Haseeb949/fluenta-backend

Fix zip extraction for Windows backslash paths

Deploy user-provided retrained models (clean)

Deploy restored user models via archive.zip

Prioritize final scaler and feature_count to avoid LFS pointers

Prioritize stutter_model_final.pkl to avoid LFS pointers

Enhance debug-fs to inspect zip file

Fix critical bug: Call startup_load() globally for Gunicorn

Rename models.zip to archive.zip

Add models.zip with LFS tracking

Add /debug-fs endpoint

Fix model loading: Extract models.zip at startup to bypass LFS

Replace corrupt retrained models with valid final models

Upload real model files (LFS tracked)

Move model file to models_retrained for correct Docker deployment

Add Hugging Face README config

Track pkl files with LFS

Switch to port 7860 for Hugging Face

Move files to root for Hugging Face

Initial backend commit