JupyterLab × RStudio × AI — on your desktop. No cloud required.

Features • Quick Start • Project Layout • Architecture

---

What is AcademiaML?

AcademiaML is a notebook-first desktop research workbench designed for scientists who have tabular data and need help choosing, running, and interpreting machine learning workflows — without leaving their local environment.

Think of it as JupyterLab + RStudio + AI advisor, packaged as a native desktop app:

Your tabular data
      │
      ▼
AcademiaML opens it in a real .ipynb notebook
      │
      ├──▶  Monaco editor  (VS Code-quality editing)
      ├──▶  Local Jupyter kernel  (real Python execution)
      ├──▶  AI advisor  (OpenAI-compatible or GitHub Copilot)
      ├──▶  Agent queue  (approve before any code runs)
      └──▶  KaTeX math rendering  (publication-ready output)

---

✨ Features

📓 Real Notebooks, Real Kernel Native .ipynb format — fully portable Local Jupyter kernel execution via @jupyterlab/services Monaco Editor (same engine as VS Code) Per-project Python virtual environments bootstrap_runtime.py auto-configures each project	🤖 AI Research Advisor GitHub Copilot SDK integration Local Copilot CLI agent queue OpenAI-compatible provider support Approvals gate: every AI action requires explicit sign-off KaTeX math rendering in AI responses
📊 Tabular Data First CSV parsing via PapaParser Auto-infers column types and schema Summary stats without exposing raw data data/raw/ → data/derived/ pipeline convention Sampled slice preview — no full upload	🔒 Local-First, Privacy-Respecting All computation on your machine Only schema + inferred types + summary stats used by AI by default Explicit approval required for any broader data access Project folders are plain directories — readable without the app

---

🚀 Quick Start

# Clone and install
git clone https://github.com/Xueyang-Song/academia-ml.git
cd academia-ml
npm install

# Run in development
npm run dev

# Bootstrap a new project (Python runtime + venv)
python scripts/bootstrap_runtime.py /path/to/your/project

# Build for production
npm run build

# Package as desktop app
npm run pack

---

📁 Project Layout

AcademiaML projects are plain directories — readable, portable, version-controllable:

my-research-project/
├── .academiaml/
│   └── project.json        # Project config and metadata
├── notebooks/              # .ipynb files (real Jupyter format)
├── data/
│   ├── raw/                # Original data, never modified
│   └── derived/            # Processed / feature-engineered data
├── generated/              # AI-generated scripts and outputs
├── artifacts/              # Plots, model files, exports
└── logs/                   # Kernel and agent execution logs

---

🏗 Architecture

academia-ml/
├── electron/               # Main process
│   ├── kernel/             # Jupyter kernel management
│   ├── agent/              # Copilot SDK + agent queue
│   └── python/             # Virtualenv bootstrap, script runner
├── src/                    # Renderer process (React + Vite)
│   ├── components/
│   │   ├── notebook/       # Cell editor (Monaco), output renderer
│   │   ├── advisor/        # AI chat panel (KaTeX, markdown)
│   │   ├── data/           # Table viewer, schema inspector
│   │   └── queue/          # Agent approval drawer
│   └── lib/                # Shared utilities
├── python_templates/       # Starter ML scripts per workflow type
├── prompts/                # AI advisor prompt templates
└── scripts/
    └── bootstrap_runtime.py

Tech Stack:

---

Inspiration

Designed to feel closer to how scientists actually work:

Tool	What AcademiaML borrows
JupyterLab	Real .ipynb notebooks, kernel execution
RStudio	Project-centric workspace, data panel
MATLAB	Integrated environment, no context switching
GitHub Copilot	AI that asks before acting

---

For scientists who want to run ML, not manage infrastructure.