https://github.com/fru-dev3/memosa
https://github.com/fru-dev3/memosa
Last synced: 1 day ago
JSON representation
- Host: GitHub
- URL: https://github.com/fru-dev3/memosa
- Owner: fru-dev3
- License: gpl-3.0
- Created: 2026-03-07T21:21:03.000Z (4 months ago)
- Default Branch: main
- Last Pushed: 2026-06-15T00:13:04.000Z (7 days ago)
- Last Synced: 2026-06-15T01:18:56.771Z (7 days ago)
- Language: TypeScript
- Size: 79.9 MB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- Changelog: CHANGELOG.md
- Contributing: CONTRIBUTING.md
- License: LICENSE
- Code of conduct: CODE_OF_CONDUCT.md
- Security: SECURITY.md
Awesome Lists containing this project
README
# Memosa
**Local-first meeting recorder for macOS.** Records audio, transcribes on-device with Whisper, and organises everything into a searchable library — nothing leaves your machine unless you choose.
**[⤓ Download the latest `.dmg`](https://github.com/fru-dev3/memosa/releases/latest/download/Memosa-mac-arm64.dmg)** · [Mac App Store](https://apps.apple.com/us/app/memosa-meeting-memory/id6760445178?mt=12) · [all releases](https://github.com/fru-dev3/memosa/releases)
## What it does
- Records meetings and calls on your Mac
- Transcribes audio locally using Whisper (on-device after initial model download)
- Organises recordings by date, tags, people, and custom folders
- **AI summaries, action items & decisions** — local-first via Ollama (private by
default) or your own Anthropic/OpenAI key (opt-in)
- **Chat with your meetings** — ask questions across your transcript library
(local retrieval + your chosen engine)
- **Calendar auto-record** — connect Google Calendar (read-only) to record
meetings automatically, with a 2-minute heads-up
- **AI speaker labels** — speaker-attributed transcript on demand
- **Sync** a meeting to an Obsidian vault (local) or a Notion database (your token)
- Exports transcripts and notes to any folder or tool you already use
> **Privacy posture:** the default engine (heuristic, or local Ollama) keeps
> everything on your Mac. Cloud options (BYOK summaries, Notion) only send data
> after you explicitly opt in, and they say so in the UI. Calendar uses read-only
> scope and stores its token in the macOS Keychain.
## Tech stack
- **Frontend** — React + TypeScript + Vite
- **Backend** — Rust via Tauri 2
- **Transcription** — whisper-rs (on-device)
- **Storage** — SQLite (FTS5 full-text search) + local file system
- **Audio** — CoreAudio via Objective-C (no subprocess, MAS sandbox compliant)
## Requirements
- macOS 13.0 or later
- Rust (latest stable)
- Node.js 18+
## Development
```bash
# Install dependencies
npm install
# Start dev server (Vite only — see note below)
npx vite
# Build Rust binary
cd src-tauri && cargo build --no-default-features --features whisper-rs
# For microphone access in dev, wrap in a signed .app bundle
# See: src-tauri/src/audio/permissions.rs
```
> **Note on mic permissions in dev:** macOS requires a signed `.app` bundle to show
> the microphone permission dialog. Running the binary directly will silently deny access.
> Build the bundle and launch via `open` to trigger the TCC prompt.
## Release build
```bash
npm run tauri build
# Output: target/release/bundle/macos/Memosa.app
```
Signed, notarized `.dmg` releases are produced in CI by pushing a `v*` tag — see
[docs/RELEASING.md](docs/RELEASING.md).
## Project structure
```
src/ React frontend
src-tauri/src/ Rust backend
audio/ Recording, mic permissions, CoreAudio helpers
transcription/ Whisper integration, job queue, insight orchestration
insights/ AI insight engines (Ollama / BYOK Anthropic·OpenAI)
calendar/ Google Calendar OAuth (PKCE), auto-record scheduler
chat/ "Chat with your meetings" — FTS retrieval + LLM
sync/ Obsidian (local) + Notion sync
storage/ SQLite DB, file system, settings, cleanup
export/ Export providers (local bundle)
macos.rs ObjC bridge (audio, URL open, Finder reveal)
macos_helpers.m Objective-C implementations
src-tauri/entitlements.plist App sandbox entitlements (incl. loopback OAuth)
src-tauri/Info.plist macOS metadata and privacy strings
```
## Configuration
All optional; configured in **Settings**. Secrets (OAuth/Notion tokens, BYOK
keys) live in the macOS Keychain, never on disk.
- **AI Insights** (Settings → AI Insights): choose *Built-in* (offline),
*Ollama* (local — run `ollama pull llama3.1`), or *Cloud (BYOK)* with your own
Anthropic/OpenAI key. "Regenerate all" re-runs insights across your library.
- **Calendar** (Settings → Calendar): paste a Google Cloud **OAuth desktop
client ID** (see `docs/google-calendar-setup.md`), connect, and toggle
auto-record. Read-only scope; loopback redirect `http://localhost:8899/callback`.
- **Integrations** (Settings → Integrations): pick an Obsidian vault folder
and/or add a Notion integration token + database ID. Sync per-meeting from a
recording's sidebar.
## Privacy
All audio and transcripts are stored locally at the path you configure in Settings.
Whisper models are downloaded from the internet on first use, then all processing runs on-device.
The app is sandboxed and targets the Mac App Store.
## License
Memosa is free software, released under the **GNU General Public License v3.0**.
You may use, study, share, and modify it; derivative works must also be released
under the GPL-3.0. See [LICENSE](LICENSE) for the full text.
Copyright © 2026 Fru Louis · fru.dev
The name "Memosa", its logo, and the Mac App Store listing are not covered by the
code license.