sanchitishere/Deep-NIDS ? reverse-engineered prompt

Build me a local network intrusion detection app that uses machine learning to spot unusual traffic from network flow data. I want it to take raw network captures in parquet or csv, clean and prepare the data, train an autoencoder only on normal traffic, then use that model to flag suspicious activity based on reconstruction error.

When something looks abnormal, save it to a database so there’s a permanent audit trail, and make a simple live dashboard that shows the latest alerts and lets me adjust the detection threshold. Use Python and keep the whole thing runnable locally, with clear scripts for preprocessing, training, detection, database logging, and the dashboard.

If you need to, look up current docs online for TensorFlow, Streamlit, SQLAlchemy, and parquet handling. Please make it feel like a real SOC style demo, with sensible defaults, clean code, and enough setup instructions that I can run it end to end on my machine.