🔒 Security & Privacy First

SemanticSearch - Apple Notes AI Search

A macOS menu bar application that provides semantic search for Apple Notes using local embeddings and OpenAI's text-embedding-3-small model.

Download

👉 Download for macOS (DMG)

First run (not notarized): after dragging the app to Applications, right‑click SemanticSearch.app → Open → Open once.

Demo

Features

• SwiftUI Menu Bar App: Clean, modern interface accessible from the menu bar
• Semantic Search: AI-powered search that understands meaning, not just keywords
• Local Data Storage: All embeddings and search data stored locally in Application Support
• Auto-sync: Automatically scans and indexes notes on first run and subsequent launches
• Background Processing: Runs background sync to keep search index up-to-date
• Run at Login: Optional LaunchAgent for automatic startup

Architecture

Python Backend (app/)

• AppleScript/JXA Integration: Exports notes with time-range filters and metadata
• Incremental Sync: Window batching, parallel embeddings, smart caching by modification date
• Local Vector Store: Efficient storage and retrieval of embeddings
• Streaming Search: Real-time search results with confidence scoring
• Interest Line Extraction: Finds the most relevant lines within notes

SwiftUI Frontend (SemanticSearchMenuBar/)

• Menu Bar Integration: Native macOS menu bar app with popover interface
• Bridge Communication: NDJSON-based communication with Python backend
• Dynamic Sizing: Content-driven popover sizing with gradient background
• Search Interface: Real-time search with streaming results

Quick Start

Development Setup

# Create virtual environment
python3.11 -m venv venv_notes
source venv_notes/bin/activate
pip install -U pip
pip install -r requirements.txt

API Key Configuration

# .env at repo root
OPENAI_API_KEY=your_openai_api_key

Run the App

• Open SemanticSearchMenuBar/SemanticSearch.xcodeproj in Xcode
• Build and run the project
• The app will appear in your menu bar

CLI Usage (Development)

# Interactive CLI
python3 app/cli.py

# Commands available:
# sync          - Scan and index notes
# sync all      - Full re-index
# watch         - Auto-sync on file changes
# list          - List indexed notes
# open <id>     - Open specific note
# refresh-meta all - Refresh metadata

Manual full backfill (CLI)

python3 -u "/Users/siddharthasehgal/Documents/Work/SearchNotes/SemanticSearchMenuBar/SemanticSearch/Resources/bridge.py" sync --days 3650 --full | tail -n 1

Packaging & Distribution

Build Distributable App

# Ensure venv is up-to-date
python3.11 -m venv venv_notes
venv_notes/bin/pip install -U pip
venv_notes/bin/pip install -r requirements.txt

# OpenAI API key
export OPENAI_API_KEY=your_openai_api_key
# or create a .env at repo root with: OPENAI_API_KEY=...

/bin/bash SemanticSearchMenuBar/Scripts/package.sh | cat

Optional Notarization

export NOTARIZE=1
export NOTARY_PROFILE=your_keychain_profile
bash SemanticSearchMenuBar/Scripts/package.sh

Auto-Start on Login

bash SemanticSearchMenuBar/Scripts/install_launch_agent.sh

Using the App (End Users)

• Open the generated SemanticSearch.dmg and drag SemanticSearch.app into Applications.
• Eject the mounted SemanticSearch disk after copying.
• Launch from /Applications.
• On first run, the app will ask for Apple Events permission and begin an initial sync; ensure OPENAI_API_KEY is configured in your shell environment or placed in ~/Library/Application Support/SemanticSearch/.env.

Data Storage

• User Data: ~/Library/Application Support/SemanticSearch/
  • state.json - Last sync timestamp and metadata
  • embeddings.jsonl - Note embeddings
  • vector_store/ - Vector search index
  • .env - Optional local API key override

Permissions Required

• Apple Events: To interact with Apple Notes
• Full Disk Access: To read Notes database (if needed)

Reset Automation permission via CLI

# Reset Apple Events permissions for the app (bundle id)
tccutil reset AppleEvents okaysidd.SemanticSearch

# Then quit and relaunch the app; the next Notes action will prompt again

Notes:

• This clears all Apple Events approvals for SemanticSearch to control other apps.
• The command requires macOS 12+; if it does nothing, open Settings → Privacy & Security → Automation and remove the app manually.

Technical Details

Search Process

1. User types query in menu bar interface
2. Query is sent to Python backend via bridge
3. Backend generates embedding for query
4. Vector similarity search against indexed notes
5. Results stream back with confidence scores
6. Most relevant lines extracted from matching notes

Sync Process

1. On first launch: Full scan of all notes
2. Subsequent launches: Incremental sync since last run
3. Background sync: Continuous monitoring for changes
4. Smart caching: Only re-embed notes that have changed

Bridge Communication

• NDJSON protocol for real-time communication
• Process-based architecture for stability
• Error handling and graceful degradation

Development

Project Structure

SearchNotes/
├── app/                          # Python backend
│   ├── notes.py                  # Core sync/search logic
│   ├── vector_store.py           # Vector storage
│   ├── cli.py                    # Command-line interface
│   └── watcher.py                # File system monitoring
├── SemanticSearchMenuBar/        # SwiftUI app
│   ├── SemanticSearch.xcodeproj  # Xcode project
│   ├── Bridge/                   # Swift-Python bridge
│   ├── Views/                    # SwiftUI views
│   └── Scripts/                  # Build and packaging scripts
└── requirements.txt              # Python dependencies

Building from Source

1. Clone repository
2. Set up Python environment (see Quick Start)
3. Open Xcode project
4. Build and run

License

This project is for personal use. Please ensure you have appropriate OpenAI API access and comply with Apple's developer guidelines for distribution.