Source: Wikimedia Commons

My home lab has a problem, and it’s me. If I create documentation of the experiments I’ve been doing, it invariably ends up as a Notepad file in the Dropbox folder I use for home lab things. While at least it’s saved, it’s not very searchable for future use. Other files are saved to GitHub, some are backed up from the folders I use for Docker, and it’s all a glorious mess.

But I realized something. NotebookLM isn’t just for learning new things, it’s also a searchable repository of whatever I decide to upload to it. That means I can use it to dump all the files from my home lab, add the official FAQ and technical documentation, and tease out insights into what I did right — or, more commonly — what I did wrong. It’s fantastic when you mix what you have already done with what you still have to discover, and I can’t work any other way now.

NotebookLM is my second brain

It knows what’s inside my home lab even if I don’t

It’s a wonder that my home lab works at all, because I’m scatterbrained and spend more time looking through documentation on things I’ve done a million times before than actually running any experiments. I’ve got bookmarks in folders that are in no sort of order, folders of PDFs and other documentation that are barely appropriately titled, and who knows how many folders of simple text files containing YAML and other code snippets to use. And we all know how good Windows Search is these days, when it barely works and probably makes my life harder.

Some of those resources are going to stay like that, because there aren’t enough hours in the week for me to organize things. But I’m making a concerted effort to add documentation, how-tos, and relevant add-ons to new NotebookLM notebooks that encompass different aspects of my home lab, from code to hardware, so that it’s all readily available for the AI agent to sort, summarize, and search when I need it.

And it sticks to the sources I want it to

There is nothing worse than crafting a search prompt and getting low-quality results because some websites have crafted their content to be more palatable to GPT web crawlers. But with NotebookLM, it sticks to the sources I add, which means I’ve either vetted the site, or PDF, or created the Google Docs that it is searching in to give me context, troubleshooting tips, or further insight into the intricacies of my home lab setup.

And I can use mixed media, so each notebook includes YouTube videos, podcast audio, and more, all as easily searchable as the text resources inside. I wish this had existed when I was at school, and it’s making it so much easier for me to build the home lab I want.

Everything has a place and now I can find it again

Notebooks to manage documentation intertwined with reference guides

I’m in the middle of tearing down my home lab and rebuilding from scratch, as I made some errors while installing the foundation layer that make it easier to rebuild than repair. I’ve got robust backups of my virtual machines and containers, which is the important part, but I didn’t have much documentation.

As I rebuild, I’m adding each device, tool, container, and other resource to their related notebook, along with the resources that I used to set it up with. That gives me an instantly searchable repository for every single thing in my home lab going forward. Currently I have notebooks for:

• Infrastructure as Code: Ansible, Terraform, and Kubernetes manifests go in here, along with source documentation so I can query when things go wrong
• **Home Assistant: **Mini how-to documents for each add-on or integration live here, along with links to the main Home Assistant help pages
• Network management: Remote access documentation, PDF manuals for every network appliance and software tool, and manifests of VLANs with which devices are where are all kept here
• Containers: Docker Compose files and Kubernetes deployments, along with refreshers on how to use them, links to the relevant Docker Hub pages, etc are all where I need them for troubleshooting
• Proxmox: Links to the main documentation, plus any LXC containers I’ve added, additional software packages, and documentation on the VMs that I’m running

This is already a huge improvement over what I was doing, but it doesn’t stop there. Suppose I run into a problem with part of the home lab. In that case, I can feed log files into the relevant notebook, and ask NotebookLM to cross-reference those errors with how I set things up and the official documentation, and suggest a plan to fix the issue. It’s now a feedback loop with a subject-matter expert who also knows what was created, and while I’m not under any illusions that it will fix everything, it should get me part of the way there.

It’s not perfect, but it’s much better than my old Notepad files

I’m enjoying the process of setting up actual documentation for my home lab, and a knowledge base to work from. Parsing log files to surface relevant data and offer insight into remediation methods is exactly the kind of useful function that natural language processing AI agents should be handling, and I’m enjoying being able to talk without specific keywords and narrowing down search terms as if I were talking to an assistant. Which, really, I am, and it’ll only get better as I add more in-depth and relevant sources for it to draw from.