v1.0.4

1.0.4

• Configurable LLM request timeout (advanced setting, default 60s, range 5–600s) for slow local LLM cold starts (#8).
• Settings UI enforces the same 5–600s range as runtime validation.
• Test coverage for custom timeout behavior.

v1.0.3

1.0.3

• Phase 5 panel UI: notebook picker, inline notebook creation, title/to-do/due controls, and Joplin-themed styling.
• Resilience: mandatory raw-transcript fallback when LLM polish fails (zero data loss).
• Debug logging to dictate.log in the plugin data directory (optional setting).
• Panel state fixes: controls stay locked through stop/cancel, file transcribe, polish, and save.
• Vitest coverage for note creation, notebook sentinels, and polish-failure fallback.

v1.0.2

1.0.2

• Fix: store icon returned 404 because icons were only under src/icons/. Added root-level icons/ (icon-16/32/48/128) so the Joplin registry resolves them.
• Bump version to 1.0.2 across manifest.json and package.json.

v1.0.1

Security & reliability hardening

• Local-by-default polishing: the default LLM server URL is now http://localhost:1234. Removes a hardcoded LAN address from the shipped defaults and keeps transcript polishing on your machine unless you deliberately point it elsewhere.
• Polish request timeout: the optional LLM polish step now has a 60-second timeout, so a slow or unreachable endpoint no longer hangs the dictation flow — you get a clear timeout error instead.
• Privacy documentation: the README now spells out that transcription is always local, and that enabling polishing sends transcript text to your configured LLM URL (keep it local/HTTPS to avoid cleartext egress).

Notes

No changes to the recording/transcription behavior from v1.0.0. Recommended upgrade for anyone using the optional LLM polishing feature.

Full changelog: v1.0.0...v1.0.1

v1.0.0

First public release of Joplin Dictate — record your voice, transcribe it locally with whisper.cpp, and turn it into a Joplin note or to-do. Audio never leaves your machine.

Highlights

• One-click dictation panel with Pause / Resume and Cancel.
• Local transcription via whisper.cpp (whisper-cli).
• Optional transcript polishing via any OpenAI-compatible LLM endpoint.
• Transcribe an existing WAV file instead of recording.

Requirements

• Linux with PipeWire (pw-record) for audio capture.
• whisper.cpp built locally with a ggml model.
• Joplin 3.6+.

Install

• Store: Tools → Options → Plugins → search Dictate (available once the Joplin bot indexes it).
• npm: joplin-plugin-dictate
• Manual: download the attached .jpl and install via Tools → Options → Plugins → Install from file.

See the README for full setup and configuration.