Problem: Traditional technical interview practice often lacks realism, immediate feedback, and interactive voice-based engagement.
Solution: InterviewLab delivers AI-driven technical interviews using real-time voice conversations, live code execution, and in-depth feedback, powered by LangGraph and LiveKit.
Python3.11+TypeScript5.0+LangGraph0.0.40+LiveKit0.11.0+OpenAI1.0.0+LicenseGNUStatusPortfolio-Project
Portfolio Project — Production-ready codebase demonstrating AI system architecture.
Provide candidates with realistic interview practice through:
- Natural voice conversations with AI interviewer
- Live code execution in isolated sandbox
- Comprehensive feedback on communication, technical knowledge, problem-solving, and code quality
- Resume-based questions tailored to candidate background
graph TB
subgraph Frontend
FE[Next.js React App]
end
subgraph Backend
API[FastAPI Server]
ORCH[LangGraph Orchestrator]
end
subgraph Voice
LK[LiveKit Server]
AGENT[LiveKit Agent]
TTS[OpenAI TTS]
STT[OpenAI STT]
end
subgraph Services
SB[Docker Sandbox]
LLM[GPT-4o-mini]
DB[PostgreSQL]
REDIS[Redis Cache]
end
FE -->|HTTP REST| API
FE -->|WebSocket| LK
API -->|HTTP| LK
API -->|SQL| DB
API -->|Cache| REDIS
LK -->|WebSocket| AGENT
AGENT -->|LangGraph| ORCH
ORCH -->|API| LLM
ORCH -->|Docker| SB
AGENT -->|API| TTS
AGENT -->|API| STT
| Component | Technology | Purpose |
|---|---|---|
| Orchestrator | LangGraph | State machine managing interview flow |
| Agent | LiveKit Agents | Real-time voice agent (STT/TTS) |
| LLM | GPT-4o-mini | Question generation, decision making |
| Sandbox | Docker | Isolated code execution |
| Database | PostgreSQL | Interview state, checkpoints |
| Cache | Redis | State caching, session management |
sequenceDiagram
participant U as User
participant F as Frontend
participant A as API
participant LK as LiveKit
participant AG as Agent
participant O as Orchestrator
participant LLM as GPT-4o-mini
U->>F: Start Interview
F->>A: POST interviews
A->>LK: Create Room
F->>LK: Connect WebSocket
LK->>AG: Bootstrap Agent
AG->>O: Initialize
O->>LLM: Generate Greeting
LLM->>O: Response
O->>AG: next_message
AG->>LK: TTS Audio
LK->>U: Hear Greeting
loop Conversation
U->>LK: Speak
LK->>AG: STT Text
AG->>O: execute_step
O->>LLM: Detect Intent
O->>LLM: Decide Next Action
O->>LLM: Generate Response
O->>AG: Response
AG->>U: TTS Audio
end
- LangGraph MemorySaver: In-memory state per interview (
thread_id) - Database Checkpoints: Persistent state after each turn
- Reducers: Append-only fields (conversation_history, questions_asked)
- Single Writer: Critical fields (next_message, phase) written by one node
- ✅ Real-time voice with <3s latency
- ✅ State persistence via checkpoints
- ✅ Concurrent interviews (isolated by thread_id)
- ✅ Code execution in isolated Docker containers
- ✅ Comprehensive feedback with skill breakdowns
InterviewLab/
├── src/ # Backend (Python/FastAPI)
│ ├── agents/ # LiveKit agent logic
│ ├── api/ # REST API implementation
│ │ └── v1/
│ │ └── endpoints/ # Endpoints for interviews, resumes, voice, sandbox
│ ├── core/ # Configuration, database, and authentication utilities
│ ├── models/ # Database models for core entities
│ ├── schemas/ # Pydantic schemas for data validation
│ └── services/ # Business logic and subsystems
│ ├── analysis/ # Interview response and code analysis
│ ├── analytics/ # Analytics functionality
│ ├── data/ # Checkpointing and state management
│ ├── execution/ # Secure code sandboxing
│ ├── logging/ # Interview activity logging
│ ├── orchestrator/ # State orchestration using LangGraph
│ └── voice/ # LiveKit voice management
├── frontend/ # Frontend (Next.js + React)
│ ├── app/ # App routing and authentication
│ ├── components/ # UI components (interview, analytics, UI kit)
│ ├── lib/ # API client and store utilities
│ └── hooks/ # Custom React hooks
├── docs/ # Documentation and guides
├── alembic/ # Database migration scripts
├── docker-compose.yml # Local dev orchestration
├── Dockerfile # Production build configuration
└── pyproject.toml # Backend dependencies and settings
- Architecture - System architecture and component relationships
- API Reference - REST API endpoints
- Frontend - Next.js frontend architecture and development
- Voice Infrastructure - LiveKit setup and agent architecture
- User Guide - How to use InterviewLab
- Local Development - Setup and development workflow
- LangGraph Guide - State, nodes, and orchestration
- Deployment - Railway and Vercel deployment
# Backendcd src
uvicorn main:app --reload
# Frontendcd frontend
npm install
npm run dev
# Agent (requires LiveKit server)
python -m src.agents.interview_agentSee Local Development for detailed setup.
- FastAPI - Modern async web framework
- Python 3.11+ - Programming language
- LangGraph 0.0.40+ - State machine orchestration
- SQLAlchemy 2.0+ - ORM with async support
- Alembic - Database migrations
- LiveKit Agents - Real-time voice agents
- OpenAI GPT-4o-mini - LLM for question generation
- Instructor - Structured LLM outputs
- PostgreSQL - Primary database
- Redis - Caching and state management
- Docker - Code sandbox execution
- Next.js 16.1 - React framework
- TypeScript 5.0+ - Type safety
- React 19.2 - UI library
- Tailwind CSS 4 - Styling
- Zustand - State management
- TanStack Query - Data fetching
- Monaco Editor - Code editor
- Framer Motion - Animations
- LiveKit Client - WebRTC integration
- Railway - Backend and agent hosting
- Vercel - Frontend hosting
GNU General Public License v3.0