off-grid-ai-mobile: a comprehensive offline AI suite for mobile and Mac providing on-device text, image, and vision capabilities

What it solves

Off Grid is a comprehensive offline AI suite for mobile devices and macOS, allowing users to run powerful AI models locally without needing an internet connection or sending data to the cloud. It eliminates the privacy concerns associated with cloud-based AI by ensuring all processing happens on-device.

How it works

The app leverages native hardware acceleration (NPU, Core ML) and optimized libraries like llama.cpp and whisper.cpp to run various AI models locally. It supports text generation (GGUF models), image generation (Stable Diffusion), vision AI (VLM), and speech-to-text (Whisper), all integrated into a single interface. It also allows connection to remote LLM servers (like Ollama) for those who prefer a local network setup.

Who it’s for

Privacy-conscious users, developers, and anyone who wants an AI assistant that works entirely offline, on their phone or Mac.

Highlights

• Multimodal Capabilities: Supports text, image generation, vision AI, and voice transcription.
• On-Device RAG: Includes a project knowledge base where PDFs and text documents are embedded and retrieved locally using a MiniLM model and SQLite.
• Tool Calling: Built-in tools for web search, calculator, and knowledge base search with runaway prevention.
• ** uma NPU-accelerated Image Generation**: Real-time previews and support for 20+ Stable Diffusion models.
• Broad Model Support: Compatible with Qwen, Llama, Gemma, and Phi models, as well as custom .gguf files.