Documentation

Everything we have written for you

19 documents, in four groups: one for each of the three assistants, and one for the hardware they share. Every one opens in your browser, and every one has its own contents list at the top so you can jump straight to a section.

SmartMarketSmartServiceSmartCommunityThe hardware underneath

Nothing matches that. Clear the box to see all 19 again.

SmartMarket, the grocery assistant

Ordering a week's shopping by describing it: search, cart, checkout, sourcing and delivery, and the machine it runs on.

SmartMarket: specificationSpecification

What it does, and the two things it is actually judged on: how fast it finds something, and whether the money is right

31 August 2026

What it does, what holds it back, where it can goBrief

The brief: what the product does today, what makes it slow, what it would take to sell it, and what to build next

8 August 2026

Setting up your own copySetup guide

Start a new machine from a snapshot, give it its own address, and check it in a browser. Written for somebody who has never used a server

17 August 2026

Product search and cartMilestone

The search and cart milestone: what was asked for, what was built, and the result of every test

10 August 2026

The interface redesignWalkthrough

A redesign of the shopper's screens, and the four faults it uncovered

8 August 2026

Payment, sourcing and speedWalkthrough

What was built across three phases, how it was done, and what each change did to the numbers

8 August 2026

Word versions

SmartService, the booking platform

Describing a problem and ending up with a tradesperson booked, paid and confirmed. Also where parking passes and the hardware switch live.

The Service Assistant: specificationSpecification

What the system does, who uses it, and how it must behave under load and under failure, ending in a live console you can run

29 August 2026

Break the assistantInteractive

Our AI infrastructure under load, as something you can drag: turn the traffic past a million requests a day, starve the cache, kill a card, and watch what the design does about it

29 August 2026

The booking platformWalkthrough

Describe a problem, see who can do it, pick a time, book it. What a customer does, what a provider does, and where it runs

12 August 2026

Requirements, for approvalRequirements

What the booking assistant was going to be, before it was built. Kept so the finished thing can be checked against it

11 August 2026

Spin-off and plumbing handoverHandover

Two applications from one codebase: what differs between them, and how to stand either one up on a new machine

14 August 2026

Word versions

SmartCommunity, the association assistant

Answering a resident out of their own association's documents, and naming the page it came from. Its own site since August 2026.

SmartCommunity: specificationSpecification

What it answers, what it refuses, and how one association is never answered from another's rules

31 August 2026

How the community assistant answersExplainer

In plain language: where the answers come from, what happens between a question and the reply, and what it does when your documents do not cover it

27 August 2026

Community documents: how the assistant answersTechnical reference

Uploading a document yourself, downloading one, parking passes, and how a resident's question is routed so that one community is never answered from another's rules

26 August 2026

Adding a community documentHow to

How to add one yourself in a minute, what can and cannot be used, and what happens to a document once it is in

26 August 2026

The hardware underneath

Running the model on a machine we control, what it costs, and how the applications carry on when it is switched off. Shared by all three agents.

Spinning up a GPU, and wiring it to FastAPIStep by step

From an empty AWS account to a resident's question answered by our own model, in nine steps. Written from a real attempt, including the two things that stopped it

31 August 2026

Our own GPU, in plain wordsExplainer

What it means to run the AI on hardware we control, what it costs when it is on and when it is off, and why forgetting to switch it off does not cost anything

29 August 2026

Running the AI on our own GPU: setupTechnical setup

Creating the machine and wiring it to the Service Assistant: the quota that blocks everything else, the AWS permissions, the two mechanisms that stop it when nobody is using it, and how to check each one works

29 August 2026

Installing the open model, and wiring it to FastAPIInstallation

The software half: installing Ollama, choosing a model, making it reachable, the four modules that decide where a question goes, and how to run the whole thing with no GPU at all

29 August 2026