Fix seed: run prisma generate before db push

Seed database with model metrics during Docker build

Revert sklearn to 1.5.1, fix multi_class at load time

Docker: remove tf-keras, keep native keras 3

Fix LSTM: better loader fallbacks, verify tf-keras

Fix Dockerfile: separate torch CPU install, add tf-keras

Fix: resilient predictions, sklearn 1.3.2, LSTM compat, CPU torch

Pin sklearn to 1.3.2 for model compatibility

Initial commit: Sentiment Analysis HF Space