amal66 brings local AI into the Mike stack

amal66 has explored optional support for Ollama, software that runs language models locally, through Mike's existing AI-provider setup. Turn it on deliberately and selected local models become available; leave it off and the standard experience stays unchanged.

The proposal also puts guardrails around custom AI-service endpoints, blocking unsafe internal or private-network destinations in production unless they are explicitly permitted. That matters because flexible integrations can otherwise create an unexpected route into systems that should remain unreachable.

This work was tested but closed without being merged upstream, so it is a useful signal of the fork's direction rather than a capability in the main project.