R&D: Research, Photos, And Parts Facts

What this bot is for

R&D is the facts underneath the site and the counter: specs, research, reference photos, and catalog reading. When a Localizer sentence or a Designer price chart needs to be true, this is the bot that checks. On the pulse map, the helpers around this work — Site Reader, Supplier Reader, Photo Sourcer, and Render — belong with R&D as supporting bots. They read, they fetch, they draw. They do not set prices and they do not publish.

This guide explains the public pattern. Confidential methods, supplier names, and part costs stay off the page.

Why a shop needs a research bot

Repair shops argue with the internet every day. A customer has read a forum post. A listing has an old spec. A guide you wrote last year used a phrase you would not use now. Without a research bot, you either spend the evening checking, or you hope yesterday's memory is still right.

R&D's job is to keep a shared, verified set of notes that the rest of the team can use. This shop already maintains a shared knowledge index and refreshes it so every bot can find the same facts. When a claim cannot be sourced, the bot says so. "I'm not sure" is a valid research result. A confident guess is not.

Specs and a clean parts list

Public price menus are built from a pricing curve and from model and parts facts, not from a vendor's marketing name. The long-read version of this idea is simple: you keep a clean internal list of devices, repairs, and tiers; you use that list to fill the public chart; you never put cost or a supplier name on the chart the customer sees.

R&D can:

  • Check a model name, a screen size, or a battery description against verified notes and official manuals.
  • Flag a listing that mixes two devices or two tiers.
  • Tell Designer and the Localizer which public wording is safe.
  • Watch for catalog changes that might affect a price later, and report those changes internally — without writing the vendor's name into a public draft.

It should not paste an internal catalog row onto a blog post. It should not explain a tier by saying where the part came from.

Helper bots: reading, photos, renders

The pulse map shows four helpers that sit near R&D.

Site Reader checks live pages and source material: is this sentence still on the official manual, is this guide citing a page that moved, did a public page accidentally include a vendor name or a $0 price. It is a reader, not a publisher.

Supplier Reader looks at catalogs and change lists so the shop can see when a cost or an availability line moved. That work stays internal. The public article should never name the supplier. The report goes to the Accountant and the Designer, not to a customer email.

Photo Sourcer finds reference photos you are allowed to use: your own shop photos first, then official-safe references when you need a clear parts shot. It does not scrape random social posts, and it does not use a customer's lock-screen photo.

Render (and related diagram work) happens on Steve's own computer. The pulsing team diagram on this series page is that kind of work: a picture of how the team is set up, not a live camera on the shop. If you want a new diagram, R&D and the Designer plan it, render it locally, and leave it in a draft theme or an unpublished page until you say it can go live. (See also the later guide on running Python and a graphics card on your own computer.)

What is public vs what stays internal

Public: verified specs that help a customer choose a tier, photos you own or may use, diagrams that explain a process, source-backed sentences in a repair guide.

Internal: supplier names, costs, margin, wholesale availability, customer devices, credentials, and any method you do not want copied into a forum post.

If a helper bot's output contains both, R&D should split the file. The Localizer only receives the public half.

How R&D hands facts to Localizer and Designer

A clean handoff looks like this:

  • The question (what we want to say).
  • The source (what we checked, quoted briefly).
  • The verdict (safe to publish, needs a rewrite, or do not say this).
  • The public wording, if any, already stripped of vendor names and costs.
  • The internal note, stored separately, if someone on the team needs the private half.

Designer uses that to update a draft theme or a price page. Localizer uses it to write the sentence the customer reads. Front Desk can use a one-line version at the counter. Nobody should be rewriting the same fact three different ways from memory.

Sample prompts

Check this repair-guide claim against our verified notes and official manuals. Quote the source, say whether we can publish it, and list anything we should not say (vendor names, costs, or guesses).

Make a public-safe facts sheet for this device and repair: what the customer needs to know, what the tiers mean in our public language, and what we will not mention. Keep costs and supplier names out of the public sheet.

Find a reference photo we are allowed to use for this guide. Prefer our own shop photos. If you cannot find one, say so. Do not invent an image and do not use a customer photo.

Site Reader: scan this draft page for broken links, a $0 price, or a vendor name in customer copy. Report only. Do not edit the live page.

Summarize today's catalog-change watch for the Accountant and Designer. Keep supplier names out of any draft that might be published. Do not change a price.

How this bot works with the rest of the team

R&D is upstream of almost every public sentence. Localizer writes, Designer places, Front Desk quotes, Social links only what is already live and sourced. Cortana only needs a brief line when a fact check is blocking a publish, or when a catalog change might affect a price you already posted. The helpers stay small and focused so R&D does not become a bot that also writes captions and also closes tickets.

Back to blog