🧠 DevMentor AI

Real-Time AI Coding Mentor for Beginner Developers

Built for WeMakeDevs Hack All February Hackathon — powered by Vision Agents SDK + Stream + Gemini

---

🎯 The Problem

Millions of beginner developers get stuck on simple bugs and concepts — and have nobody to ask. Stack Overflow is intimidating. YouTube tutorials don't answer YOUR specific question. Hiring a tutor costs money.

DevMentor AI solves this — a real-time AI mentor that listens to your voice, watches your code, and teaches you like a patient senior developer. Free. Always available.

This directly aligns with WeMakeDevs' mission: "Quality education, free for all."

---

✨ Features

Feature	Description
🎤 Voice Interaction	Speak your question naturally — AI responds with voice
👁️ Screen Monitoring	AI watches your code in real-time via Stream Video
💬 Text Chat	Type questions and get instant AI responses
🌐 Multi-Language	Python, JavaScript, Java, C++, C, TypeScript
▶️ Code Execution	Run code directly in the browser terminal
📊 Progress Tracking	Skill tags, progress bar, session stats
⏱️ Session Timer	Live session tracking
🔴 Live Connection	Real-time Stream edge connection

---

🏗️ Architecture

┌─────────────────────────────────────────────────────┐
│                    DevMentor AI                      │
├──────────────────┬──────────────────────────────────┤
│   Frontend       │         Backend                  │
│   React + Vite   │   Vision Agents SDK (Python)     │
│                  │                                  │
│  Stream Video    │   Agent ──► Gemini Realtime      │
│  React SDK       │   Edge  ──► GetStream Edge       │
│                  │   STT   ──► Deepgram             │
│  Voice Input     │                                  │
│  Code Editor     │   Token Server (FastAPI)         │
│  Terminal        │                                  │
└──────────────────┴──────────────────────────────────┘

---

🛠️ Tech Stack

• Vision Agents SDK — Agent orchestration, voice pipeline, video monitoring
• Stream Video — Real-time audio/video edge network (333k free minutes!)
• Gemini 2.5 Flash Native Audio — Speech-to-speech AI model
• Deepgram — Speech-to-text transcription
• React + Vite — Frontend
• FastAPI — Token server backend

---

🚀 Getting Started

Prerequisites

• Python 3.11
• Node.js 18+
• API Keys: Stream, Gemini, Deepgram

1. Clone the repo

git clone https://github.com/jass2422/DevMentorAI.git
cd DevMentorAI

2. Backend Setup

cd backend
python -m venv venv
venv\Scripts\activate  # Windows
pip install -r requirements.txt

3. Create .env file in backend/

STREAM_API_KEY=your_stream_api_key
STREAM_API_SECRET=your_stream_api_secret
GEMINI_API_KEY=your_gemini_api_key
DEEPGRAM_API_KEY=your_deepgram_api_key

4. Frontend Setup

cd frontend
npm install

5. Create frontend/.env.local

VITE_STREAM_API_KEY=your_stream_api_key

6. Run Everything

Terminal 1 — AI Agent:

cd backend && python agent.py

Terminal 2 — Token Server:

cd backend && python token_server.py

Terminal 3 — Frontend:

cd frontend && npm run dev

Open http://localhost:5173 🎉

---

🎬 How It Works

1. Student opens DevMentor AI and enters their name
2. Selects their programming language (Python, JS, Java, C++, C, TypeScript)
3. Joins the live session — connects to Stream's edge network
4. The Vision Agents AI agent is already waiting in the call
5. Student speaks their coding question OR types it in chat
6. Gemini Realtime processes audio and responds with voice
7. Session tracks questions asked, bugs fixed, and skill progress

---

💡 Vision Agents SDK Usage

This project showcases multiple Vision Agents SDK features:

# Agent with Gemini Realtime voice
agent = Agent(
    edge=GetStreamEdge(...),       # Stream edge connection
    agent_user=agent_user,         # AI participant
    instructions="...",            # Mentor personality
    llm=GeminiRealtime(            # Voice AI
        model="gemini-2.5-flash-native-audio-preview-12-2025"
    ),
)

# Join the call and listen
async with agent.join(call):
    # AI is now live — listening and responding with voice!

---

🌍 Impact

• Free — powered by free tiers of Stream (333k minutes), Gemini, and Deepgram
• Accessible — works in any browser, no install needed
• Multi-language — serves developers learning any major language
• Voice-first — natural interaction, just like talking to a real mentor

---

👩‍💻 Built By

Jasmeen — WeMakeDevs Community Member

Built with ❤️ for the WeMakeDevs Hack All February Hackathon 2025

---

📄 License

MIT