🎙️ VoiceIQ

AI-powered call center compliance platform for Indian call centers

---

🇮🇳 The Problem

Indian call centers handle thousands of calls daily. Supervisors manually listen to recordings to check compliance — slow, inconsistent, expensive, and completely incompatible with Hinglish and Tanglish speech.

VoiceIQ automates this entirely. Upload a call recording → get a full compliance report in under 30 seconds. No manual listening. No missed violations. No bias.

---

📋 Table of Contents

• The Problem
• Features
• Tech Stack
• How It Works
• Project Structure
• Database Schema
• Dashboard Pages
• Authentication
• Setup Instructions
• Backend Repository
• Known Limitations
• License

---

✨ Features

🗣️ Hinglish & Tanglish STT

Groq Whisper large-v3 transcribes mixed-language Indian call recordings with high accuracy - no manual effort.

✅ SOP Compliance Validation

Every call is scored against your custom SOP rules. Violations are flagged with exact rule references, timestamps, and severity levels.

😊 Sentiment Analysis

Detect caller mood and agent tone across every interaction. Spot negative patterns before they escalate.

💳 Payment Preference Detection

Automatically categorize whether customers prefer EMI, Full Payment, Partial Payment, or Down Payment.

🚩 Instant Call Flagging

Calls below 70% SOP compliance are automatically flagged for manager review — zero manual listening. Mark calls as reviewed to track progress.

📊 Live Analytics Dashboard

Real-time charts on compliance trends, sentiment distribution, language breakdown, and payment preferences. Includes best/worst performing day and most violated rule.

👤 Agent Performance Tracking

Leaderboard ranked by compliance %, with trend arrows (↑↓→), per-agent violations, sentiment trends, and full call history.

📋 Custom SOP Rules Manager

Define your own compliance rules per category — greeting, identity, payment, closing, and more. Up to 10 rules. Shows violation count per rule.

🎙️ Audio Player

Listen to the original call recording side by side with the AI transcript. Click any violation timestamp to jump directly to that moment in the audio.

💡 AI Coaching Suggestions

Every violation includes a specific, actionable coaching tip — with example scripts the agent can use next time.

⚠️ Violation Severity Levels

Each violation is classified as Critical, Major, or Minor — sorted by severity so the most important issues surface first.

---

🛠️ Tech Stack

Layer	Technology
Frontend	React 18 · Vite · TypeScript
Backend	Node.js · Express (ESM)
Database	Supabase (PostgreSQL · Auth · Storage)
AI / STT	Groq Whisper large-v3 · Llama 3.3 70B
Charts	Recharts
Deploy	Vercel (frontend) · Render (backend)

---

⚙️ How It Works

Upload audio file (MP3 / WAV / M4A)
        ↓
Saved to Supabase Storage (private bucket)
        ↓
Groq Whisper transcribes → auto-detects language + generates timestamped segments
        ↓
Custom SOP rules fetched from DB (or defaults used)
        ↓
Llama 3.3 70B analyzes timestamped transcript against SOP rules
        ↓
Returns: compliance %, violations (with timestamps + severity + coaching),
         sentiment, payment preference, passed checks, summary
        ↓
Results saved to DB → displayed on dashboard with audio player

---

📁 Project Structure

voiceiq-frontend/
├── src/
│   ├── components/
│   │   ├── ui/              # Button, Card, Badge, Loader, EmptyState
│   │   ├── layout/          # Navbar, Sidebar, MobileSidebar, DashboardLayout
│   │   └── charts/          # ComplianceTrend, Sentiment, Language, Payment
│   ├── pages/
│   │   ├── LandingPage.tsx
│   │   ├── LoginPage.tsx
│   │   ├── SignupPage.tsx
│   │   └── dashboard/
│   │       ├── DashboardHome.tsx
│   │       ├── CallsPage.tsx
│   │       ├── CallDetailPage.tsx
│   │       ├── FlaggedCallsPage.tsx
│   │       ├── AgentsPage.tsx
│   │       ├── AgentCallsPage.tsx
│   │       └── SOPRulesPage.tsx
│   ├── context/             # AuthContext
│   ├── hooks/               # useAuth, useCalls
│   ├── services/            # api.ts, supabase.ts
│   └── utils/               # helpers.ts
├── vercel.json
└── package.json

---

🗃️ Database Schema

calls       — recordings, transcripts, compliance results, agent assignment, reviewed status
agents      — agent profiles per user account
sop_rules   — custom SOP rules per user (max 10)

All tables have Row Level Security (RLS) enabled via Supabase.

---

📊 Dashboard Pages

Dashboard Home — Stats overview, compliance trend chart, best/worst day, most violated rule

All Calls — Upload calls, assign agents, filter by date, view all recordings

Call Detail — Audio player with clickable violation timestamps, severity badges, coaching tips, transcript

Flagged Calls — Calls below 70% compliance, filter by reviewed/pending, mark as reviewed

Agents — Leaderboard with trend arrows, per-agent compliance, sentiment, top violation

SOP Rules — Create/edit/toggle rules, violation count per rule, default rules preview

---

🔐 Authentication

• Email and password with confirmation
• Google OAuth one-click sign in
• All routes protected via Supabase Auth
• JWT passed as Bearer token to backend on every request

---

⚙️ Setup Instructions

1. Clone the repository

git clone https://github.com/JexanJoel/VoiceIQ.git
cd VoiceIQ

2. Install dependencies

npm install

3. Set environment variables

cp .env.example .env

Fill in your .env:

VITE_SUPABASE_URL=your_supabase_project_url
VITE_SUPABASE_ANON_KEY=your_supabase_anon_key
VITE_API_URL=https://your-render-backend-url.onrender.com

4. Run locally

npm run dev

---

🔗 Backend Repository

The backend API (Node.js + Express + Groq) is in a separate repository:

👉 VoiceIQ Backend

The backend exposes POST /api/call-analytics — the hackathon evaluation endpoint that accepts Base64 MP3 audio and returns structured compliance JSON.

---

⚠️ Known Limitations

• Audio files above ~25MB may hit Whisper API limits
• Very noisy recordings reduce transcription accuracy
• Frontend requires an active backend on Render (free tier may cold-start on first request — allow ~30s)

---

📄 License

---

Built with ❤️ for HCL GUVI Intern Hiring Hackathon 2026 — Track 3: Call Center Compliance