llampart 1.0.0 polishes local llama-server UI

This is a UX-first release, not a backend breakthrough, but that is exactly the point: local LLM stacks get much more usable when the front end stops feeling like a throwaway demo.

• –The strongest differentiator is polish: Frosted Glass, wallpaper customization, and a denser sidebar make the app feel closer to a desktop client than a server UI
• –MCP flows, extended settings, and minimal reasoning/tools display suggest it is aimed at power users who live inside local workflows
• –The translation support and selective conversation management are practical quality-of-life additions, especially for long-running local sessions
• –Because it sits on top of `llama-server`, its value depends on how much you already want the llama.cpp ecosystem, but it lowers the friction for adopting it
• –The project being MIT-licensed and explicit about attribution/licensing makes it easier to evaluate and reuse than many local AI UI side projects