AI Inference VM

activesnapshotverified 2026-07-20

Status: active Last verified: 2026-07-20 (against the current hardware inventory and AI platform overview) Scope: VM spec, inference stack, and service inventory for the primary AI inference host

Purpose

This Proxmox guest is dedicated to local GPU inference and AI control-plane services. It provides OpenAI-compatible model access for editor integrations, chat interfaces, and automation agents.

RAG services (Qdrant, embeddings, and indexing) run on a separate compute appliance and no longer share these GPUs. Media serving also moved to a dedicated guest in July 2026. This VM is now inference and control-plane only.

Current Shape

ComponentSpec
HypervisorProxmox VE on AMD EPYC 7402P (24C/48T)
vCPUs48 (1:1 with host threads, no overcommit)
RAM96 GiB dedicated (1G hugepages)
GPUs2x NVIDIA RTX 3090 (24 GB each) with NVLink bridge
GPU passthroughFull PCIe passthrough (GPU + audio), no mediated devices
OSUbuntu 24.04.4 LTS, kernel 6.8.x (PREEMPT_DYNAMIC)
NVIDIA driver580.126.20
CUDA12.9
ContainerDocker
StorageLocal SSD-backed VM disk + NFS mounts from TrueNAS (models) and dev host

Inference Stack

LayerModel / Role
Deep vLLM laneQwen3.6-27B AWQ, tensor parallel across both GPUs (TP=2)
LiteLLM front doorStable deep and fast capability aliases across separate inference lanes

All inference runs in containers. No models run directly on the guest operating system.

The deep lane (vLLM TP=2) uses both GPUs for a single 27B model. The previous one-model-per-GPU pattern (Qwen 2.5 7B + 14B) was replaced in favor of a unified tensor-parallel serving. A 32B AWQ alternative mode exists but requires CUDA graph capture narrowing (max_cudagraph_capture_size=8) to avoid Xid 13 faults on dual 3090 hardware.

Control Plane & Adjacent Services

ServiceRole
Open WebUIBrowser-based chat interface
LiteLLMAPI router for deep and fast capability lanes
Chat serviceSupplementary chat and skill dispatch
Voice notesBrowser capture and speech transcription
vLLMDeep inference engine
Matrix serviceInternal chat transport
SearXNGPrivacy-focused search engine
Agent servicesScheduled analysis and orchestration

RAG, vector storage, embeddings, and indexing run on the separate retrieval appliance. Consumers reach them through private service contracts that are intentionally omitted here.

Constraints

Drift / Unknowns