pi-antigravity-imagen

A premium image generation suite for pi-coding-agent, powered by Google's Gemini 3 Pro Image model via Antigravity.

✨ Features

• The Studio: A distraction-free ASCII TUI modal for crafting prompts.
• Magic Enhance: Press Ctrl+E to transform simple ideas into professional art prompts using Gemini 3 Flash.
• Style Prompts: Define a global artistic style (e.g., "Moebius", "Cyberpunk") that persists across all your sessions (Ctrl+S).
• Output Control: Customize save directory and filename patterns (Ctrl+O).
• Native Editing: Full text navigation, selection, and undo/redo support.

📥 Installation

Clone the repository into your pi extensions folder:

cd ~/.pi/agent/extensions
git clone https://github.com/ditfetzt/pi-antigravity-imagen.git

Restart pi to load the extension.

🚀 Usage

Run the command to open the Studio:

/imagen

Controls

Key	Action
Enter	Generate Image (in Prompt Mode)
Ctrl+E	Magic Enhance (Rewrite prompt)
Ctrl+S	Style Editor (Set global style)
Ctrl+O	Output Settings (Path & Filename)
Ctrl+R	Cycle Aspect Ratio (1:1, 16:9, etc.)
Esc	Close / Return

🎨 Style Prompts

Press Ctrl+S to open the Global Style Editor. Text entered here is sent as a system instruction to the model for every generation.

• Example: "In the style of Jean Giraud Moebius, Ligne Claire, clean black ink outlines, flat pastel colors."
• This setting persists globally across all your Pi sessions.

⚙️ Configuration

Press Ctrl+O to configure output settings:

• Save Directory: Default: ~/.pi/agent/generated-images
• Filename Pattern: Default: imagen-{timestamp}
  • Supports {timestamp} placeholder.

Configuration is saved to ~/.pi/agent/imagen-config.json.

Requirements

• Authenticated Google Antigravity session (via auth.json managed by Pi).
• @mariozechner/pi-coding-agent (Peer Dependency).

License

MIT