update backend to Bi-LSTM model and fix paths

feat: implement dual-input system and dedicated AI detection for maximum accuracy

feat: initial clean project setup (root-level)