riponcm/jharu

Jharu is a free, open source desktop app that helps developers and machine learning engineers reclaim disk space that their tools leave behind. It runs on macOS and Windows and is built with Tauri. Instead of just showing folder sizes like a generic disk analyzer, Jharu knows where development tools and machine learning frameworks store their data, classifies what it finds, and explains each item in plain language with a safety rating before removing anything. The app scans over 25 known cache locations across tools like Hugging Face, Ollama, PyTorch, pip, npm, conda, Cargo, Docker, and Xcode. One standout feature is its handling of machine learning models. Jharu can tell the difference between a model you have actually loaded and one you downloaded once and never opened again, so you can safely delete the unused ones. It also preserves shared Ollama layers so removing one model never breaks another. Beyond caches, Jharu offers a deep clean view that sizes every folder on your disk and classifies them as system, applications, your files, app data, or developer data. System locations are protected and cannot be cleaned from the app. It also finds leftover files from uninstalled applications, old installers, logs, crash reports, and duplicate Python packages spread across multiple virtual environments. Additional features include browser cache cleaning for Chrome, Edge, Firefox, and others, a desktop clutter manager, drive health monitoring with SMART status and a fill forecast, and an uninstaller that removes apps along with their leftover support files. Updates are delivered in app with a signed, verified, one-click process. Jharu is available as a universal macOS download and a 64-bit Windows installer, with no telemetry or subscription. The full README is longer than what was shown.