{"slug": "devember-2026-home-lab-edition", "title": "Devember 2026: Home Lab Edition", "summary": "A homelab practitioner has outlined a Devember 2026 project to turn a dual-R9700 Linux workstation into a local-first AI operator, nicknamed Darth, that monitors infrastructure and performs allowlisted actions through a tool broker rather than direct root access. The project follows months of work on a custom PowerShell assistant built from several million .md files, which the author says outperformed commercial tools including CoPilot on detailed PowerShell projects. The author, who built the machine in Q1 2026 and has only SSH and WebUI access while out of town following a family death, set goals of read-only observation first, then shadow mode and human approval before any autonomous action, with all tool invocations logged and monitoring designed to survive the AI system being offline.", "body_md": "Below is what I am shooting for, though my stepfather passed away last week, so I am out of town with my mother for who knows how long. I have full SSH and WebUI access to the machine I will be testing on, but physical access is relegated to how often my son is down to help be my hands and eyes. I built this machine back in Q12026 and spent several months tooling a custom PowerShell assistant from several million .md files from the usual PS places, of note, MS directly. It has worked well and helped me automate a lot of things in my daily work flow for the better without worrying about token costs. But the main reason is anyone that has used the big players to work on intricately detailed PS projects they just are not that good. Even CoPilot I would say is the worst. But I jays2cents. With all of the INSANE advancements in tooling for the ROCm and R9700 platform specifically of late, I figure why not build what I really need. Outlined below is the local AI assisted on my current build goals for this Devember. Again, if I fall short with everything going on with the family life thing, apologies for wasting your time. Admittedly the goals are lofty to say the least, but not entirely impossible.Without further ado:\n\nMy Devember 2026 project is to learn how to turn a dual-R9700 Linux workstation into a local-first AI operator for my homelab.\n\nThe goal is not simply to run a large language model. The goal is to build a system that can observe the state of my infrastructure, correlate telemetry across otherwise separate systems, explain failures in useful technical language, and eventually perform tightly controlled actions on my behalf.\n\nExamples of the questions it should ultimately answer include:\n\nWhy is this NAS not responding?\n\nWhat is the uptime and current load of that Docker server?\n\nDoes that Windows host appear to need a reboot?\n\nWhich containers are unhealthy?\n\nWhy has a requested movie not reached Plex yet?\n\nWhen was a monitored license plate last observed by UniFi Protect?\n\nWhich smart-home devices are unreachable?\n\nWhat changed immediately before a service failure?\n\nThis is a learning project first.\n\nAI may assist with research, code review, debugging, documentation, and boilerplate, but important architectural decisions, configuration, testing, troubleshooting, and integration work must remain understandable to me.\n\nI should be able to explain:\n\nwhy each component exists;\n\nhow data moves through the system;\n\nwhat permissions each service has;\n\nwhat happens when a dependency fails;\n\nhow a decision or action can be audited;\n\nand how to recover the system without AI assistance.\n\n**Local first.**\n\nInference, telemetry, automation, logs, and household data should remain local wherever technically practical.\n\n**Observe freely; act narrowly.**\n\nDarth may receive broad read-only visibility into the homelab, but write operations will be explicitly allowlisted.\n\n**No unrestricted root shell for the model.**\n\nThe language model will not receive direct root SSH, unrestricted Docker socket access, full Home Assistant administrator access, or blanket UniFi administrator credentials.\n\n**Tool broker between AI and infrastructure.**\n\nInfrastructure actions will be exposed as discrete tools such as:\n\n`get_host_uptime(host)`\n\n`get_container_health(host, container)`\n\n`pause_download(client, id)`\n\n`restart_allowlisted_container(host, container)`\n\n`set_home_assistant_entity(entity, state)`\n\nThe broker, not the model, defines what is possible.\n\n**Read → Suggest → Approve → Automate.**\n\nNew capabilities begin read-only. They progress through Shadow Mode and human approval before any policy-driven autonomous action is considered.\n\n**Everything important is logged.**\n\nTool invocation, arguments, requesting user, result, errors, approval state, and execution time should be auditable.\n\n**Monitoring must survive Darth being offline.**\n\nPersistent telemetry and logs should live on lower-power always-on infrastructure rather than depending exclusively on the AI workstation.\n\n**Recovery before cleverness.**\n\nNo change to kernels, GPU drivers, storage, networking, or remote access is accepted until rollback is understood.