SafeRag SE Support

1. Install from the Mac App Store

Search for "SafeRag SE" on the Mac App Store, or follow the link from our product page. Click Get/Install to download and install the app.

2. Download Your First AI Model

On first launch, SafeRag SE will guide you through downloading an AI model. You can also use Apple Foundation Models on macOS 26+ without any downloads.

• Go to Settings and browse available models
• Choose a model that fits your hardware (smaller models for 8GB RAM, larger for 16GB+)
• Wait for the download to complete (models range from 2-20GB)
• On macOS 26+, Apple Foundation Models are available immediately with no download

3. Start Chatting

Once a model is ready, you can start a conversation right away. Upload documents to enable RAG mode for context-aware answers from your own files.

SafeRag SE won't start — what should I check?

• Verify you are running macOS 15.0 (Sequoia) or later (Apple menu → About This Mac)
• Confirm you have an Apple Silicon Mac (M1, M2, M3, or M4)
• Ensure at least 5GB free disk space
• Try deleting and reinstalling from the Mac App Store
• Check Console.app for error messages mentioning SafeRag SE

How do I upload documents for RAG?

To use RAG (Retrieval-Augmented Generation) with your own documents:

1. Open the Documents section in the sidebar
2. Click the upload button or drag and drop files
3. Supported formats include PDF, DOCX, TXT, and Markdown
4. Wait for processing (the app creates vector embeddings using sqlite-vec)
5. Enable RAG mode in chat to query your documents

AI responses are slow — how can I speed them up?

• Use smaller models: Choose compact models (3B-8B parameters) for faster responses
• Try Apple FM: On macOS 26+, Apple Foundation Models are optimized for your hardware
• Upgrade RAM: 16GB+ helps significantly with larger models
• Close other apps: Free up system memory and GPU resources
• Reduce context length: Shorter conversations process faster

Where is my data stored?

All SafeRag SE data is stored locally within the app sandbox:

• Location: ~/Library/Containers/io.corixa.SafeRagSE/
• Database: GRDB.swift (SQLite-based)
• Vector store: sqlite-vec for document embeddings
• Encryption: AES-GCM when enabled

No data ever leaves your Mac. Everything runs locally.

How do I enable encryption?

SafeRag SE uses AES-GCM encryption to protect your chat history and documents:

1. Go to Settings
2. Enable the encryption toggle
3. Set your encryption passphrase
4. Messages are encrypted at rest in the local database

When encryption is enabled, your data is protected even if someone gains access to your Mac's filesystem.

What is the difference between llama.cpp and Apple Foundation Models?

llama.cpp (Built-in AI Engine):

• Available on macOS 15.0+ (Sequoia)
• Supports a wide range of open-source models (Llama, Mistral, Gemma, etc.)
• Models must be downloaded (2-20GB each)
• Full control over model selection and parameters

Apple Foundation Models:

• Requires macOS 26+ (Tahoe)
• No model download needed — built into macOS
• Optimized by Apple for Apple Silicon hardware
• Seamless integration with the operating system

You can use both engines side by side and switch between them per conversation.

How do I connect Ollama (optional)?

SafeRag SE works without Ollama, but you can optionally connect to a running Ollama instance for additional model support:

1. Install Ollama separately from ollama.com
2. Ensure Ollama is running on your Mac
3. In SafeRag SE settings, enable the Ollama provider
4. Ollama models will appear alongside your local llama.cpp models

This is entirely optional — SafeRag SE's built-in AI engine handles everything on its own.

Model Download Failed

Symptoms: Download stops, progress bar freezes, or error message appears

Solutions:

1. Check your internet connection
2. Ensure you have enough free disk space (models can be 2-20GB)
3. Cancel and retry the download
4. Try a smaller model if you have limited storage
5. If the issue persists, restart SafeRag SE and try again

RAG Not Finding My Documents

Symptoms: RAG mode is enabled but the AI does not reference your documents

Solutions:

1. Verify documents have finished processing (look for a completion indicator)
2. Ensure RAG mode is enabled in the chat input area
3. Try asking more specific questions related to your document content
4. Re-upload the document if processing appears to have failed
5. Check that the document format is supported (PDF, DOCX, TXT, MD)