Summary

This project provides comprehensive, open-source guides for building personal AI computers, addressing the need for self-hosted, cloud-independent AI model execution. It targets individuals, research teams, and businesses seeking full control over their AI hardware and data. The benefit is permanent ownership of AI compute capabilities, free from external service disruptions or restrictions, enabling users to "own their intelligence."

How It Works

The project offers three distinct hardware configurations (2x, 4x, 8x GPU) with detailed build instructions, including bills of materials, 3D-printable/CNC housing files, wiring diagrams, and assembly photos. This approach empowers users to construct powerful AI machines tailored to their budget and performance needs, running open-source models locally using standard software stacks like Ollama or vLLM. The core advantage is achieving true AI hardware autonomy and data sovereignty.

Quick Start & Requirements

• Primary Install/Run: Select a configuration (e.g., "2x config", "4x config", "8x config"), source parts from its Bill of Materials, fabricate housing (3D print/CNC), assemble hardware, and set up software (OS, NVIDIA drivers, BIOS tuning, model serving via /software).
• Non-Default Prerequisites: High-end GPUs (e.g., NVIDIA RTX 5090, RTX PRO 6000 Blackwell, RTX 4090), substantial VRAM (64GB to 384GB+), compatible server/workstation motherboards and CPUs (Intel Xeon W5, AMD EPYC), and tools for housing fabrication.
• Links: Configuration-specific guides (e.g., → 2× config), Software setup details (/software).

Highlighted Details

• Three scalable configurations: "Home" (2x RTX 5090, 64GB VRAM), "Team" (4x RTX PRO 6000 Blackwell, 384GB VRAM), and "On-prem business" (8x RTX 4090/5090, 192-256GB VRAM).
• Comprehensive build documentation includes BOM, 3D/CNC housing files, wiring, BIOS settings, and photo-based assembly guides.
• Enables running large open-source models locally, from quantized Llama 70B to frontier-class models like Llama 405B, with VRAM dictating model capacity.
• Focuses on data sovereignty and preventing AI compute from being "switched off" by cloud providers or governments.

Maintenance & Community

Community contributions are encouraged via CONTRIBUTING.md, with successful builds featured. Questions can be directed through GitHub issues. No specific community channels (e.g., Discord/Slack) or roadmap details are provided in the README.

Licensing & Compatibility

Licensed under the permissive MIT License, allowing for modification, distribution, and commercial sale of derived works.

Limitations & Caveats

Guides for local model-serving software setup are currently in progress, suggesting potential complexity or incomplete documentation in that specific area.