IKEA assembly, step by step: animated, manual-style 3D guides. Scan any paper manual with Together AI.
TypeScript
/**
* Reads the measurements IKEA gives on each product's own page, for every guide we ship, so build.ts can use real
* sizes where the product listings give none (most seats and sofas) and check.ts can catch a drawing that is off.
*
* npm run ikea:measure fetch what is not cached yet, then write measures.json
* npm run ikea:measure -- --from-end only fill the cache, last guide first (a second runner alongside the first)
*
* One page per guide (its first article number), one at a time, at most about two a second. Every page is cached
* under scripts/ikea/cache/ already trimmed to facts, and measures.json is git ignored like facts.json: the page
* title, the measurements, the package count, the short product description and the addresses of its assembly
* instructions. No prices, reviews or images are kept, except the address of the main photo, for side by side
* checks on this machine only.
*/
import { existsSync, mkdirSync, readFileSync, writeFileSync } from "node:fs";
import path from "node:path";
import { metric } from "./us";
const DIR = import.meta.dirname;
const CACHE = path.join(DIR, "cache");
const OUT = path.join(DIR, "measures.json");
const ITEMS = path.join(DIR, "../../src/data/ikea-items.ts");
const UA = "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/130.0 Safari/537.36 bygga-catalog/1.0";
const GAP_MS = 500;
export interface Page {
/** the page's own title: name, type and what makes this variant ("with chaise longue/Knäbäck grey-green") */
title: string;
market: "gb" | "us";
item: string;
/** IKEA's measurement rows as they read ("Seat height": "45 cm"), inches turned into cm */
m: Record<string, string>;
packages?: number;
about: string[];
photo?: string;
/** the product's assembly instructions on ikea.com, one PDF per box that needs building */
manuals?: string[];
}
const decode = (s: string) =>
s
.replace(/"/g, '"')
.replace(/'|'/g, "'")
.replace(/&/g, "&")
.replace(/</g, "<")
.replace(/>/g, ">")
.replace(/ /g, " ")
.trim();
/** The facts on one product page, or null when it isn't a product page. */
function read(html: string, market: "gb" | "us", item: string): Page | null {
const title = html.match(/<title>([^<]*)/)?.[1];
// an article the market doesn't sell lands on a listing page instead of its own
const canonical = html.match(/rel="canonical" href="([^"]*)"/)?.[1] ?? "";
if (!title || !/\/p\//.test(canonical) || !/pipf-measurements/.test(html)) return null;
const m: Record<string, string> = {};
for (const r of html.matchAll(/measurement-name">([^<]*)<\/span><span class="pipf-measurements-tab__measurement-value">([^<]*)</g)) {
const k = decode(r[1]);
let v = decode(r[2]);
if (/"\s*$/.test(v)) v = metric(v) ?? v;
if (!(k in m)) m[k] = v.replace(/\s+/g, " ");
}
const pk = html.match(/This product comes as (\d+) packages?/) ?? html.match(/This product comes as (one|1) package/);
const details = html.slice(html.indexOf('id="product-details"'));
const about = [...details.slice(0, details.indexOf("Designer") > 0 ? details.indexOf("Designer") : 4000).matchAll(/<p[^>]*>([^<]{12,400})<\/p>/g)].map((p) => decode(p[1])).slice(0, 6);
const photo = html.match(/https:\/\/www\.ikea\.com\/[a-z]{2}\/en\/images\/products\/[a-z0-9_-]+__\d+_pe\d+_s5\.jpg/)?.[0];
const manuals = [...new Set([...html.matchAll(/https:\/\/www\.ikea\.com\/[a-z]{2}\/[a-z]{2}\/assembly_instructions\/[a-z0-9_.-]+\.pdf/gi)].map((r) => r[0]))];
return { title: decode(title).replace(/\s+-\s+IKEA.*$/, ""), market, item, m, packages: pk ? (/\d/.test(pk[1]) ? Number(pk[1]) : 1) : undefined, about, photo, manuals };
}
let last = 0;
async function page(market: "gb" | "us", item: string): Promise<Page | null> {
const file = path.join(CACHE, `page-${market}-${item}.json`);
// pages read before the manuals were kept are read again
if (existsSync(file)) {
const cached = JSON.parse(readFileSync(file, "utf8")) as Page | null;
if (cached === null || cached.manuals) return cached;
}
const wait = last + GAP_MS - Date.now();
if (wait > 0) await new Promise((r) => setTimeout(r, wait));
last = Date.now();
for (let attempt = 1; attempt <= 3; attempt++) {
const r = await fetch(`https://www.ikea.com/${market}/en/p/-${item}/`, { headers: { "User-Agent": UA, Accept: "text/html", "Accept-Language": market === "us" ? "en-US" : "en-GB" } }).catch(() => null);
if (r?.status === 404) {
writeFileSync(file, "null");
return null;
}
const html = r?.ok ? await r.text().catch(() => null) : null;
if (html !== null) {
const p = read(html, market, item);
writeFileSync(file, JSON.stringify(p));
return p;
}
console.log(` ${r?.status ?? "network"} on ${market} ${item}, retry ${attempt}`);
await new Promise((res) => setTimeout(res, 3000 * attempt));
}
return null;
}
async function main() {
mkdirSync(CACHE, { recursive: true });
const text = readFileSync(ITEMS, "utf8");
const fromEnd = process.argv.includes("--from-end");
const all = [...text.matchAll(/^([a-z0-9-]+)\|([\d ]+)$/gm)].map((g) => ({ slug: g[1], items: g[2].split(" ") }));
const guides = fromEnd ? all.reverse() : all;
const prev: Record<string, Page> = existsSync(OUT) ? JSON.parse(readFileSync(OUT, "utf8")) : {};
const out: Record<string, Page> = {};
let n = 0;
for (const g of guides) {
let found: Page | null = null;
// the first article that has a page: the UK one, or the US one for products only the US sells
for (const item of g.items.slice(0, 3)) {
found = (await page("gb", item)) ?? (await page("us", item));
if (found) break;
}
if (found) out[g.slug] = found;
// a page that couldn't be read this time keeps what was read last time, if that was a product page
else if (prev[g.slug] && /^[A-ZÅÄÖÜÉØÆ]{2,}/.test(prev[g.slug].title)) out[g.slug] = prev[g.slug];
if (++n % 50 === 0) {
console.log(`${n}/${guides.length}`);
if (!fromEnd) writeFileSync(OUT, JSON.stringify(out, null, 0).replace(/\},"/g, '},\n"'));
}
}
if (fromEnd) return;
writeFileSync(OUT, JSON.stringify(out, null, 0).replace(/\},"/g, '},\n"'));
console.log(`${Object.keys(out).length} of ${guides.length} guides measured`);
}
main().catch((e) => {
console.error(e);
process.exit(1);
});