homestead-ai — Your Local AI Harness

Discover, track, and serve any local LLM from any engine — with one CLI.

homestead discover   # Scan Ollama, HF cache, GGUF files, MLX, running engines
homestead list       # Tabular view of all models (or filtered)
homestead serve qwen # Auto-selects best engine, serves on :8080
homestead ui         # Web dashboard on :3030

homestead is the universal model agent harness for local AI. It treats Ollama, llama.cpp, MLX, and Apple Core AI as first-class citizens in a single SQLite registry — so you don't have to remember which engine has which model, or where that GGUF file is buried.

Quick Start

# Requires Bun
curl -fsSL https://bun.sh/install | bash

# Clone and run
npm install -g homestead-ai
git clone https://github.com/clawcenter/lai.git
cd lai
bun install
homestead discover   # Find all your local models
homestead serve model-name  # Serve any model on :8080
homestead ui         # Launch the web dashboard

Features

• Zero-config discovery — finds models in Ollama, HuggingFace cache, GGUF files, MLX directories, and running engines.
• Cross-engine serving — picks the best engine for any model (OllamaAdapter for Ollama models, LlamaCppAdapter for standalone GGUFs).
• Audit trail — every model load, unload, and error is logged to SQLite and streamed via SSE. Prove no data left your machine.
• Observability dashboard — real-time event stream, session tracking, and model health monitoring in the browser.
• CLI-first — designed for scripts, CI/CD, SSH, and agentic workflows. The UI is a monitoring dashboard, not a chatbot wrapper.

Architecture

[Ollama] -----\
[HF Cache] ----> [ Scanner ] --> [ Registry ] --> [ Unified API ] --> [ Agent/UI ]
[Local GGUF] --/                    |
                                    v
                             [ SSE Event Stream ] --> [ Audit Dashboard ]

homestead is a horizontal orchestration layer — it does not replace your inference engine. It sits above them all and provides a single pane of glass.

License

Apache 2.0 OR MIT — your choice. See LICENSE.

The Claws Center also offers enterprise plugins (SSO, RBAC, audit retention, BiFrost gateway, multi-user model pools) and the Apple Core AI runtime for .aimodel serving. See ENTERPRISE.md or visit https://clawcenter.io/lai.