Integrating Specialized Tool Serving with TorchServe

Refactor one of your simulated tools (e.g., a more complex `anomaly_detection_model(metrics: dict)` that classifies abnormal behavior) to be served by TorchServe. Your Claude agent should now make an HTTP request to this TorchServe endpoint when it needs to use this specialized model. Ensure the agent can correctly parse the output from the TorchServe API and incorporate it into its reasoning. Document the steps to package and deploy a simple PyTorch model with TorchServe.

This will demonstrate how agents can interact with independently deployed microservices.