "hook" on UI elements, preventing broken click actions.

⌨️ Human-Like Input Simulation

Built to bypass the "React State Trap." When the AI fills a form, it simulates a full keyboard lifecycle:

• keydown → keypress → input → change → keyup This ensures your useState and validation logic capture every character perfectly.

🇮🇳 Multi-Language & Accents

Automatic detection and selection of Indian English (en-IN) and Hindi (hi-IN) voices. Supporting "Hinglish" commands right out of the box.

🛠️ Installation

npm install @voice-ai/frontend-sdk

📖 Basic Usage

1. Simple Setup (React/Next.js)

import { useEffect } from 'react';
import Talkify from '@voice-ai/frontend-sdk';

export default function App() {
  useEffect(() => {
    // 1. Initialize
    Talkify.init({
      backendUrl: "https://your-api.com/voice",
      mode: "auto", // Automatically falls back to offline if server is down
      wakeWord: "hey website" // Optional
    });

    // 2. Start the Mic
    Talkify.start();

    return () => Talkify.stop();
  }, []);

  return <div>Voice Enabled Website</div>;
}

♿ Why Inclusive Design? (The Benefits)

This SDK was built to make the web unconditionally accessible:

• 🧑‍🦯 Visually Impaired: Users can ask "What is on this page?" or "Tell me about [Item]" and receive a high-fidelity vocal description.
• ♿ Motor Impaired: Navigate complex drop-downs, sliders, and buttons without ever touching a mouse or keyboard.
• 🕒 Hands-Free Convenience: Perfect for cooking sites, repair manuals, or "lean back" browsing.
• 🤝 Global Inclusion: Support for localized accents (Indian English, Hindi) ensures technology speaks the user's language.

🔒 Reliability & Stability

• Deadlock Prevention: Specific fixes for Windows/Chrome TTS audio driver bugs.
• Mic Persistent: Integrated auto-restart logic that handles browser-enforced microphone timeouts gracefully.
• Memory Efficient: Utility memory management to prevent memory leaks during long-running voice sessions.

🔗 Architecture

This package works best when paired with the @voice-ai/backend-sdk which acts as the "Brain" for complex natural language reasoning.

Built with ❤️ for a more accessible web.