🎯 AI Technical Interviewer

Powered by Google Agent Development Kit (ADK) & Gemini

Features • Quick Start • Architecture • Deploy • Contributing

---

v5.0.0 - A2UI Web Interface ✅ PRODUCTION READY (2025-12-30)
Beautiful, component-based web UI via A2A-ADK bridge. Cards rendering successfully - verified production-ready. Includes Sequential Safety, guided learning, and multi-dimensional scoring. Powered by Google's Agent Development Kit, Gemini, and A2UI. See A2UI Integration Journey.

Latest Update (2025-12-30 18:52 IST):

• ✅ A2UI Card rendering verified working with interview flow screenshots
• ✅ Multi-turn conversation with scoring (8.5/10) tested successfully
• ✅ Bridge fixes: path corrections, session handling, data variable initialization
• ✅ Production-ready implementation

---

✨ Features

Core Capabilities

Feature	Description
🎓 Guided Learning Mode	Study CS concepts with Socratic method & progressive hints
🎯 Multi-Agent Scoring	Parallel evaluation: technical, communication, problem-solving
🎚️ Difficulty Modes	Quick Screen (15min), Standard (45min), Deep Technical (90min)
💬 Answer Critique	Get improvement suggestions & validation feedback
🧠 Adaptive Questions	Dynamic difficulty based on performance
💻 Code Analysis	Review and analyze Python code logic
🛡️ Safety Screening	Content moderation & bias detection
⚡ Sequential Safety	Automated risk assessment blocks dangerous code (v4.6.0)
🌐 A2UI Web Interface	Beautiful Lit-based web UI with A2A-ADK bridge (v4.7.1)
📎 Resume Support	Paste resume text for analysis (file upload limited by Gemini)

Technical

Feature	Description
🌐 ADK Web UI	Beautiful interface out of the box
☁️ Cloud Ready	One-click deploy to GCP Cloud Run (Free Tier)
📊 Session State	Persistent interview sessions
🔄 Multi-Agent	6 specialized sub-agents with orchestration
📋 Context Files	Conductor-style config (.adk/) for team alignment (v4.6.0)

---

🚀 Quick Start

Prerequisites

• Python 3.11+
• Node.js 18+ (for A2UI frontend)
• Google AI Studio API Key (Free)

Installation

# Clone
git clone https://github.com/VIKAS9793/ai-interviewer-google-adk.git
cd ai-interviewer-google-adk
git checkout google-adk

# Setup
python -m venv .venv
.venv\Scripts\activate  # Windows | source .venv/bin/activate # Linux/Mac
pip install -r requirements.txt

# Configure
cp .env.example .env
# Add your GOOGLE_API_KEY to .env

# Run (IMPORTANT: Must run from src/ directory)
cd src
python -m google.adk.cli web

Open http://127.0.0.1:8000 🚀

A2UI Web Interface (v5.0.0) ✅

Status: Production-ready, verified 2025-12-30 with interview flow testing.

Run 3 terminals simultaneously:

# Terminal 1 - ADK Backend (Port 8000) - MUST run from src/
cd src
python -m google.adk.cli web

# Terminal 2 - A2A Bridge (Port 10002)
python src/adk_interviewer/a2ui/bridge.py

# Terminal 3 - A2UI Frontend (Port 3000)
cd a2ui-repo/samples/client/lit/shell
npx -y vite dev --port 3000

Open http://localhost:3000/?app=interviewer 🎨

✅ Verified Working: Cards render correctly, schema compliant, production-ready
⚠️ Windows Users: Use npx vite dev directly, NOT npm run dev (wireit incompatibility)

Interview Flow Screenshots

See the complete interview experience in action:

