Fix ZeroGPU duration quota for Vision API image detection.

Add Wizara Vision API endpoints without changing UI or inference.

Add Wizara Vision API endpoints without changing UI or inference.

ui: default OCR task to slow inference mode

ui: widen layout and give media its own dedicated canvas zone

fix(example): correct mislabeled Sweet example to Sushi

ui: make Detected Target Overlays a fixed-height scroll panel

ui: expand decoding trace to full width without scroll clipping

feat(log): restore user upload logging to HF dataset

ui: restore staggered detection pop-in and add usage guide

gradio server (#1)

fix(log): write one JSONL file per record with session_id to prevent overwrite

docs: switch arXiv citation to tags+body link format

docs: cite arXiv 2605.27365 in README to link Space from paper page

ui: add resolution note next to header about magi-attention / 1K resize

fix(requirements): bump torch to 2.8.0 / torchvision to 0.23.0 for ZeroGPU compatibility

Update README.md

fix private auth and zerogpu build

demo

initial commit