fix: use compatible Gradio runtime

chore: remove retired case-study image artifacts

release: publish claim-safe Q9UQB9 custom prediction workflow

Deploy BioInteract Gradio demo

fix: monkey-patch gradio_client bool-schema bug, remove ssr_mode

fix: eager ESM-2 load at startup, ssr_mode=False

fix: sdk_version 5.20.0, show_api=False to suppress /info bug

fix: upgrade sdk_version to 5.9.1 for huggingface_hub 1.x compat

fix: remove huggingface_hub.py shadow file and pycache

fix: remove gradio from requirements (managed by sdk_version)

fix: python_version, requirements, academic UI redesign

Minimise requirements: remove heavy deps so Space can build

Lazy-load model and heavy deps to avoid build-time installs

Set python_version: 3.10 to avoid Py3.13 audioop issue

Add shim huggingface_hub.py to satisfy gradio.oauth imports

Pin huggingface-hub==0.14.1 for gradio oauth compatibility

Add audioop-lts for Python 3.13 pydub/gradio compatibility

Revert python_version key; use default runtime

Set python_version to 3.10 to fix gradio/pydub audioop error on py3.13

Fix: torch==2.6.0+cpu, drop C++ PyG extensions (use native scatter fallback)

Guard PyG import; use PyGBatch alias

Fix deps: add PyG CPU find-links + torch_scatter/sparse/cluster

Deploy BioInteract Gradio demo

Fix requirements: use torch CPU index, remove scatter deps and gradio duplicate

Deploy BioInteract Gradio demo

initial commit