Step	Screenshot
1. Question
2. Answer
3. Evaluation
4. Feedback
5. Summary

---

🏗️ Architecture

Three-Tier Bridge Architecture

System Components

┌─────────────────────────────────────────────────────────────────────┐
│                    A2UI Frontend (v4.7.1)                           │
│  ┌─────────────────┐           ┌────────────────────────────────┐  │
│  │   Lit Renderer  │──────────▶│  A2A-ADK Bridge (:10002)       │  │
│  │   :3000         │   A2A     │  FastAPI · JSON-RPC Translator │  │
│  └─────────────────┘           └────────────────────────────────┘  │
└──────────────────────────────────────────┬──────────────────────────┘

                                           │
                                           ▼
┌─────────────────────────────────────────────────────────────────────┐
│                      ADK Backend (:8000)                            │
│  ┌─────────────┐  ┌──────────────────┐  ┌─────────────────────┐    │
│  │   Web UI    │  │ Session Service  │  │  run_sse Endpoint   │    │
│  └─────────────┘  └──────────────────┘  └─────────────────────┘    │
└──────────────────────────────────────────┬──────────────────────────┘
                                           │
                                           ▼
                  ┌────────────────────────────────────────────┐
                  │      root_agent (Orchestrator)             │
                  │  Routes tasks to specialist sub-agents     │
                  └───────────────┬────────────────────────────┘
                                  │
            ┌─────────────────────┴─────────────────────┐
            │           6 Specialist Sub-Agents         │
            ├───────────────────────────────────────────┤
            │ • interviewer_agent (Questions/Eval)      │
            │ • resume_agent     (Resume/JD Analysis)   │
            │ • coding_agent     (Code + Safety v4.6)   │──▶ Gemini 2.5
            │ • safety_agent     (Content Moderation)   │
            │ • study_agent      (Guided Learning)      │
            │ • critic_agent     (Answer Critique)      │
            └───────────────────────────────────────────┘

Tech Stack

Layer	Technology
Frontend (v4.7)	A2UI Lit Renderer
Bridge (v4.7)	FastAPI + httpx
Framework	Google Agent Development Kit
LLM	Gemini 2.5 Flash-Lite
Web UI	ADK Web (adk web)
Deployment	Google Cloud Run
State	ADK SessionService

---

💡 Usage Examples

Interview Mode

👤 "Start a system design interview"
🤖 Interviews you with adaptive questions

Study Mode (v4.2)

👤 "Explain binary search trees"
🎓 In-depth concept explanation with examples

👤 "Give me a hint for two-sum problem"
🎓 Level 1: Gentle direction
🎓 Level 2: Algorithm suggestion  
🎓 Level 3: Detailed pseudocode

Difficulty Modes (v4.4)

👤 "Quick screen for junior developer"
⚡ 15-min, 3-5 easy/medium questions

👤 "Standard technical interview"
🎯 45-min, comprehensive assessment

👤 "Deep technical for senior engineer"
🔬 90-min, expert-level questions

---

☁️ Deployment

Google Cloud Run (Free Tier)

# Authenticate
gcloud auth login

# Deploy (one command!)
gcloud run deploy ai-interviewer \
  --source . \
  --region us-central1 \
  --allow-unauthenticated \
  --set-env-vars GOOGLE_API_KEY=your_key_here

Free Tier: 2M requests/month, 360,000 GB-seconds

See docs/DEPLOYMENT.md for details.

---

📚 Documentation

Core Docs

• Architecture
• Setup Guide
• Deployment
• Changelog
• Security
• A2UI Integration Journey - Protocol bridge story

Research & Planning

• A2UI Analysis - Google's agent UI protocol
• A2UI Value Impact - ROI assessment
• Kaggle Insights - v4.6.0 inspiration

Architecture Decisions

• ADR-001: Migration to ADK

---

🤝 Contributing

Contributions welcome! See CONTRIBUTING.md.

---

� License

MIT License - see LICENSE

---

🙏 Acknowledgments

• Google ADK Team for the amazing framework
• Google A2UI Team for agent-to-user interface protocol
• Gemini AI for powering intelligence
• Built with Google Antigravity
• Open source community for inspiration

---

Built with ❤️ using Google's Agent Development Kit

⭐ Star on GitHub