Fix scikit-learn version mismatch (upgrade to 1.6.1)

Add version diagnostics to debug model loading

Better error parsing in frontend

Add detailed error logging to back debugging

Improve error logging in frontend

Fix Dockerfile: Remove lfs model copy, using hub download instead

Update frontend to use HuggingFace backend URL

Add HuggingFace Space metadata to README

Initial commit: Sentiment analysis app (models on HuggingFace)