v0.6.0-alpha — first runnable build

First runnable build. Code-complete across all six plans; end-to-end speech round-trip is not yet verified against a real microphone. You're the first human to try it.

Install

Download the .msi below and run it. Windows 11 22H2+ (build 22621+), x64 required.

The MSI is unsigned, so SmartScreen will warn on launch — "More info" → "Run anyway".

After install:

• App: C:\Program Files\Winpepper\Winpepper.exe
• User data + downloaded models: %LOCALAPPDATA%\winpepper\
• Autostart entry under HKCU\…\Run\Winpepper

First-run

1. Click the tray icon to open the main window.
2. Navigate to Models tab → "Download Missing Models". Pulls ~1.2 GB from HuggingFace (Parakeet TDT v3 + Qwen 2.5 0.5B Q4_K_M GGUF), SHA-256 verified.
3. Hold Right Ctrl + Right Shift while speaking. Release. Cleaned text should land in the focused window.

Bug reports

Inside the app: Diagnostics tab → Copy diagnostics bundle. Zips logs + system info (never audio, never transcripts) into a file you can attach to an issue.

What's verified vs not

• ✅ Builds clean on Windows 11 24H2 + .NET 9 + WinAppSDK 1.8
• ✅ MSI installs/uninstalls clean (verified msiexec /qn round-trip)
• ✅ Winpepper.exe --selftest returns WINPEPPER_SELFTEST_OK
• ✅ App boots, tray icon registers, hotkey hook installs, pipeline reaches idle/ready
• ✅ ~365 cross-platform unit tests pass on Linux
• ❓ Actual mic → ASR → cleanup → keystroke injection round-trip
• ❓ All XAML pages render and bind correctly
• ❓ Post-paste learning toast prompts
• ❓ Lab rerun panels
• ❓ Model downloader UI (the downloader logic itself is tested)