v0.1.4

Wisteria v0.1.4

Fixes:
- Hotkey recorder: reliable capture (Ctrl now records), Fn-lock-safe keys
  (Insert/PrintScreen/Numpad/arrows), commit-on-first-release, Esc to cancel,
  and a warning when a modifier-only binding would consume global shortcuts.
- Dock: removed the ghost shape that appeared above the pill on hover.
- History: strip UTF-8 BOM before JSON parsing, so history/config no longer
  vanish when written by Windows tooling.
- GUI: added a portfolio link (rjav.in) to the Settings footer.

Wisteria v0.1.3 — Windows

Windows installer for Wisteria v0.1.3.

Faster cloud formatter (BYOK)

• Reuses one warm HTTPS connection across dictations and pre-opens it at startup, so a cloud model skips the TLS handshake.
• On OpenRouter, routes to the highest-throughput backend for your model.
• Tip: pick a small, fast, non-reasoning model (e.g. gpt-4o-mini, or Groq's llama-3.1-8b-instant) for the snappiest cleanup.

Memory efficiency

• When a cloud provider is selected, local Ollama is never loaded.
• Switching away from a local Ollama model (to a cloud provider, a different model, or turning the LLM off) now unloads it from Ollama to free RAM/VRAM.

Download Wisteria-Setup-Windows-x64.exe below and run it. Updating over an existing install keeps your settings, history, and models (stored separately in WisteriaData). Linux .AppImage/.deb are attached by CI.

Wisteria v0.1.2 — Windows

Windows installer (locally built) for Wisteria v0.1.2.

Fixes

• Dock now anchors just above the taskbar on any monitor/DPI (no more random placement on other machines).
• Main window fits small / high-DPI laptop screens instead of overflowing off-screen.
• "Report an Issue" now actually reaches the reporting endpoint in distributed builds.

New

• BYOK cloud formatter: link OpenAI-compatible API services (OpenRouter, OpenAI, Groq, …) from Settings and use a hosted model for cleanup / Ask AI. Local Ollama stays the default.

Download Wisteria-Setup-Windows-x64.exe below and run it. On first launch it downloads the speech-to-text model automatically.

Wisteria v0.1.0 — Windows

Wisteria — open-source, local-first voice dictation. Hold a key, speak, release; clean text appears at your cursor. 100% offline by default.

Install (Windows 10/11, 64-bit)

1. Download Wisteria-Setup-Windows-x64.exe below.
2. Run it. It installs per-user (no admin needed) and adds the WebView2 runtime if missing.
3. During install it will offer to set up Ollama for the optional local AI formatter — you can decline; Wisteria still works transcription-only.
4. On first launch, Wisteria downloads its speech-to-text model (Parakeet) automatically. You can pick and download a formatter model later from inside the app.

SmartScreen note: this installer isn't code-signed yet, so Windows may show a "Windows protected your PC" prompt. Click More info → Run anyway to proceed. Code signing is on the roadmap.

No Rust, build tools, or extra commands required.