jeyprabu/systemlogs-anomaly-detection ? reverse-engineered prompt

Build me a small containerized anomaly detection system for system logs.

I want a backend that accepts metrics like CPU, memory, disk I O, network latency, process count, thread count, and a few other system values through a REST endpoint, then sends that data to a separate Python machine learning service for real time anomaly detection. The ML service should use a pretrained Isolation Forest model and return whether the input looks normal or anomalous, an anomaly score, a severity level like low, medium, or critical, and the main metrics that seem to be driving the result.

The main API should validate the request, call the ML service, store the log and prediction in MongoDB Atlas, and return one combined response to the client. Please make it runnable with Docker Compose so both services start together. Include a simple test request example I can try in Postman, and if you need current library details or setup steps, look up the latest docs online.