MedOtter/medotter-sdk-fixtures

cmunext_busi.pth: model sidecar (medotter/model-sidecar/v1) and provenance

Add model sidecars: unet_busi_tiny_cfg.pth.config.json, unet_busi_tiny_cfg.pth.provenance.json (tools/hf_sidecars.py)

unet_busi_tiny_cfg.pth: tiny U_Net fixture with embedded top-level 'config' dict so InferenceAgent.from_checkpoint can drive it from a single hf: file (1-epoch/2-batch BUSI training, stem conv expanded 3->4ch for the 4-modality canary bundle; keys: epoch/model_state_dict/config; survives torch.load weights_only=True)

Add CMUNeXt/BUSI golden weight for tier5 SDK inference regression (val_dice 0.6656)

initial commit