Resonance

Resonance is the clean-slate v2 of the Obsidian recording plugin. The product is now built around a local-first pipeline, settings-first setup, and manifest-backed session storage.

Fastest Setup

If this machine already has FFmpeg, a built whisper.cpp, and at least one local ggml model:

1. Open the plugin settings page.
2. Work through Step 1, Step 2, and Step 3 from top to bottom.
3. Use the Detect buttons to fill paths automatically when possible.
4. Pick a microphone and run Quick test.
5. Leave Ollama as the provider unless you intentionally want a cloud provider.

The settings page is now the full setup flow. Nothing important is hidden behind a separate setup modal.

What Changed

• src/ is the only active application tree for v2.
• The primary entrypoint is now the plugin settings surface instead of a simple start/stop flow.
• Recording state is driven by an explicit session controller with ordered live transcription.
• Sessions are stored as structured manifests with audio/, transcript/, summary/, and diagnostics.log.
• The session library is manifest-backed. It no longer scans raw media files blindly.

Product Direction

• Desktop-only Obsidian plugin
• Local-first path: FFmpeg + whisper.cpp + Ollama
• Cloud summary providers remain available as secondary adapters
• Native Obsidian settings are now the primary surface, with horizontal tabs for diagnostics, library, and setup

Current Information Architecture

• Diagnostics
  • Health status and blocking issues
  • Quick test
  • Startup behavior
  • Setup guidance when the local path is incomplete
• Setup & Settings
  • Horizontal tabs for Diagnostics, Library, Capture, Transcription, Summary, and Output
  • Capture tab for FFmpeg, devices, and quick test
  • Transcription tab for whisper.cpp, CLI, and model
  • Summary tab for Ollama or cloud providers
  • Output tab for vault, retention, and startup behavior
• Session Library
  • Filters for done and failed
  • Artifact availability
  • Transcript, diagnostics, audio, and summary actions

Repository Layout

src/            Active Resonance v2 source
tests/          Pure unit tests for v2
legacy/         Archived v1 runtime files kept for reference
dist/           Build output

Local Development

Requirements:

• Obsidian desktop
• Node.js
• FFmpeg
• whisper.cpp with a local model
• Ollama if you want the full tier-1 local path

Commands:

npm run typecheck
npm test
npm run build

Install In Obsidian

1. Build the plugin with npm run build.
2. Copy dist/main.js, dist/manifest.json, and dist/styles.css into your vault plugin folder.
3. Enable Resonance in Obsidian community plugins.
4. Open the plugin settings page.
5. Use the Capture tab for FFmpeg and devices, then Transcription and Summary.
6. When setup is done, use the Diagnostics tab for health and the Library tab for saved session artifacts.

Minimum Local Requirements

To record and summarize locally, Resonance needs:

• FFmpeg
• whisper.cpp CLI
• a readable local ggml model file
• Ollama running locally for summaries

If auto-detect fails, the manual fallback is:

1. Point Step 3 in plugin settings at your whisper.cpp repo.
2. Detect or set the whisper.cpp CLI path.
3. Detect or set the ggml model path.
4. Return to Step 2 and run the quick test.

Runtime Model

Each session persists:

• session.json
• audio/recording.mp3
• audio/segments/
• transcript/live-transcript.txt
• summary/summary.md
• diagnostics.log

The live transcription queue commits segments strictly in order and the stop flow waits for the queue to flush before summary generation starts.

Notes On Legacy Code

The previous implementation is archived under legacy/. It is not part of the active v2 runtime, build entry, or test surface.