Flat dataset: one record per ID, feedback updates in place

Fix share link labels, wrap long text in result card image

Switch to synchronous dataset push (no CommitScheduler), add HF_TOKEN for private dataset access

Add cross-session ID lookup: check memory, local files, then HF dataset

Remove ?url= param support, only ?id= for share links

Add persistent predictions dataset, ID-based share links, user feedback buttons

Add share link + result card to text tab, download image button for mobile, hide Gradio toolbar via CSS

Hide Gradio image toolbar (download/share/fullscreen)

Add copy link/image buttons, fix card footer, detect Google/redirect pages

Add shareable result card image for social media sharing

Add copy button to share link

Add server-side ?url= param: auto-fetches, switches to URL tab, shows results

Revert share link to hidden until result, keep JS url param

Fix share link: always visible, JS-based ?url= param handling

Remove demo.load callback, disable SSR - fix startup timeout

Add shareable URL links (?url= param), JS-site detection, fix examples

Fix text examples: use NHS news (human 0.646) and medical text (AI 0.664)

Revert to simple from_pretrained, ONNX files removed from model repo

Fix startup: download only needed files, skip 1.35GB ONNX

Switch to PyTorch - float32 ONNX too large for free CPU Space

Use float32 ONNX (exact PyTorch match, faster CPU inference)

Switch back to PyTorch inference - ONNX quantization flips borderline predictions

Fix URL examples: Wikipedia (human) + garryslist (AI), fix error handling

Fix: css/theme go in Blocks constructor not launch()

Fix error handling for gr.Label, swap broken ChatGPT link, fix deprecation warnings

Add URL examples, fix tab layout jumping

Switch to ONNX inference for faster CPU performance

Redesign UI with dark editorial theme, add URL tab, fix examples

Add URL fetching tab, remove share link button

Fix SDK version and requirements

Add AI text detector Gradio app

initial commit