>=9.0.0 <10.0.0 · 1 bounded recipe
Proxmox VE 9.x
- supported deployment
- Proxmox VE on bare metal
- known issues in scope
- 0
- Proxmox host with two light service VMs
version-scoped application and deployment ledger
Browse the maintained application releases, supported deployment methods, bounded workload recipes, known issues, and the pilot's primary deployment pattern.
Each record fixes a release range, supported or discouraged deployment method, bounded workload recipe, and known-issue scope. A newer release, different runtime chain, or unmodeled workload needs a reviewed record; it does not inherit a positive result from the same product name.
This hub is intentionally non-index-candidate during the pilot. It helps people navigate maintained records without increasing the exceptional detail-page ceiling.
>=9.0.0 <10.0.0 · 1 bounded recipe
>=2026.7.4 <2026.8.0 · 1 bounded recipe
>=10.11.0 <10.12.0 · 1 bounded recipe
Proxmox host with isolated appliance and application guests. The pilot's maintained topology for one exact mini PC: Proxmox VE on bare metal, Home Assistant OS in its own KVM virtual machine, and containerized applications inside a separate Debian KVM virtual machine. It is an architecture contract, not a performance or resilience result.
candidate · >=3.0.0 <4.0.0
This release stays in the private graph for rule and maintenance testing. It has no graduated application page and cannot expand the pilot by implication.
deferred · >=0.17.0 <0.18.0
This release stays in the private graph for rule and maintenance testing. It has no graduated application page and cannot expand the pilot by implication.
Grade A count: 0. A Grade B source can establish a documented specification or requirement. Grade D records are modeling inputs. Neither is proof of measured throughput, thermals, power, reliability, or physical fit.
a grade describes how well attested a statement is, not how good the hardware is. grade a is a lab run we reproduced with a published method. grade d is a model, a reasoned inference, or a single unverified report.
evidence grade b · official sourcesupports
For production, Proxmox VE documents Intel 64 or AMD64 with hardware virtualization, at least 2 GB for host services plus designated guest memory, fast redundant storage with SSDs preferred, and redundant multi-gigabit network interfaces. Its separate evaluation-only minimum is not a production recommendation.
evidence grade b · official sourcesupports
Proxmox VE 9.0 is an official stable release line based on Debian 13, and its ISO is installable directly on bare metal.
evidence grade b · official sourcesupports
The generic x86-64 Home Assistant OS image requires x86-64, UEFI boot, compatible storage, and a supported network adapter; Secure Boot must be disabled.
evidence grade b · official sourcesupports
Home Assistant documents a KVM/Proxmox QCOW2 deployment with minimum VM resources of 2 GB RAM and 2 vCPUs, requires UEFI, and directs KVM users to a non-Secure-Boot OVMF image. The example KVM command allocates 4 GB, while the guide says resources should be sized to the expected workload rather than asserting a universal higher recommendation.
evidence grade b · official sourcesupports
The official July 2026 release notes list Home Assistant Core patch release 2026.7.4.
evidence grade b · official sourcesupports
Jellyfin recommends 8 GB RAM, SSD storage for application data and transcode cache, wired gigabit networking, and properly configured hardware acceleration for transcoding.
evidence grade b · official sourcesupports
Jellyfin publishes an official Debian-based container image and documents 10.11 tags, but does not establish the pilot's additional KVM guest layer.
evidence grade b · official sourcesupports
Jellyfin 10.11 is a documented stable server release line and requires a 64-bit operating system on ARM.
evidence grade b · official sourcesupports
Immich v3 requires a 64-bit Linux-like environment, at least 6 GB RAM and 2 CPU cores, Docker Compose, and x86-64-v2 for the amd64 machine-learning container; 8 GB and 4 cores are recommended.
evidence grade b · official sourcesupports
Immich recommends keeping PostgreSQL data on a local SSD and estimates thumbnails and transcodes can add roughly 10–20 percent to library size.
evidence grade b · official sourcesupports
Frigate recommends Docker on bare-metal Debian-like Linux for low-overhead access to GPU and detector devices and does not recommend a virtual machine as the default path.
evidence grade b · official sourcesupports
Frigate is designed around a supported detector and documents several accelerator families; camera count depends on streams, codecs, models, activity, and detector throughput.
evidence grade b · official sourcesupports
Frigate 0.17 adds and changes accelerator support, including Intel OpenVINO improvements and NPU support.
evidence grade b · official sourcesupports
Jellyfin says Intel 7th through 10th generation integrated graphics remain capable for existing owners, but are no longer recommended for a new purchase because Intel deprecated the relevant toolkit.
evidence grade b · official sourcesupports
Immich does not recommend Docker inside LXC and recommends a supported full virtual machine when virtualization is required.
evidence grade b · official sourcesupports
Frigate recommends H.264 camera video for the broadest feature compatibility and notes that multi-stream cameras can avoid unnecessary re-encoding.
evidence grade b · official sourcesupports
Frigate highly recommends an integrated or discrete GPU for hardware-accelerated video decoding and documents Intel integrated graphics as a supported acceleration path.
evidence grade d · modeled or inferredsupports
Resource allocations in candidate recipes and the starter stack are planning budgets, not measured throughput or latency claims.
limitation: This is a modeled or inferred claim. It cannot establish measured performance.
Official facts and first-party modeling are shown separately. Source prose, tables, and diagrams are not republished.
This projection is generated from homelab graph version 0.1.0. The generated lifecycle index exposes only history events that are explicitly present in the structured record. It does not invent an earlier revision when the graph contains only its creation baseline.
inspect the complete generated lifecycle and change-history index