添加本地 AI：运行经过您的 GPU 验证的模型，并在其上打开编程代理

Add Local AI: run the model validated for your GPU and open a coding agent on it Replaces #12845. Local AI is a bar widget and one command. It finds the GPUs in the machine and, for each card, offers the models that were accepted on that exact card, the recommended one first. One click downloads and checks a model, runs it, and opens a coding agent on it in the folder you pick: Would be awesome if it could also automatically configure/reconfigure custom endpoints in Hermes especially whenever the model is changed. I proposed a design pass on the panel (APCA contrast, layout rhythm, button centering) as a PR onto this branch: https://github.com/0xSero/omarchy/pull/2