GGUFloader/gguf-loader — reverse-engineered prompt

Build me a beginner friendly desktop app called GGUF Loader for running local AI chat models from GGUF files. I want someone to download the app, pick any GGUF model on their computer, load it with a progress indicator, then chat with it privately without sending anything online.

The app should have a clean main window, a floating chat button that stays available on screen, and a simple chat panel where users can type messages and get streaming replies. Please support common local models like Mistral, LLaMA, DeepSeek, Gemma, and Qwen as long as they are GGUF files.

Add an Agentic Mode where the user can choose a workspace folder and let the assistant read, create, edit, and organize files inside that folder only. Make this feel safe and clear, with permission prompts and visible logs of what the assistant is doing.

Please include easy launch scripts, basic settings, optional GPU support if available, and clear beginner instructions. Look up current docs online if you need to.