Proxmox smart-home virtualization

Home Assistant + Frigate Proxmox Server Build

Proxmox can consolidate Home Assistant, Frigate and local AI, but it also turns every hardware accelerator into a device-ownership decision. Home Assistant OS is comfortable in a VM; Frigate’s documentation says virtual machines are generally not its preferred path but specifically recommends a QEMU VM over LXC when using Proxmox. Local AI adds GPU passthrough and memory reservations on top.

See live hardware Size my workload

Quick answer

What this combined workload needs

Use Home Assistant OS as its own VM when you want the recommended Home Assistant lifecycle and apps. If Frigate is virtualized, favor a Proxmox QEMU VM with clean iGPU/GPU/accelerator passthrough rather than an unsupported LXC setup. Keep Ollama in the VM/container that owns the GPU it needs.

Live Amazon hardware

Current products that fit this decision

Compare current Amazon listings relevant to this guide. Product availability and prices can change.

Checking the dedicated Home Assistant + Frigate + AI catalogue…

Buying decision

Match the hardware to the bottleneck

Do not build the VM layout before the device map. Write down who owns the iGPU, discrete GPU, Hailo/Coral device, Zigbee/Thread/Z-Wave coordinator and recording storage. If two guests both depend on exclusive access to one device, the architecture needs to change before installation.

Interactive decision tool

Proxmox Smart-Home Resource Planner

Size the combined workload before comparing products. Results are planning tiers, not benchmark guarantees. Verify current Home Assistant installation support, Frigate detector/device access and Ollama hardware support for the exact system.

Compatibility checklist

Four checks before you purchase

Assign every accelerator a job

Video decoding, Frigate detection, enrichments and Ollama inference are separate workloads. Do not assume one NPU/GPU automatically covers all of them.

Protect Home Assistant headroom

Size the combined host for peak overlap so camera or LLM activity cannot starve automations, databases or radio services.

Verify device access

USB radios, iGPU/NPU/GPU and Hailo devices must be visible to the correct host, container or VM before their benchmark performance matters.

Keep data roles separate

Use fast SSD space for system/models and size Frigate recordings independently so retention cannot crowd the smart-home environment.

Home Assistant OS is the straightforward VM

Home Assistant OS is the recommended installation type for most users and is designed to run on virtual machines as well as dedicated hardware. In Proxmox, that gives Home Assistant its Supervisor/apps lifecycle without asking you to maintain Home Assistant Container manually.

Allocate stable CPU, RAM and storage, then focus attention on USB radios and network access rather than overprovisioning the automation VM.

Frigate in Proxmox should favor QEMU over LXC

Frigate’s current installation guide says VMs are not generally recommended because hardware access can be harder, but in its Proxmox section it points users toward running application containers inside a QEMU VM. It explicitly says LXC is not officially supported and can be complex.

That distinction matters because many homelab tutorials promote LXC for lower overhead. Choose the supported/maintainable path that matches your risk tolerance rather than the smallest possible memory footprint.

Disable memory ballooning for GPU-passthrough Frigate VMs

Frigate’s Proxmox guidance specifically warns to disable ballooning, especially when passing through a GPU. Predictable guest memory helps hardware-accelerated workloads avoid surprises.

Reserve enough RAM for camera buffers, enrichments and the guest OS instead of planning to reclaim it dynamically whenever the host is busy.

The iGPU can only help if the guest can actually see it

Intel Quick Sync or other hardware decoding is one of the most valuable resources in a Frigate host. Passing it through changes how the Proxmox host and other guests can use it.

Confirm IOMMU/device behavior for the exact platform before buying a mini PC solely because its iGPU looks attractive on paper.

Dedicated accelerators simplify GPU allocation

A Hailo or similar Frigate detector can reduce pressure to give Frigate the same discrete GPU that Ollama needs. The detector becomes a small pass-through device while the larger GPU belongs to the AI guest.

This is often cleaner than attempting to share one GPU among several VMs without a supported partitioning method.

USB radios need persistent identity

Zigbee, Thread and Z-Wave coordinators should be passed to the Home Assistant VM in a way that survives reboots and device enumeration changes. Use stable device identifiers where possible.

A network coordinator can avoid USB passthrough entirely, which may be valuable when Proxmox migrations or hardware maintenance are part of the plan.

