sumit7521/cyber-ids-api

Update additional models and configs

Add robust_xgb_model API endpoint with LFS tracking

Add missing mlp_ids_model.keras file

Fix MLP model to use hybrid preprocessing

Fix architecture to support dual scalers for old vs new models

Fix scaler feature name matching error by passing values array

Update models to new PKL/Keras versions and add new API endpoints

Fix Random Forest (110 features) and CNN (118 features) dimension mismatches by dynamically reindexing input features to their specific training subsets

Add Hugging Face metadata config to README.md

Track large models with Git LFS

Add Dockerfile for Hugging Face

Implement real-time SHAP Explainable AI (XAI) feature importances for Hybrid model endpoint + add shap to requirements.txt

Create comprehensive README.md detailing multi-model architectures, FastAPI endpoints, and Render memory optimizations

Optimize RAM usage by implementing Lazy Loading for heavy models (CNN, KNN, Random Forest, CatBoost, Autoencoder) to prevent Out of Memory on Render Free Tier

Fix XGBoost and CatBoost model loading by switching from .load_model to joblib.load for pickle compatibility

Update Python version to 3.11.9 to satisfy Keras 3.14.1 requirement of Python >= 3.11

Pin keras to 3.14.1 to fix quantization_config deserialization crash when loading Keras models

Specify Python 3.10.14 version for Render deployment to ensure package compatibility

Optimize requirements.txt for Render deployment by removing heavy PyTorch and switching to tensorflow-cpu

Initial commit with all trained models and FastAPI endpoints