vaultlab-meetings

Smart Windows meeting recorder with per-app audio capture, voice-activity detection, and local-or-cloud transcription. Originally built standalone; now also serves as the meeting-context layer for vaultlab.

Install name: vaultlab-meetings (PyPI) Import name: meeting_recorder (Python)

pip install vaultlab-meetings

import meeting_recorder

The split is intentional — meeting-recorder was already taken on PyPI, so the package ships under the vaultlab-meetings name while keeping the original import name for stability across existing scripts.

What it does

• Captures per-app audio on Windows (Zoom, Teams, browser-based meetings, anything that goes through Windows audio)
• Voice-activity detection trims silence and skips empty stretches
• Local transcription via faster-whisper or cloud via OpenAI
• Optional speaker diarization via pyannote.audio
• System-tray app with hotkey controls; saves recordings + transcripts to a configurable folder
• Optional Outlook calendar integration: auto-tags recordings with the meeting that was on your calendar

Install

Base install (recording only, no transcription backend):

pip install vaultlab-meetings

With local transcription (faster-whisper + torch):

pip install "vaultlab-meetings[local]"

With cloud transcription (OpenAI Whisper API):

pip install "vaultlab-meetings[cloud]"

With Outlook calendar tagging (Windows + Outlook Classic):

pip install "vaultlab-meetings[outlook]"

Combine extras as needed: pip install "vaultlab-meetings[local,outlook,gdrive]".

Quick start

meeting-recorder

Drops a tray icon. Hotkeys (configurable in config.toml):

• Start/stop recording
• Mark a moment (saves a timestamped flag in the transcript)
• Open the recordings folder

Recordings + transcripts land at the path set in config.toml (default: ~/MeetingRecordings/).

Use from vaultlab

When vaultlab[meetings] is installed (which pulls vaultlab-meetings), vaultlab's vaultlab.context.meetings module wraps this package and ingests transcripts into your KB:

from vaultlab.context.meetings import is_available, start_recording

if is_available():
    start_recording(project="codex-pdac")

Transcripts land at <kb>/Sources/Meetings/<YYYY-MM-DD>-<slug>.md with frontmatter (date, attendees, project, recording_path, duration, transcription_model) so they're searchable from any vaultlab slash command.

Platform

Currently Windows-only. Per-app audio capture relies on PyAudioWPatch, pycaw, and mss, which are Windows-specific. macOS/Linux support is planned via a different recording backend; PRs welcome.

License

MIT. See LICENSE.