Audit and clean 15–90 GB of ComfyUI temp tensors, duplicate HuggingFace / LM Studio GGUF weights, CUDA PyTorch pip wheels, and Cursor IDE workspace bloat without deleting active models or workflows.
Pre-loaded with the default Windows 10/11 cache directories for ComfyUI, LM Studio, Ollama & Cursor AI. Paste your own paths or click Audit Paths Now to verify deletion safety:
%UserProfile%\.cache\huggingface\hub
Python AI tools, Diffusers, and Transformers cache every downloaded safetensors checkpoint and tokenizer snapshot inside .cache\huggingface\hub.
View Detailed Folder Safety Guide →%UserProfile%\.cache\lm-studio\models
LM Studio downloads multi-gigabyte HuggingFace GGUF model checkpoints and temporary partial downloads into your user cache folder.
View Detailed Folder Safety Guide →%LocalAppData%\pip\Cache
Python's pip installer caches large AI/ML wheels (PyTorch, CUDA, TensorFlow, NumPy) in %LocalAppData%\pip\Cache.
View Detailed Folder Safety Guide →%LocalAppData%\Spotify\Storage
Spotify caches streamed song tracks locally on your C: drive to speed up playback and reduce bandwidth usage.
View Detailed Folder Safety Guide →Local AI workflows duplicate 2.5 GB CUDA torch wheels inside %LocalAppData%\pip\Cache, leave unlinked HuggingFace snapshot blobs after interrupted downloads, and retain SQLite vector indexes for deleted git branches in Cursor.