new project. rackfolio: an inventory and a portfolio for your homelab. you add every device, disk, and service once, and it gives you the numbers homelabbers actually care about — total power draw, estimated monthly cost, terabytes, cores, rack units used — then, later, a page worth showing off.
this is the routine i never had a good home for. my own lab lives in a markdown file that drifts out of date the week after i touch it.
naming it with my own tool
i ran the candidates through namescout, which is exactly what i built it for.
the .com landscape for anything homelab-shaped is a graveyard — rackdex,
racknest, rigster, labfolio, geardex, all gone. dozens checked, almost nothing
left short and clean.
rackfolio survived, and it was the only good one open on .com, .dev, and
.io all at once. it also says the product out loud: rack + folio, a portfolio
of your rack. registered the .com before writing this — naming a thing in
public before you own the domain is how you lose the domain.
the foundation
- next 16 (app router), react 19, typescript, tailwind v4 on the css-first
@themesetup - postgres 17 running locally in docker —
docker compose up -d, persistent named volume,DATABASE_URLwired. neon later; local first so the repo runs for anyone who clones it - prisma teed up, schema next
nothing fancy yet. page.tsx still says hello world. this entry is foundation
and design, not a working app — the react port and the data layer are the next
session.
drawing it before building it
before writing components, the screens started as static tailwind mockups — the lab overview with the stat block (power, est. cost at your $/kWh, devices, storage), the device strip, a device-detail page. dark, a little industrial, one orange accent. this is the prototype layer, and it’s where i let ai help: roughing out layouts fast so i can make design calls while they’re cheap.
the react port is the opposite — that’s a translation job i do by hand.
the plan
the ordering is the strategy: a genuinely good private tool first, sharing as a layer on top, community last.
- the tool — private inventory, the live stat block, devices and services. no auth, no sharing. i’ll dogfood it on my own lab.
- profiles — accounts, claim a username, flip a lab public at
/u/<name>. the part you’d post to r/homelab. - the catalog — a seeded gear database so power and cost stop being hand-typed guesses and start being real.
- community — browse other people’s labs, filter by what they run, and the leaderboards i actually want to see: most powerful, best watts-per-core, biggest storage.
the boundary on this one: ai is for support and prototypes — naming, scoping, roughing out the design — and that’s it. the application code is hand-written. i’ve spent a long stretch building with ai in the loop and i can feel a few muscles going soft, so on rackfolio the code is mine, keystroke by keystroke.
tool first. the part everyone sees comes after the part only i see is good.