ZetaNotes

ZetaNotes runs AI models entirely on your Mac. No cloud connection required. Your notes and AI processing never leave your device.

CORE CAPABILITIES

• AI Autocomplete - Predictive text completion as you type using on-device language models
• Text Summarization - Generate concise summaries from longer note content
• Language Translation - Translate to 12+ languages including English, Spanish, French, German, Chinese, Japanese, Korean, Arabic, Italian, Portuguese, Russian, and Dutch
• Grammar Correction - Fix spelling and grammatical errors in selected text
• Text Reformulation - Rewrite sentences with alternative phrasing
• Semantic Search - Find notes by meaning using local vector embeddings, not just keyword matching
• Auto Title Generation - Automatically generate descriptive titles from note content

AI CHAT ASSISTANT (PRO)

Conversational AI interface for open-ended queries. Ask questions, brainstorm ideas, or get writing assistance. Full conversation history persisted locally.

CUSTOM TOOLS (PRO)

Create application-specific AI tools with custom prompts. Configure individual temperature, token limits, and stop sequences per tool. Build workflows tailored to your specific use cases.

TECHNICAL SPECIFICATIONS

• 8 Open-Source AI Models Available - Llama 3.2 (1B/3B), Phi-3.5 Mini, Gemma 2 2B, Qwen 2.5 (1.5B/3B), SmolLM2 1.7B, Phi-2 2.7B, TinyLlama 1.1B
• Metal GPU Acceleration - Hardware-accelerated inference on Apple Silicon
• Model Format - GGUF quantized models optimized for M-series chips
• Model Size Range - 700MB to 2.3GB per model (downloaded on demand)
• Inference Engine - llama.cpp with Apple Metal backend
• Vector Search - Local embedding generation with cosine similarity ranking

PRIVACY & SECURITY

• Zero Network Connectivity - All AI processing runs locally on device
• No Telemetry - No analytics, crash reporting, or usage tracking
• No Cloud Sync - Notes stored exclusively in local SQLite database
• No Account Required - No registration, login, or personal information collected
• Keychain Integration - License keys stored in macOS Keychain with local encryption

SYSTEM ARCHITECTURE

• Native SwiftUI Interface - Optimized for macOS with standard system controls
• Core Data Persistence - Note storage with SQLite backend and automatic indexing
• Memory Management - Optimized context caching reduces RAM usage by 90%
• Batch Processing - Efficient token generation for summarization tasks
• Async Inference - Non-blocking AI operations preserve UI responsiveness

STORAGE LOCATIONS

• Notes Database: ~/Library/Application Support/ZetaNotes/notes.sqlite
• AI Models: ~/Library/Application Support/ZetaNotes/Models/
• Settings: ~/Library/Preferences/com.zetanotes.ZetaNotes.plist

FREE TIER FEATURES

• Llama 3.2 1B model (700MB)
• AI Autocomplete
• Text patch and grammar correction
• Text reformulation (rephrasing)
• Summarization
• Translation (12+ languages)
• Semantic search
• Unlimited notes
• Unlimited storage

PRO TIER FEATURES

Available via in-app purchase: Monthly ($1), Annual ($6), or Lifetime ($20)

• All Free features
• AI Chat Assistant with conversation history
• 8 AI models (download any combination)
• Custom AI tool creation
• Full prompt editing for all built-in tools
• Parameter tuning (temperature, tokens, stop sequences)
• Priority support

REQUIREMENTS

• macOS 13.0 (Ventura) or later
• Apple Silicon (M1/M2/M3/M4) recommended, Intel supported
• 4 GB RAM minimum (8 GB+ recommended for 3B models)
• 2 GB disk space (includes app + at least one AI model)

INSTALLATION

Standard macOS application. Code signed with Apple Developer ID and notarized by Apple. No security warnings or workarounds required.

UPDATES

Automatic updates via Sparkle framework. Updates download and install in background with user confirmation.

TECHNOLOGY

Built with llama.cpp inference engine and open-source language models from Meta (Llama), Microsoft (Phi), Google (Gemma), Alibaba (Qwen), and Hugging Face (TinyLlama, SmolLM2).