Turn conversations into working memory.
Record, transcribe, and summarize meetings entirely on your device.
Sivlo is a privacy-first AI meeting assistant for macOS. It captures microphone and system audio, transcribes speech in real time with on-device models, and generates summaries, actions, and decisions — with no account required and no cloud by default.
v0.1.0 is a public macOS beta (Apple Silicon + Intel, macOS 13+). Windows and Linux builds are not part of this release.
| Data | Default behavior |
|---|---|
| Audio & recordings | Stored locally only |
| Transcripts & summaries | Processed on-device (Whisper / Parakeet) |
| Meeting content | Never sent to analytics |
| Cloud AI (Claude, Groq, etc.) | Optional — only when you configure a provider |
Get the latest notarized DMG from GitHub Releases.
Or visit the landing page for an overview.
- Real-time transcription — Whisper and Parakeet, GPU-accelerated (Metal on macOS)
- Professional audio mixing — Mic + system audio with RMS ducking
- Meeting intelligence — Summaries, actions, decisions, notes, full-text search
- Flexible AI — Local Ollama by default; optional cloud providers (Claude, Groq, OpenRouter, custom endpoints)
- Secure credential storage — API keys stored in macOS Keychain, not plaintext SQLite
- Auto-updates — Signed, notarized builds with in-app update checks
- Opt-in analytics — Off by default; no meeting content in telemetry
- macOS 13+ (Ventura or later)
- Apple Silicon (M1+) or Intel Mac
- Microphone permission; screen recording permission for system audio capture
- ~2 GB disk space for transcription models (varies by model size)
cd frontend
pnpm install
./clean_run.sh # macOS dev build
./clean_build.sh # Production buildSee docs/BUILDING.md for detailed platform instructions.
cd frontend
bun test # Frontend unit tests
pnpm lint
cd src-tauri && cargo test # Rust unit testsSivlo is a self-contained Tauri 2 desktop app:
- Frontend — Next.js + React (
frontend/src) - Backend — Rust audio pipeline, transcription, SQLite storage (
frontend/src-tauri)
See docs/architecture.md for details.
The
backend/directory contains an archived Python/FastAPI implementation from upstream Meetily. It is not used by Sivlo and must not be deployed.
Contributions welcome! See CONTRIBUTING.md. Please open issues for bugs and feature requests.
Report vulnerabilities per SECURITY.md.
MIT — see LICENSE.md.
Sivlo is forked from Meetily by Zackriya Solutions. Additional credits in the upstream project for Whisper.cpp, Screenpipe, and Parakeet integrations.