paulktothicloudcom/

mac-egpu-llm-stack

A fully automated setup for running a four-component local LLM inference stack on a Mac Mini M4 Pro with an AMD RX 7900 XTX eGPU via TinyGPU. All inference runs locally — no cloud API keys, no telemetry, no code leaves the machine.

	Hardware Target




	

Component Specification

Host Apple Mac Mini M4 Pro

Unified Memory 64 GB (hard minimum — see Memory Requirements)

eGPU SAPPHIRE NITRO+ RX 7900 XTX VAPOR-X 24 GB… See the full description on the dataset page: https://huggingface.co/datasets/paulktothicloudcom/mac-egpu-llm-stack.