Graft

A command-line tool for AI-powered image generation and editing using OpenRouter API.

3G-CLI Suite: Glean (text analysis) | Glimpse (image analysis) | Graft (image generation and editing)

Installation

1. Clone this repository
2. Set up a virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

Configuration

Create ~/.graft_cfg in your home directory:

[openrouter]
api_key = your_openrouter_api_key_here
model = google/gemini-2.5-flash-image-preview
temperature = 0.4

Usage

Generate new images:

python3 graft.py -p "A sunset over mountains"
echo "A futuristic city" | python3 graft.py

Edit existing images:

# Edit image from file
python3 graft.py -i photo.jpg -p "make it black and white"

# Edit image from URL via curl
curl https://example.com/image.png | python3 graft.py -p "turn it into an orange logo"

Advanced options:

# Override model and temperature
python3 graft.py -p "Abstract art" -m google/gemini-2.5-flash-image-preview -t 0.8

# Specify output filename
python3 graft.py -p "street scene" -o street.png

# Multiple images saved as street.png, street_2.png, street_3.png, etc.
python3 graft.py -p "create 3 variations of a street scene" -o street.png

# Output to stdout for piping (only first image if multiple are generated)
python3 graft.py -p "vintage car" -o - | glimpse -p "what car model is this?"

# Process multiple prompts from a file (handles prompts with spaces)
while read p; do graft -p "$p" < /dev/null; done < prompts.txt

# List available models
python3 graft.py --list-models

Features

• ✅ Generate images from text prompts
• ✅ Edit existing images with text instructions
• ✅ Support for file input (-i image.png)
• ✅ Support for piped image data (curl | graft)
• ✅ Custom output filename (-o filename.png)
• ✅ Output to stdout for piping (-o -)
• ✅ Multiple output images (saved as image.png, image_2.png, etc.)
• ✅ Multiple image formats (PNG, JPEG, GIF, WebP, BMP)
• ✅ Base64 image handling
• ✅ Automatic filename generation with timestamps (now include sanitized model id: graft_20260103_140500_google_gemini_25_flash_image_preview_A_sunset_over_mountains.png)

Supported Models

• google/gemini-2.5-flash-image-preview (default)
• Other OpenRouter models with image output capability

Use --list-models to see all available options.