~24
Specialized services
We didn't just build websites with AI. We built our own AI — a private system that runs on our own hardware, thinks from a real knowledge base, and gives our platforms intelligence we don't have to rent.
Every "AI feature" you buy from a software vendor comes with two strings attached: a fee for every user every month, and your customers' data living on a server you don't control.
For a one-person shop running several businesses at once, that math breaks fast. Charge through the nose per seat, or hand your data to a third party — neither is a foundation to build on.
So instead of wiring every project to someone else's chatbot, Tandy Services built its own AI from the ground up: a private system on dedicated hardware that any of our platforms can plug into. What started as an experiment grew into a fleet of around two dozen specialized services working together.
The payoff is simple — own the brain, and every platform gets real AI features at almost no extra cost, with the data staying ours.
Instead of one giant black box, Tandy Services AI is many focused services — each doing one job well — coordinated behind a single secured gateway. Here's what's under the hood.
A vector knowledge base the AI searches before answering, so responses are grounded in real, specific facts — not invented.
A chat service with memory, plus an intent router that classifies each request and sends it where it belongs.
Reads and describes images, and generates original artwork — used for article visuals and tattoo placement previews.
Checks messages and scans uploaded images for unsafe content before anything reaches a user.
Condenses long documents and generates written content — from news rewrites to drafts grounded in real data.
A watchdog monitors every service and restarts anything that fails; an incident service escalates, and an auto-repair service even attempts code fixes.
The core idea is "retrieval-augmented generation" — fancy words for a simple habit: look it up before you answer.
Thousands of curated facts live in a searchable knowledge base. When a question comes in, the system finds the most relevant pieces and hands them to the model as context. The result speaks from real information instead of guessing.
Not every question needs a sledgehammer. Tandy Services AI sizes up each request and sends it to the cheapest engine that can answer it well.
Quick, factual questions get a fast local model. Medium ones step up. The genuinely hard, nuanced questions go to a top-tier 70-billion-parameter model — and if any tier struggles, it automatically escalates to the next.
Answers from a real knowledge base, with memory of the conversation so far.
Creates artwork and reads images — powering news visuals and placement previews.
Writes drafts and condenses long material into something usable.
Filters unsafe messages and scans uploaded images before they go live.
The fleet is the foundation. The goal is to turn it into something that doesn't just answer questions — it does the work.
One command center — better than a chat box — where you talk to the AI by voice and it acts: builds sites, generates images, reads and writes documents, and shows the live health of every service at a glance.
Right now the brain is wired into one platform. The plan is to plug it into more Tandy Services builds, so every new business gets grounded AI features for free instead of renting them.
Speech in and out, document understanding, and generated media — turning the system from a text tool into a hands-on assistant you can actually talk to.
A routing engine already lives in the fleet — the groundwork for transport and oversize-load products when those move off the back burner.
If you've got a platform that needs genuine AI features without the per-seat tax, let's talk about what Tandy Services AI could do for it.