\n\nDarth Omniscience:\n\nAMD Ryzen 9 9950X3D\n\n128 GB system RAM\n\n2 × AMD Radeon AI PRO R9700\n\n64 GB aggregate VRAM\n\nUbuntu Linux\n\nDedicated AI/model/data storage\n\nThe first technical phase will benchmark the current R9700 software ecosystem rather than assuming a single inference runtime is best.\n\nCandidate runtimes include:\n\nRadiance\n\nPaiton\n\nvLLM/RDNA4 variants\n\nStrata\n\nllama.cpp\n\nAMD-supported ROCm inference stacks\n\nThe architecture will be chosen from measured results on this hardware.\n\n**Infrastructure observability**\n\nLinux hosts\n\nWindows endpoints\n\nZimaOS systems\n\nTrueNAS/Synology storage\n\nDocker containers\n\nUniFi networking\n\nUPS/power state\n\nsystem logs and metrics\n\n**Home automation**\n\nHome Assistant\n\napproximately 90 WiZ lights and sensors\n\nlocal-only device control wherever possible\n\ninventory, location, naming, firmware state, and health\n\n**IT management**\n\nNinjaOne/NinjaRMM read-only telemetry initially\n\npending reboot and patch state\n\ndevice health and alerts\n\nfuture allowlisted remediation\n\n**Media automation**\n\nPlex\n\nSonarr\n\nRadarr\n\nOverseerr/Jellyseerr\n\nTransmission\n\nrelated storage paths\n\n**Security and cameras**\n\nUniFi Protect\n\ncamera health\n\nevent history\n\nALPR event research\n\nlocal retention/querying of useful event metadata where supported\n\nPrivate administration will not require direct public exposure of the homelab.\n\nA small client application, provisionally called **DarthConnect**, will eventually:\n\ndetermine whether Darth is securely reachable;\n\nestablish the approved VPN path if required;\n\nvalidate the service endpoint;\n\nopen the Darth interface;\n\nreport connection state clearly.\n\nUniFi WireGuard, Teleport, Tailscale, or another appropriate private-access method will be evaluated.\n\nJudges will not receive broad VPN or LAN access.\n\nIf a live demonstration endpoint is provided, it will use a separate Judge Mode with sanitized telemetry and restricted read-only tools.\n\nPrivate systems such as cameras, ALPR data, household automation, NAS contents, NinjaOne endpoints, and unrestricted infrastructure controls will remain inaccessible.\n\nThe default delivery profile is **Captain**:\n\ncalm\n\nmeasured\n\nprecise\n\nauthoritative\n\ndryly understated when appropriate\n\nBritish in cadence\n\nThe intent is a distinguished Shakespearean-command style rather than a literal imitation or clone of any real actor.\n\nBefore Darth receives meaningful write authority, new integrations operate in Shadow Mode.\n\nExample:\n\nUser:\n\n“Why is Zima2 behaving badly?”\n\nDarth:\n\nDiagnosis: HandBrake has been continuously CPU-bound for 47 minutes and its worker stopped producing output 19 minutes ago.\n\nRecommended action:\n\n`restart_container(\"zima2\", \"handbrake\")`\n\nShadow Mode:\n\nNo action performed.\n\nOnly after repeated correct recommendations will a tool graduate to approval-required execution.\n\nA successful Devember project does not require every planned integration to be complete.\n\nThe core project succeeds when Darth can reliably:\n\nobserve several independent homelab systems;\n\ncorrelate their telemetry;\n\nanswer useful operational questions;\n\nexplain how it reached its conclusions;\n\nrecommend a remediation;\n\nexecute a small set of explicitly authorized actions;\n\nrecord what it did;\n\nremain remotely recoverable;\n\nand demonstrate meaningful use of the dual R9700 hardware.\n\nThe project should leave behind not only a working system, but documentation sufficient for someone else to understand the architecture, reproduce the important pieces, and learn from the failures encountered along the way.", "url": "https://wpnews.pro/news/devember-2026-home-lab-edition", "canonical_source": "https://forum.level1techs.com/t/devember-2026-home-lab-edition/257902?page=4#post_79", "published_at": "2026-11-20 05:48:23+00:00", "updated_at": "2026-11-20 06:11:56.182554+00:00", "lang": "en", "topics": ["ai-agents", "ai-tools", "ai-infrastructure", "artificial-intelligence"], "entities": ["Darth", "R9700", "ROCm", "PowerShell", "CoPilot", "Plex", "UniFi Protect", "Home Assistant"], "also_reported_by": [], "takeaway": "Expose infrastructure actions as discrete broker tools like get_host_uptime(host) and restart_allowlisted_container(host, container) instead of giving the model root SSH or unrestricted Docker socket access, and progress new capabilities from read-only through shadow mode and human approval to autom", "recommendation": {"advice": "Give the model read-only visibility and route writes through an allowlisted tool broker with logged arguments, approval state and execution time, never direct root SSH or Docker socket access.", "playbook": "agent-security", "playbook_url": "https://wpnews.pro/playbooks/agent-security"}, "alternates": {"html": "https://wpnews.pro/news/devember-2026-home-lab-edition", "markdown": "https://wpnews.pro/news/devember-2026-home-lab-edition.md", "text": "https://wpnews.pro/news/devember-2026-home-lab-edition.txt", "jsonld": "https://wpnews.pro/news/devember-2026-home-lab-edition.jsonld"}}