mirror of
https://github.com/different-ai/openwork
synced 2026-05-09 16:52:02 +02:00
Images are now compressed via canvas resize (max 2048px) + JPEG transcode (quality 0.82, target <1.5MB) before base64 encoding. Uses OffscreenCanvas when available to avoid blocking the main thread. Adds a post-encoding size pre-check to show an explicit error instead of silently failing. Fixes #481