stakwork brings local inference to mikeoss with Ollama

@pitoi merged a change that plugs Ollama - a runtime for running open-source language models locally - into mikeoss as a first-class option alongside whatever cloud providers were already wired in. The fork now knows how to route requests to Ollama and can ask it which models are available, so the UI doesn't need a hardcoded list.

The practical shift is the deployment story. Hosted providers mean prompts leave your network; a local runtime keeps them on your own machines. For legal teams nervous about client data crossing into third-party LLMs, that's a different conversation with the risk committee. The change was authored by Hive, stakwork's code-generation tooling, and landed straight into main.