Getting Started

Requirements

• macOS 13 (Ventura) or later
• Apple Silicon Mac (M1, M2, M3, or M4)
• Node.js 18+
• Python 3.10+

Install from Source

git clone https://github.com/fabriziosalmi/silicondev.git
cd silicondev

# Frontend dependencies
npm install

# Backend dependencies
cd backend
python -m venv .venv
source .venv/bin/activate
pip install -c constraints.txt -e ".[dev]"
cd ..

Run in Development Mode

npm run dev

This starts both the Vite dev server (port 5173) and the Electron shell. The backend FastAPI server starts automatically, binding to 127.0.0.1. It tries port 8000 first and scans up to 8099 if 8000 is busy. The chosen port is signalled to Electron via stdout (SILICON_PORT=<port>).

Build for Distribution

npm run package

Produces a .dmg and .zip in the release/ directory. The backend is bundled via PyInstaller.

First Steps

1. Open the app. Wait for the backend health check (green dot in the top bar).
2. Go to Models and download a model from Hugging Face (e.g., mlx-community/Qwen3-1.7B-MLX-8bit).
3. Click the model name in the top bar to load it into memory.
4. Switch to Chat and start a conversation.

Storage

All user data is stored in ~/.silicon-studio/:

Path	Contents
models.json	Model registry (names, paths, status)
models/	Downloaded model files
adapters/	Fine-tuned LoRA adapters
conversations/	Chat history as JSON files
notes/	Markdown notes as JSON files
agents/agents.json	Agent workflow definitions
rag/	RAG collections and chunks
mcp_servers.json	MCP server configurations
logs/app.log	Backend rotating log (5 MB, 3 files)

No data is sent to external servers. Everything stays on disk.