<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Architecture Decision Records — AI VM Management :: Workspace Docs</title><link>https://samplayskeys.github.io/Workspace/homelab/ai-vm/adr/index.html</link><description>ADRs here capture decisions for the AI VM homelab project.&#10;# Title Status 0001 LlamaFarm as AI model management layer Proposed</description><generator>Hugo</generator><language>en-us</language><atom:link href="https://samplayskeys.github.io/Workspace/homelab/ai-vm/adr/index.xml" rel="self" type="application/rss+xml"/><item><title>ADR 0001: LlamaFarm as AI model management layer</title><link>https://samplayskeys.github.io/Workspace/homelab/ai-vm/adr/0001-llamafarm-over-ansible/index.html</link><pubDate>Mon, 01 Jan 0001 00:00:00 +0000</pubDate><guid>https://samplayskeys.github.io/Workspace/homelab/ai-vm/adr/0001-llamafarm-over-ansible/index.html</guid><description>Status: Proposed&#10;Date: 2026-07-14&#10;Context The homelab has a dedicated AI VM running Ubuntu with assorted AI tools installed ad-hoc. There is no consistent model management layer — models are pulled manually, runtimes are configured individually (Ollama, vLLM, raw HuggingFace), and there is no unified API surface or RAG pipeline.&#10;As part of a broader effort to make this VM “more than just a box with AI tools on it,” a management layer is needed for:</description></item></channel></rss>