Real-world
Homelab ops
Concrete workflows, including the tradeoffs and limitations reported by their authors.
this collection
Website fixes, research, and Linux maintenance on Strix Halo and DGX Spark
Beyond the obvious coding use case, this setup handles website debugging, deep web research, and day-to-day Linux server maintenance.
A homelab dashboard and desktop cleanup, vibe-coded with Pi
Qwen 27B and the Pi coding agent built a custom chat UI, an mp3 server, and a homelab control panel — and did the desktop tidy-up that first proved the setup was worth keeping.
Migrating an agent between machines over text message
Using Hermes to patch DeepSeek to a vision release and then move the whole agent, containers, and environment to another computer — remotely, from a phone.
Local coding for a homelab app, with ChatGPT as the second reviewer
Qwen3-Coder through Cline writes the app; a paid ChatGPT Plus subscription reviews the prompts and the output before it ships.
A sub-10GB model as the brain behind Home Assistant
One small model does double duty: correcting speech-to-text output and acting as the decision-making 'brain' for home automation triggers.
'Add this app to my cluster' — and it just does it
Hermes on Qwen 3.6 35B-A3B runs this person's entire homelab k3s cluster hands-off, including diagnosing a flaky network cable before it was told to.
Agents running well on an old Intel PC with no dedicated GPU
Countering a cost complaint elsewhere in the thread, this person runs agents on genuinely low-spec, CPU-only hardware and reports about 30 cents of electricity for a full day.
Letting a local model program ESP32 boards and read the serial output back
A hobbyist gives a local model a terminal and a serial device so it can build, flash and debug microcontroller sketches, and also runs ticket classification for work.
Life automation and lab monitoring, used as insurance against cloud shifts
An architect uses local models to automate downloads, tasks, backups, Android tools and lab log monitoring, treating the setup mainly as a safety net and a tinkering hobby.
Setting up and babysitting a Raspberry Pi camera over SSH
A local model SSHes into a Raspberry Pi to configure a kitchen security-camera stream plus a watcher that restores it after power or connection loss.
Going fully local on a home server for under $3,000
Using a gifted 600GB server and a cheap used P100, this owner reports that a capable fully local setup is doable for under $3,000 today.
Coding and homelab maintenance on a dual-R9700 server
A garage server with two Radeon Pro AI R9700 cards supports coding, homelab maintenance and an evolving Cline/Open WebUI workflow.
Making an 8GB laptop useful through DIY agent workflows
A contributor experiments with agents, tools and MCP on an 8GB laptop, using lightweight Gemma models to turn constrained hardware into a practical coding playground.
Zero cloud usage since getting a Strix Halo — almost
DevOps work, smart-home control, and paperwork all moved off the cloud once Strix Halo hardware arrived — with one exception the poster caught themselves forgetting.
A daily-refreshed RAG advisor built from your own infrastructure
A Python collector pulls read-only configs, firewall rules, and network health data every day and feeds it into a local RAG pipeline, so the advisor always knows the current state of the homelab without ever being able to touch it.
Reducing security-camera inference power use with a local model
A local model optimized a continuously running YOLO security-camera workload and reportedly cut its power use by 40–50%.