Skip to content

IgniteHackNWin/SpotlightReady

Repository files navigation

SpotlightReady 🎤

AI-Powered Immersive Performance Simulation Engine

SpotlightReady is a real-time performance simulation + analytics platform that helps users rehearse high-stakes scenarios — job interviews, public speeches, pitches — using live AI analysis and structured post-session intelligence.

This is not a chatbot. This is not a grammar checker.
This is a performance rehearsal engine.


✅ Current Status (March 2026 — Hackathon Dev)

Fully working end-to-end:

  • 🏠 Landing page with mode selection (Interview / Speech / Pitch)
  • ⚙️ Session setup — role, difficulty, duration config
  • 🎤 Live session — AI-generated questions, real-time metrics display
  • 📊 AI-powered performance report with scores, feedback, improvement plan
  • 🗄️ MongoDB Atlas persistence (sessions + reports saved)
  • 🤖 Groq LLM integration (llama-3.1-8b-instant + llama-3.3-70b-versatile)

Report sections working:

  • Overall Score Card (0–100) + tier (Beginner / Developing / Proficient / Expert)
  • Speech Analytics (WPM, filler words, rhythm, stutters)
  • Visual Presence (eye contact, head stability, expressiveness)
  • Content Intelligence (relevance, structure, depth, missed points)
  • Grammar & Language (corrections, vocabulary upgrades)
  • Session Replay timeline
  • Improvement Plan (drills + retry recommendation)

Coming next:

  • Webcam + microphone real data capture (MediaPipe, Web Speech API)
  • Confidence & flow scoring from actual audio/video
  • PDF report export

🗂️ Monorepo Structure

SpotlightReady/
├── frontend/      → Next.js 14 App Router (Simulation UI + Report pages)
├── backend/       → Express.js API (Sessions, Reports, AI orchestration)
├── shared/        → Shared TypeScript types (SessionData, PerformanceReport, etc.)
└── docs/          → Architecture, ideation, and planning docs

🏗️ System Architecture

Layer Package Tech
UI frontend Next.js 14, Tailwind CSS, Zustand
API backend Express.js, MongoDB Atlas, tsx watch
AI backend/src/ai/ Groq API (llama models via openai SDK)
Types shared TypeScript interfaces shared across packages

AI Pipeline (2 LLM calls per session):

  1. transcriptEvaluatorllama-3.3-70b-versatile — scores content, grammar, missed points
  2. feedbackGeneratorllama-3.1-8b-instant — generates improvement drills + plan

🚀 Quick Start

Prerequisites

  • Node.js ≥ 18
  • pnpm ≥ 8 → npm install -g pnpm
  • MongoDB Atlas account (free M0 tier)
  • Groq API key (free at console.groq.com)

1. Install dependencies

pnpm install

2. Set up environment variables

Create backend/.env:

PORT=4000
NODE_ENV=development
FRONTEND_URL=http://localhost:3000
MONGODB_URI=your_mongodb_atlas_connection_string
MONGODB_DB_NAME=spotlightready
LLM_PROVIDER=groq
GROQ_API_KEY=your_groq_api_key

Create frontend/.env.local:

NEXT_PUBLIC_API_URL=http://localhost:4000/api

3. Run in development

# Terminal 1 — Backend
pnpm --filter backend dev

# Terminal 2 — Frontend  
pnpm --filter frontend dev

🎯 Feature Status

  • Landing page — mode selection (Interview / Speech / Pitch)
  • Session setup — role, topic, difficulty, duration
  • Live session — AI questions, timer, real-time metrics UI
  • Session submission + processing screen
  • AI report generation (Groq LLM)
  • Report page — score card, speech analytics, content intelligence, grammar, improvement plan
  • MongoDB persistence (sessions + reports)
  • Webcam capture — MediaPipe eye contact / head tracking
  • Microphone capture — Web Speech API transcript + WPM
  • Confidence scoring from real audio/video data
  • PDF report export

📊 Scoring Model

Category Weight Source
Speech Delivery 30/100 AI transcript analysis
Visual Presence 20/100 Webcam (coming soon)
Content Quality 30/100 AI content evaluation
Confidence & Flow 20/100 Combined metrics

🌿 Branching Strategy

Branch Purpose
main Production-ready code only
develop Integration branch — all PRs merge here
feature/* Individual features
fix/* Bug fixes
release/* Release prep

Never push directly to main.
All work → feature/* → PR to develop → PR to main for releases.


🧑‍💻 Team

Name Role
Yuvraj Lead Developer
Riya Co-Developer

📄 License

Private – Hackathon project. All rights reserved.

About

An AI-powered performance simulation platform that helps users practice interviews and speeches with real-time feedback and structured post-session analytics.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors