Meltdown: A desktop LLM client written in Python and Tk, supporting ChatGPT, Gemini, Claude, Kimi, and local models via llama.cpp

Meltdown is a desktop application for interacting with large language models (AI systems that understand and generate text), written 100% in Python and using tkinter for its graphical interface. It supports remote models from OpenAI (ChatGPT), Google (Gemini), Anthropic (Claude), and Moonshot (Kimi), as well as local models via llama.cpp inference engine.

The application features hundreds of arguments and commands, conversation tabs (up to 100 open at once by default), markdown rendering, snippets, search, memory, keyboard shortcuts, and profile support for independent configurations. Responses can be instant or streamed, and local models can be auto-unloaded after a configurable time period.

Installation is available via pipx for remote-only features or with optional local model support (llama.cpp). GPU acceleration via Vulkan is available for AMD GPUs; NVIDIA GPUs have not been tested. Data and config files are stored in the system's standard directories (~/.local/share and ~/.config on Linux).