EdukaAI Studio

Fine-tune Large Language Models on Apple Silicon with an intuitive web interface

EdukaAI Studio is a user-friendly application for fine-tuning Large Language Models (LLMs) on your Apple Silicon Mac. Built on Apple's MLX framework, it provides a simple 5-step wizard to train custom AI models - no coding required!

Perfect for:

• Students learning about AI/ML
• Professionals customizing models for specific domains
• Researchers experimenting with fine-tuning
• Hobbyists exploring LLM capabilities

Quick Start (5 Minutes)

Installation

Recommended install:

mkdir edukaai-studio
cd edukaai-studio
curl -fsSL https://raw.githubusercontent.com/elgap/edukaai-studio/main/install.sh | bash

Non-interactive install (for automation/CI):

mkdir edukaai-studio
cd edukaai-studio
curl -fsSL https://raw.githubusercontent.com/elgap/edukaai-studio/main/install.sh | bash -s -- --yes

The installer will:

• Check your Mac compatibility (Apple Silicon required)
• Install Python 3.10+ and Node.js 18+ (if missing)
• Download EdukaAI Studio to the current directory
• Install all dependencies
• Create a Desktop shortcut

Launch the Application

After installation:

# Option 1: Double-click 'EdukaAI-Studio.command' on your Desktop
# Option 2: Run from the install directory:
./launch.sh

Then open your browser: http://localhost:3030

Check out the Quick Start Guide for a visual step-by-step walkthrough with screenshots!

Your First Fine-Tuning (10 Minutes, no-code)

Zero-Prep Option: EdukaAi Started Pack

Don't have training data? No problem! Download EdukaAi starter pack from eduka.elgap.ai to get you going immediately.

EdukaAi Starter Pack includes:

• real world data (JSONL) ready for fine tuning
• an AI-generated story (PDF)
• Basic instructions (PDF)

⚡ Literally: Download Starter Pack → Import to EdukaAi Studio → Click Configure → Click Train → Sip Coffee → Click Chat to test your fine tuned model in dual chat

Manual Option: Create Your Own Data

Create a text file with training examples:

Example 1 - Q&A Format:

Question: What is machine learning?
Answer: Machine learning is a method of data analysis that automates analytical model building.

Question: How does a neural network work?
Answer: A neural network works by processing information through interconnected nodes organized in layers.

Example 2 - Instruction Format:

{"instruction": "Translate to French", "input": "Hello, how are you?", "output": "Bonjour, comment allez-vous?"}
{"instruction": "Summarize this article", "input": "The quick brown fox...", "output": "A fox jumped over a lazy dog."}

Save as my-training-data.jsonl (100-1000 examples recommended)

All-in-one option: AI Curator

Your local-first data preparation layer. Import samples from various sources. Review, rate, create custom Q&A pairs, and craft the perfect dataset before exporting to EdukaAI Studio.

Why Curator?

• 🏠 100% Local — Your data never leaves your Mac
• 🔌 Universal Import — Capture API accepts data from any source (Kaggle, Hugging Face, APIs, files)
• ✏️ Visual Editor — Review samples and create custom question/answer pairs
• 🎯 Studio Integration — One-click export to EdukaAI Studio
• ☕ Coffee-Break Speed — From raw data to fine-tuned model in minutes

---

Getting Help

Quick Start Guide: QUICKSTART.md - Visual step-by-step guide with screenshots

Report Issues: GitHub Issues

Discussions: GitHub Discussions

---

🛠️ Development

Tech Stack

Backend:

• FastAPI - High-performance async web framework
• MLX - Apple's machine learning framework
• SQLAlchemy - Database ORM
• Pydantic - Data validation
• Python-multipart - File uploads

Frontend:

• Vue.js 3 - Progressive JavaScript framework
• Pinia - State management
• Recharts - Data visualization
• Tailwind CSS - Utility-first CSS

ML Libraries:

• mlx-lm - Language model fine-tuning
• transformers - Model utilities
• safetensors - Efficient model storage

Quick Setup for Developers

# Clone the repository
git clone https://github.com/elgap/edukaai-studio.git
cd edukaai-studio

# Run the installer (automatically detects developer mode)
./install.sh

# Start the application
./launch.sh

The installer will:

• Set up Python virtual environment
• Install all dependencies
• Run the test suite
• Create storage directories

Project Structure

├── backend/
│   ├── app/
│   │   ├── routers/        # API endpoints
│   │   ├── ml/            # Training pipeline
│   │   ├── models.py      # Database schemas
│   │   └── main.py        # FastAPI app
│   ├── tests/             # Test suite
│   └── storage/           # Data storage
│
├── frontend/
│   ├── src/
│   │   ├── views/         # Page components
│   │   ├── components/    # Shared components
│   │   ├── stores/        # Pinia stores
│   │   └── router/        # Vue Router
│   └── public/           # Static assets
│
└── docs/                 # Documentation

Running Tests

# Backend tests
cd backend
source .venv/bin/activate
python -m pytest tests/ -v

# Frontend build check
cd frontend
npm run build

Environment Variables

EdukaAI Studio can be configured via environment variables. Copy .env.example to .env and customize:

cp .env.example .env

Key Configuration Options:

# Server Ports (change if ports are already in use)
EDUKAAI_PORT=8000                    # Backend API port (default: 8000)
VITE_PORT=3030                       # Frontend dev server port (default: 3030)

# API Connection (for custom backend locations)
VITE_API_URL=http://localhost:8000   # Backend API URL
VITE_WS_URL=ws://localhost:8000     # WebSocket URL

# Security
EDUKAAI_ALLOW_REMOTE=false          # Allow remote connections (not recommended)

# HuggingFace
EDUKAAI_HF_TOKEN=your-token         # For private models and higher rate limits

# Storage
EDUKAAI_STORAGE_PATH=./storage      # Data storage location
EDUKAAI_MAX_STORAGE_GB=50           # Max storage limit

# Database
EDUKAAI_DATABASE_URL=sqlite:///./storage/app/edukaai.db

# MLX / Training
EDUKAAI_MLX_DEVICE=gpu              # gpu or cpu
EDUKAAI_MAX_DATASET_SAMPLES=10000
EDUKAAI_MAX_CONTEXT_LENGTH=4096

See .env.example for all available options.

Custom Port Configuration

If the default ports (8000 backend, 3030 frontend) are already in use:

Option 1: Using .env file (Recommended) Edit .env in project root:

EDUKAAI_PORT=8080
VITE_PORT=3000
VITE_API_URL=http://localhost:8080
VITE_WS_URL=ws://localhost:8080

Then start normally:

./launch.sh

Option 2: Environment Variables

# Terminal 1 - Backend on port 8080
export EDUKAAI_PORT=8080
cd backend && source .venv/bin/activate && python run.py

# Terminal 2 - Frontend on port 3000, connecting to backend on 8080
export VITE_PORT=3000
export VITE_API_URL=http://localhost:8080
export VITE_WS_URL=ws://localhost:8080
cd frontend && npm run dev

Contributing

We welcome contributions! Contributing guide coming soon.

License

MIT License - see LICENSE file for details.

SECURITY

EdukaAI Studio security audits are ongoing. Report is here