Ollama GPU ownership should be explicit

A local AI VM with an NVIDIA or AMD GPU needs that device, its drivers and the appropriate container/runtime configuration. Passing the GPU through to one VM generally means other guests do not casually share it.

If Frigate also needs a discrete GPU, either allocate a second device or redesign Frigate around iGPU/Hailo/OpenVINO so the Ollama VM owns the large GPU.

Storage should follow failure domains

Keep Home Assistant VM disks and backups small and recoverable. Give Frigate recordings a storage path sized for sustained video writes. Give Ollama models fast SSD capacity without pretending model files need the same redundancy as irreplaceable Home Assistant configuration.

A single ZFS pool can host all of these, but separate datasets/volumes and backup policies make recovery clearer.

Do not back up terabytes of Frigate recordings like VM state

Proxmox VM backups are useful for Home Assistant and service configuration, but huge recording volumes can make backup jobs impractical. Exclude or separately manage disposable footage according to your retention policy.

Important incident clips should be exported intentionally instead of relying on the rolling NVR volume as the only copy.

Network architecture becomes easier to enforce

Proxmox bridges and VLAN-aware networking can separate camera traffic, trusted Home Assistant traffic and model services. That is useful for security, but unnecessary complexity can also create hard-to-diagnose integration failures.

Keep the routes Home Assistant needs to Frigate and Ollama explicit. Use DNS names and firewall rules you can document rather than a web of NAT exceptions.

Leave host-level headroom

The hypervisor itself needs memory and CPU, and storage scrubs, backups or migrations can overlap with camera activity. Size the physical host from the sum of guest peaks plus host reserve rather than adding each VM’s minimum allocation.

A 32GB machine is workable for modest stacks; 64GB gives much more freedom once an Ollama VM and several other services join the host.

Virtualize only when the management benefits are real

Snapshots, backups, isolation and service separation are strong reasons to use Proxmox. If the box exists only to run Frigate and Ollama with direct hardware access, bare-metal Linux/Docker may be simpler and more efficient.

Use Proxmox because you need virtualization, not because every home server diagram is expected to include it.

Questions people ask

Home Assistant + Frigate Proxmox Server questions

Should Home Assistant OS run in Proxmox?

Yes, Home Assistant OS is designed to run as a VM and remains the recommended Home Assistant installation type for most users.

Should Frigate run in an LXC?

Frigate states that LXC is not officially supported and can be complex. Its Proxmox guidance favors a QEMU VM for application containers.

Should ballooning be enabled for Frigate?

Frigate warns to disable ballooning, especially when passing a GPU through to the VM.

Can one GPU serve Frigate and Ollama in separate VMs?

Not automatically. Device passthrough typically assigns the GPU to one guest. Plan separate accelerators or a different architecture.

Can I pass a Zigbee USB stick to Home Assistant OS?

Yes. Use stable USB/device mapping so the coordinator remains attached after reboot.

Is Hailo useful in Proxmox?

It can be because a small dedicated detector can keep Frigate off the GPU that Ollama needs. Passthrough still must work for the exact device.

How much RAM for the Proxmox host?

32GB can handle a modest stack; 64GB is more comfortable for Ollama, multiple VMs and future services.

Should Frigate recordings live inside the VM disk?

Usually use a dedicated volume/mount or storage path sized for recordings rather than bloating the VM system disk.

Can Home Assistant talk to Frigate in another VM?

Yes. The integration is network-based, so the services can be separated as long as routing and authentication are configured.

When is bare metal better?

When direct device access, lower complexity and predictable performance matter more than VM isolation and snapshots.

Official references and methodology

Verify current Home Assistant, Frigate and Ollama support

Cloudzat treats Home Assistant, Frigate and local AI as separate workloads that share one hardware envelope. Home Assistant installation guidance, Frigate device-access and detector guidance, and Ollama GPU/context documentation are used as the technical baseline. Calculator outputs are planning tiers rather than benchmark guarantees, and live Amazon listings come only from this sprint's dedicated catalogue.

As an Amazon Associate, Cloudzat may earn from qualifying purchases. Prices, seller terms, exact configurations and Home Assistant, Frigate or Ollama hardware support can change.