comfyui src 5daadca94e57fae8dadd78e14c14d2eeb763b36d99a4b91e2de4d1409278a9a4 Modular diffusion model GUI, API and backend ComfyUI is a node-based interface for running diffusion models (Qwen-Image, Flux, SDXL, …). The RPM ships the application only; checkpoints go in /srv/ai or /media/space/ai (see /etc/comfyui/extra_model_paths.yaml). User data (output, input, custom_nodes, temp) is under $XDG_DATA_HOME/comfyui (default ~/.local/share/comfyui). comfyui Qwen-Image needs a diffusion GGUF/safetensors, the Qwen Image VAE and Qwen2.5-VL 7B as the text encoder. Stop llama.service first if the GPU is already full. bero <bero@lindev.ch> https://github.com/Comfy-Org/ComfyUI