macOS Offline AI Server Setup

I want to turn my Mac into a completely self-contained AI workstation that never sends data to the cloud. The goal is to run several kinds of models locally—large-language chat models, image generation models, and speech-to-text / voice models—and interact with them through a single, lightweight desktop application (no browser interface).

Here’s the workflow I have in mind:

• Your solution installs and configures the best-fit open-source stacks for each model type (for example, llama.cpp or similar for text, Stable Diffusion for images, Whisper or comparable for speech).
• The models stay on-device; internet access is blocked once everything is downloaded.
• A native or Electron-based desktop app lets me select a model, drop in a prompt or audio/image input, and view the output. No command-line juggling once it’s in place.
• You document every step so I can update models or add new ones later without breaking the setup.

Deliverables
1. Fully working local server and desktop front-end packaged for macOS (Apple Silicon).
2. Scripts or containers that reproduce the environment from scratch.
3. Clear “how-to” guide covering installation, usage, and future model updates.

If this matches your experience in macOS, Python, PyTorch, Core ML, or related tooling, I’m ready to get started right away.