@lukisxyz/pi-core-vision

Transparent vision fallback for text-only pi models — overrides read so image files are described by a vision model when the active model cannot see images. Supports any OpenAI-compatible endpoint.