whyslow

Diagnose why a Windows machine feels slow, in the terminal.

A Windows-first diagnostic CLI that identifies likely bottlenecks, explains the evidence, and suggests what to do next.

Install

1. GitHub Releases (recommended for most users)
   • Download the latest whyslow-windows-x86_64.exe from GitHub Releases.
2. crates.io

   cargo install whyslow

3. Build from source

   git clone https://github.com/MuhammedZohaib/whyslow.git
   cd whyslow
   cargo build --release
   .\target\release\whyslow.exe

Quick Start

whyslow
whyslow --json
whyslow --watch 5

Example Output

whyslow report
=============
Generated: 2026-01-15 10:30:00 UTC
Samples: 40 over 20s
Host: workstation-01 | Windows 11 Pro

1) System Summary
- CPU avg/peak: 89.2% / 98.7%
- Memory avg/peak used: 86.1% / 92.4%
- Avg process count: 254.0
- Disk read/write: 45.0 MB/s / 12.0 MB/s
- Disk busy avg/peak: 87.4% / 98.2%
- Disk latency avg/peak: 24.5 ms / 80.0 ms
- Network down/up avg: 2.5 MB/s / 0.8 MB/s

2) Likely Bottlenecks (Ranked)
1. cpu_saturation (confidence 91%, duration 18s)
   Explanation: CPU stayed high during most of the sampling window.

4) Suggested Actions
- Pause or close the top CPU-heavy process first
- Check whether antivirus scans, indexing, or build tasks are running

Why Not Task Manager?

Task Manager is excellent for raw counters. whyslow is for diagnosis.

• It samples over time instead of a single moment.
• It ranks likely bottlenecks with confidence scores.
• It maps process evidence to plain-language explanations.
• It outputs stable JSON for automation and incident notes.

Features

• Windows-first system sampling (CPU, memory, disk, network, process families)
• Deterministic diagnosis rules (no cloud dependency)
• Ranked bottlenecks with evidence and suggested actions
• Terminal watch mode for repeated checks
• Export to JSON or Markdown report

Limitations

• Windows-first support today (Linux/macOS are not a target yet)
• Some disk counters may require elevated privileges or may be unavailable on specific hosts
• Thermal metrics are not currently included
• Confidence scores are heuristic, not ground truth

Roadmap

• Add network bottleneck diagnosis
• Add per-process disk I/O table in default output
• Add snapshot export mode for scheduled runs
• Improve confidence calibration and rule explainability
• Add WSL-specific diagnosis rules

CLI Help

whyslow --help
whyslow --version

Contributing

See CONTRIBUTING.md for development setup, coding standards, and how to add diagnosis rules.

Security and Privacy

See SECURITY.md. whyslow performs local-only analysis and does not upload telemetry by default.

License

Licensed under MIT OR Apache-2.0. See LICENSE.