DeviceHealth：给系统进程打分并识别臃肿软件的开源桌面工具

Tells you which of the 200 things running on your machine you can safely get rid of.

The machine is slow, Activity Monitor lists two hundred processes, and you recognise maybe thirty of them. The rest have names that could equally be a driver you need or something an installer left behind three years ago. So nothing gets removed, because guessing wrong breaks the machine.

DeviceHealth scans, scores the system out of 100, and sorts what it found into bloatware, zombie processes, autostart clutter and security risks. Anything you do not recognise, you can ask a local Ollama model to explain in plain words before you decide.

Autostart detection covers macOS LaunchAgents and Linux systemd units. On Windows it currently finds nothing, since neither the registry Run keys nor the Startup folder are read yet (see ROADMAP.md).

Snapshots are kept locally, so "it has been getting slower" becomes a curve instead of a feeling. Nothing is uploaded, and there is no account.

Not for you if you already know what is on your machine. htop, Activity Monitor and Autoruns give you the same data faster; the scoring and the explanations only earn their keep when the process list is unfamiliar.

How it runs: DeviceHealth is a native desktop app, not a server or browser tool. It opens as its own window and has no tray icon or background service; it only scans your system while you actively trigger a run.

💾 Download: macOS (DMG) · Windows (Installer) · Linux (AppImage): always the latest release, not code-signed/notarized (Gatekeeper/SmartScreen will warn on first run). Or build from source, see Getting Started below.

In practice: you trigger a scan, get a 0 to 100 health score with categorized findings (bloatware, zombie processes, autostart clutter, security risks), and can ask Ollama to explain any process in plain language before deciding what to fix. Every snapshot is stored locally so you can track the trend over time.

🌱 New here? → Step-by-step guide for beginners

Features

Feature	Description
Health Score	Numerical 0 to 100 rating with grade (Excellent → Critical)
Process Analysis	Says for every process where it comes from: part of the operating system, belongs to app X (including its helpers), or installed outside both. Of about 750 processes on a typical Mac, around 40 are left that are not system. Also flags bloatware, telemetry, zombies, CPU spikes and high memory use
Findings	Categorized issues (Critical / High / Medium / Low / Info) with fix recommendations
Hardware Monitor	CPU, RAM, disk usage, temperatures (where the system reports them; implausible sensor values are dropped), network I/O
Autostart Detection	Lists LaunchAgents and LaunchDaemons (macOS) and enabled systemd units (Linux) in their own view. Listing only: removing an entry is done in System Settings or by deleting the file shown
Security Findings	Processes with unusual names, known cryptominer names, unknown processes with high CPU. No virus scanner, no driver or update checks
AI Explanations	Ollama explains every process in plain language: results cached locally
History	Score trend and system snapshots over time with charts
Recommendations	Processes worth quitting (telemetry, apps with high CPU or memory), each with a "Quit" button behind a confirmation. System processes are never offered

Requirements

Rust 1.77+
Node.js 20+
Tauri CLI v2: cargo install tauri-cli
Ollama (optional, for AI explanations)
macOS / Windows / Linux

Quick Start

git clone https://github.com/9t29zhmwdh-coder/DeviceHealth
cd DeviceHealth

# Optional: pull AI model for process explanations
ollama pull qwen3.5:4b

cd frontend && npm install && cd ..
cargo tauri dev

Uninstall / Cleanup

macOS: delete the app bundle, then remove ~/Library/Application Support/com.raystudio.devicehealth/ (snapshot history and AI cache)
Linux: delete the app binary, then remove ~/.local/share/devicehealth/
Windows: uninstall via Settings → Apps, then remove %LOCALAPPDATA%\RayStudio\DeviceHealth\

No registry entries or background services are left behind.

Privacy

DeviceHealth processes all system data locally on your machine. No data is sent to any external server. Ollama runs entirely offline; AI explanations never leave your device.

Architecture

DeviceHealth/
├── crates/dh-core/      # Rust: analyzer, hardware, process detection, DB
├── crates/dh-cli/       # CLI binary
├── src-tauri/           # Tauri v2 backend + IPC commands
└── frontend/            # React + TypeScript + Tailwind + Recharts

Analysis Pipeline

sysinfo ──► spawn_blocking ──► ProcessAnalyzer + HardwareAnalyzer
                                       │
                               SecurityAnalyzer + ServiceAnalyzer
                                       │
                               HealthScore (0 to 100) + Findings
                                       │
                               SQLite (snapshots, AI cache)

Author: Rafael Yilmaz · Status: Active · · License: MIT