# 0.2.4

## Features

- **image_gen** now uses the higher-quality grok-imagine-image-quality model.

## Bug Fixes

- **read_file** now correctly passes embedded base64 images to the model as vision tokens instead of truncated fragments.