Horizon

AI-powered mock interview practice platform. Practice interviews with an AI interviewer via real-time video/audio calls and receive detailed feedback to improve your skills.

Features

• AI Voice Interviewer - Conversational AI powered by Gemini Live
• Real-time Video/Audio - High-quality calls via Stream Video SDK
• Multiple Interview Types - Technical, Behavioral, System Design, Product, and General
• Multi-language Support - English, Spanish, and Bilingual modes
• Voice Selection - 6 distinct voices (Puck, Charon, Kore, Fenrir, Aoede, Leda)
• Smart Turn Detection - Reduced latency with 1s silence detection
• Session History - All sessions stored in PostgreSQL with transcripts
• AI Summaries - Automated feedback with strengths/weaknesses analysis
• User Authentication - Secure accounts via Better-auth
• Quiz System - Practice questions for interview preparation
• Resume Optimizer - AI-powered resume analysis and improvement

Architecture

Frontend (Next.js) --> API Routes --> Python Agent --> Gemini Live
                   |                        |
                   v                        v
              PostgreSQL            Stream Video Cloud

Tech Stack

Layer	Technologies
Frontend	Next.js 16, React 19, TypeScript 5, Tailwind CSS 4, Radix UI
Backend	Next.js API Routes, FastAPI (Python)
Database	PostgreSQL (Neon), Drizzle ORM
AI	Google Gemini Live, Vision Agents
Video	Stream Video SDK
Auth	Better-auth

Project Structure

/aimeeting
├── src/
│   ├── app/           # Next.js pages and API routes
│   ├── components/    # React components (ui/, meeting/, etc.)
│   ├── lib/           # Utilities (auth, stream, gemini)
│   └── db/            # Database schema
├── agent/
│   ├── server.py      # Python FastAPI agent
│   └── requirements.txt
└── package.json

Prerequisites

• Node.js 20.9.0+
• Python 3.11+
• Stream Video account
• Google AI (Gemini) API key
• PostgreSQL database (Neon recommended)

Environment Variables

Next.js (.env.local)

STREAM_API_KEY=your_stream_api_key
STREAM_API_SECRET=your_stream_api_secret
NEXT_PUBLIC_STREAM_API_KEY=your_stream_api_key
DATABASE_URL=your_neon_database_url
GOOGLE_API_KEY=your_gemini_api_key
BETTER_AUTH_SECRET=your_auth_secret
BETTER_AUTH_URL=http://localhost:3000

Python Agent (agent/.env)

STREAM_API_KEY=your_stream_api_key
STREAM_API_SECRET=your_stream_api_secret
GOOGLE_API_KEY=your_gemini_api_key
NEXTJS_API_URL=http://localhost:3000

Installation

Next.js Frontend

npm install
npm run dev

Python Agent

cd agent
python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate
pip install -r requirements.txt
python server.py

The app will be available at http://localhost:3000/dashboard

Configuration

Interview Types

Type	Description
technical	Algorithms, data structures, coding
behavioral	STAR method, past experiences
system_design	Architecture, scalability
product	Product thinking, estimation
general	Mix of technical and behavioral

Languages

• en - English
• es - Spanish
• bilingual - English/Spanish mix

Voices

Voice	Style
Puck	Upbeat, energetic
Charon	Calm, informative
Kore	Warm, friendly
Fenrir	Strong, confident
Aoede	Soft, bright
Leda	Clear, neutral

Turn Detection (agent/server.py)

turn_detection = smart_turn.TurnDetection(
    silence_duration_ms=1000,           # Wait 1s of silence
    speech_probability_threshold=0.4    # Sensitivity threshold
)

API Reference

Endpoint	Method	Description
/api/meeting/create	POST	Create a new meeting
/api/meeting/end	POST	End meeting, trigger summary
/api/session/[id]	GET	Get session details
/api/session/transcript	POST	Store session transcript
/api/session/summary	POST	Generate AI summary
/api/stream/token	POST	Generate Stream token
/api/auth/*	Various	Better-auth endpoints

Database Commands

npm run db:push    # Push schema changes to database
npm run db:studio  # Open Drizzle Studio

License

MIT