100% Open-Source • Zero API Costs • Self-Hosted LLM
AI-powered career guidance platform that generates personalized learning roadmaps, skills analysis, and interview preparation — all running on your own infrastructure with zero usage costs.
| Feature | Description |
|---|---|
| 🎯 Career Roadmaps | 6-month personalized learning paths with weekly breakdowns |
| 📚 Skills Analysis | Identify transferable skills and learning priorities |
| 💼 Interview Prep | Technical & behavioral questions with answer frameworks |
| 🧠 Knowledge Quizzes | 15 MCQs per step with explanations (80% pass rate) |
| 🎬 YouTube Tutorials | Auto-curated full course videos (no shorts!) |
| 📊 Progress Tracking | Visual completion tracking with PDF export |
┌─────────────────┐ ┌──────────────────┐ ┌─────────────────┐
│ Frontend │ │ Backend API │ │ LLM Service │
│ (React/Vite) │────▶│ (FastAPI) │────▶│ (Ollama) │
│ Port: 5173 │ │ Port: 8000 │ │ Port: 11434 │
└─────────────────┘ └──────────────────┘ └─────────────────┘
│
▼
┌─────────────────┐
│ Mistral 7B │
│ (Apache 2.0) │
└─────────────────┘
| Model | License | VRAM | Speed | Context |
|---|---|---|---|---|
| Mistral 7B ✓ | Apache 2.0 | 6GB | 127 tok/s | 32K |
| LLaMA 3 8B | Meta License* | 8GB | 67 tok/s | 8K |
| Mixtral 8x7B | Apache 2.0 | 24GB+ | 45 tok/s | 32K |
Mistral 7B is truly open with zero licensing friction for commercial use.
- macOS/Linux or WSL on Windows
- 8GB+ RAM (16GB recommended)
- Python 3.10+
- Node.js 18+
# Install Ollama
curl -fsSL https://ollama.com/install.sh | sh
# Start Ollama service
ollama serve
# Download Mistral 7B (4.1GB, takes 2-5 minutes)
ollama pull mistral:7b-instruct-v0.3-q4_K_Mcd backend
# Create virtual environment
python -m venv venv
source venv/bin/activate # Windows: venv\Scripts\activate
# Install dependencies
pip install -r requirements.txt
# Copy environment file
cp .env.example .env
# Run the server
uvicorn main_v2:app --reload --port 8000cd frontend
npm install
npm run devVisit http://localhost:5173 🎉
| Method | Endpoint | Description |
|---|---|---|
POST |
/api/ai/roadmap |
Generate career roadmap |
POST |
/api/ai/skills |
Analyze skills & recommendations |
POST |
/api/ai/interview-prep |
Interview preparation guide |
POST |
/api/ai/quiz |
Generate knowledge quiz |
| Method | Endpoint | Description |
|---|---|---|
GET |
/api/health |
System health status |
GET |
/api/health/llm |
LLM service status |
GET |
/api/health/models |
Available models |
| Method | Endpoint | Maps To |
|---|---|---|
POST |
/generate-path |
/api/ai/roadmap |
POST |
/generate-quiz |
/api/ai/quiz |
Full API docs: http://localhost:8000/docs
# Clone and run everything
docker-compose up -d
# Wait for model to download (first run only)
docker-compose logs -f model-loader# Just the LLM service
cd llm-service
docker-compose up -d
# Verify it's running
curl http://localhost:11434/api/tags- Hardware: 8GB RAM, any modern CPU
- Performance: ~3-5s per roadmap generation
- Best for: Development, personal use
Provider RAM vCPUs Cost/mo
─────────────────────────────────────
Hetzner 16GB 4 $18
Contabo 16GB 4 $12
DigitalOcean 16GB 4 $48
Provider GPU VRAM Cost/hr
────────────────────────────────────────
RunPod RTX 3060 12GB $0.20
Vast.ai RTX 3080 10GB $0.25
Lambda RTX 4090 24GB $0.50
GPU gives 10-20x faster inference!
career-forge/
├── backend/
│ ├── main_v2.py # FastAPI app (open-source LLM)
│ ├── config.py # Environment configuration
│ ├── api/
│ │ ├── routes/ # API endpoint handlers
│ │ └── schemas/ # Request/response models
│ ├── services/
│ │ ├── llm_service.py # Ollama integration
│ │ ├── roadmap_service.py # Roadmap generation
│ │ ├── skills_service.py # Skills analysis
│ │ ├── interview_service.py # Interview prep
│ │ └── quiz_service_v2.py # Quiz generation
│ ├── prompts/ # Prompt templates
│ └── utils/ # Helpers
│
├── llm-service/
│ ├── docker-compose.yml # Ollama deployment
│ ├── Modelfile # Custom model config
│ └── scripts/
│ ├── setup.sh # Installation script
│ ├── health-check.sh # Monitoring
│ └── benchmark.sh # Performance testing
│
├── frontend/ # React/Vite app
├── docker-compose.yml # Full stack deployment
└── ARCHITECTURE.md # Detailed architecture docs
All settings via environment variables:
# LLM Service
LLM_BASE_URL=http://localhost:11434
LLM_MODEL=mistral:7b-instruct-v0.3-q4_K_M
LLM_TIMEOUT=120
# API Server
API_PORT=8000
CORS_ORIGINS=http://localhost:5173
# Features
YOUTUBE_ENABLED=true
RATE_LIMIT=30See .env.example for all options.
- ✅ Rate limiting on AI endpoints
- ✅ Input validation & sanitization
- ✅ Prompt injection protection
- ✅ No secrets in code
- ✅ CORS configuration
- ✅ Non-root Docker containers
# deploy.yaml (Kubernetes)
apiVersion: apps/v1
kind: Deployment
metadata:
name: careerforge-api
spec:
replicas: 3 # Scale API pods
...Add Redis for caching common roadmaps:
# Cache similar profile queries
cache_key = hash(user_profile)
if cached := redis.get(cache_key):
return cachedHot-swap to stronger models when resources available:
# Upgrade to Mixtral (needs 24GB+ VRAM)
ollama pull mixtral:8x7b-instruct-v0.1-q4_K_M
# Update .env
LLM_MODEL=mixtral:8x7b-instruct-v0.1-q4_K_MBeautiful gradient UI with purple theme, 3D loading animations, interactive quiz modals, and progress visualization.
cd backend
pytest tests/ -vblack .
isort .
mypy .cd llm-service/scripts
./benchmark.sh- Fork the repository
- Create feature branch (
git checkout -b feature/amazing) - Commit changes (
git commit -m 'Add amazing feature') - Push to branch (
git push origin feature/amazing) - Open a Pull Request
- This Project: MIT License
- Mistral 7B Model: Apache 2.0 License
Harsh Saini
Made with ❤️ for developers who want AI without the API bill.
- Mistral AI for the amazing open-source model
- Ollama for making local LLM deployment easy
- The open-source community for making this possible
"BC… ye banda serious hai." 🚀