GameSummoner: ease of local use update

Hey LocalLLaMA, I released GameSummoner a bit over a week ago as an open source program and said "it works locally, but is a bit hard to use." (Post) Over the last few days I've cleaned up my prod host (https://gamesummoner.com) and made it much easier to run locally. It's now one line: git clone https://github.com/ndamiano/GameSummoner.git && cd GameSummoner && ./install.sh && GameSummoner install.sh lists the models your llama-swap serves, asks which one plays each part, checks ComfyUI has every node and model file the workflows need, and asks before installing anything. You need: llama-swap, which switches between the models. LLM Anything that handles tool calls. I use Qwen 3.8 Flash-Next IQ3_S ComfyUI with the models listed in models.md, also behind llama-swap A decision model. I use JEV-27B-VL Setup details are in docs/llama-swap.md. Once everything has been picked, run GameSummoner for a terminal dashboard of every game you've built (the video), or go straight to GameSummoner summon "roguelike deckbuilder where you're a wizard cat climbing a tower". I've been using it locally with Qwen3.8 Flash-Next (IQ3_S on Basalt) and JEV 27B on my 5090 with 64 gb of RAM, and it's been a blast. I also ran the same ask through frontier models in the builder seat: Claude Opus 5.5: https://gamesummoner.com/g/r12ctNpfCeWH (the game in the video) DeepSeek V4.1 Flash: https://gamesummoner.com/g/jMvHoZ1bCo0h Gemini 4 Argon: https://gamesummoner.com/g/96sbF0IP6HI4 Local Flash-Next, for comparison: https://gamesummoner.com/g/GZ7rQoheqoE9 I will be adding a "bring your own key" type local inference solution in the next couple of days so that you can use whatever inference provider you want without it being behind llama-swap