Skip to content

Releases: qkm2000/Scuttlebutt

1.0.3 - All Hands

Choose a tag to compare

@qkm2000 qkm2000 released this 12 Aug 07:36
b4a9f20

All Hands

Added

  • Mix in system audio — a new "System audio device" picker records a loopback input (e.g. BlackHole on macOS) alongside the mic, mixed into one track. Screen-share capture remains as an alternative.
  • Cancel a running job — transcription and summarization can be cancelled mid-flight from a button beside the progress bar; your audio and existing work are kept.
  • Configurable request timeouts per server (Transcription, Summary), where 0 = wait indefinitely, each with a reset-to-default button.
  • Per-recording "Identify speakers" toggle in the sidebar, seeded from the global default.

Fixed

  • Tag/participant inputs keep focus after pressing Enter.
  • Settings reset buttons and "Detect input devices" no longer scroll the page to the top.
  • Device/model dropdowns no longer stretch the settings row.

1.0.2 - Roll Call

Choose a tag to compare

@qkm2000 qkm2000 released this 27 Jul 08:13

All about knowing who's aboard the conversation.

New

  • Speaker identification (diarization). Flip on Identify speakers (Settings → Transcription, off by default). With a diarizing endpoint, your transcript splits into Speaker 1: / Speaker 2: turns. If the endpoint can't diarize, it falls back to a flat transcript and tells you why — nothing breaks.
  • Reset-to-default buttons on the Transcription and Summary endpoint URLs — one click to restore the default.

Notes

  • Diarization needs an endpoint that supports it (e.g. a self-hosted WhisperX server). Plain Whisper/vLLM keep working exactly as before.

1.0.1 - Maiden Voyage

Choose a tag to compare

@qkm2000 qkm2000 released this 24 Jul 10:15

Scuttlebutt — record, transcribe, and summarize meetings right in your Obsidian sidebar. Local-first: everything runs against your own Whisper/vLLM + LLM endpoints. No cloud, no accounts. Inspired by anarlog.

Highlights

  • Sidebar workflow: record (or import) → transcribe → auto-summarize → review Summary / Transcript / Memo → save one tidy note
  • Two configurable OpenAI-compatible endpoints (transcription + summary), each with a connection test and model dropdown
  • LLM-generated title and tags that reuse your existing vault tags first
  • In-sidebar audio player, re-transcribe, context-file picker, participant chips
  • Notes: an overview + ## sections, with "date created" / tags / participants frontmatter, a collapsible transcript & memo, and a configurable date format

Requires Obsidian 1.4.0+. Desktop only.