Welcome to AI Server

AI Server, also called the AI Command Interface, is a local-first AI and homelab dashboard. It brings local AI chat, model management, system monitoring, and an optional Discord bot together in one interface.

The project is being developed on Windows, with a Linux/Proxmox deployment being prepared and tested in a nested lab.

What can it do today?

✅ Working — Windows development environment

Local AI chat, model switching, system telemetry, user authentication, private conversation history, and on-demand, read-only diagnostics are implemented. The optional Discord bot can report status and control the active model.

🧪 Partially validated — nested Proxmox lab

The dashboard and backend have been deployed in a nested lab, with HTTPS access and read-only Proxmox inventory exercised there. Lab results do not establish acceptance on the physical server.

📋 Planned — not yet implemented

Concurrent AI profiles, live remediation, and production vision and image generation integrations remain future work.

⚠️ Not yet validated — target host

Linux GPU/ROCm inference and the physical Proxmox deployment still need target-host validation.

See the roadmap for detailed status and sequencing. The project overview explains the components and their current capabilities.

Get started

• Try the Windows development setup: Follow the Windows quick start and the detailed testing guide.
• Understand the project: Read the project overview and system architecture.
• Explore deployment: Start with the deployment guide and its Linux/Proxmox runbook.
• See what's next: Browse the project roadmap.

Explore the documentation

• Backend and API reference — backend modules and API contracts.
• Frontend — dashboard panels and browser modules.
• Security — authentication, sessions, and access, including the authentication guide.
• Operations — configuration, testing, and infrastructure contracts.
• Diagnostics — current observations and the proposed self-healing design.
• GPU — hardware integration and validation.
• Project context — detailed contributor and AI session onboarding.

How to read status callouts

Throughout these guides, Working identifies implemented behavior in the specified environment; Partially validated identifies scoped evidence, not full acceptance; Planned identifies future work; and Not yet validated identifies behavior that still needs testing. Always check the environment and scope stated in each callout.