Files
zeekay edb6963531 Plans, prices and the model table are read from api.hanzo.ai when a page opens, and no page carries a copy
The root layout handed every page the plan catalog (lib/rooms.ts configure
with lib/data/plans.json), and /pricing, /calculator, /customers and the
benchmark pages bundled the 202 KB price list. lib/api is now the one reader:
plans() GET /v1/billing/plans, pricing() /v1/pricing, summary(), served()
/v1/models, templates(), blueprints() /v1/blueprint/sbom, one promise a path a
page load, each drawn as loading, ready or "api.hanzo.ai is unreachable"
(components/api) and never an older number. plans.json, pricing.json,
models.json, templates.json, blueprints.json and sync-pricing, sync-models,
sync-templates, sync-blueprints are gone.

What a static export must know before anyone asks, which model pages exist and
their titles, is read by the build itself (lib/api/build, Next's fetch cache
keyed by the build's stamp: one answer per build) and renders no rate.
lib/moved reads the retired model pages against the export, and the 153 pages
the old snapshot kept for models the gateway dropped are in retired.json.
Kai's price beside Jev's, Enso's tier rates, a model page's rate and the GPU
table read the price list in the browser; the site's plan math no longer
imports @hanzo/rooms/lib/plans, whose bundled @hanzo/plans catalog rode the
home, /pricing and the picker.

Home first load 452.7 -> 437.4 KB gzip (budget 440); pages over
perf.budget.json 32 -> 9, the rest over for settings code, connectors and the
commerce catalog. The gates serve invented answers (e2e/gates/api) and block
them; e2e/gates/read.spec.ts holds that no script carries a plan, price or
model table.
2026-10-15 11:07:56 -07:00

21 lines
1.0 KiB
JavaScript

#!/usr/bin/env node
// The sync scripts rewrite their snapshots on every build and stamp each with the
// time it was taken. The stamp is bundled, so an unchanged snapshot still renames
// its chunk, which renames the webpack runtime and with it every page: a one-page
// change then publishes the whole site. A snapshot whose data equals the committed
// one, stamps aside, goes back to the committed file.
import { execFileSync } from 'node:child_process'
import { readFileSync } from 'node:fs'
const git = (...args) => execFileSync('git', args, { encoding: 'utf8', maxBuffer: 1 << 28 })
const strip = (text) => JSON.stringify(JSON.parse(text), (k, v) => (k === 'fetched' ? undefined : v))
let kept = 0
for (const file of git('diff', '--name-only', '--diff-filter=M', '--', 'lib/data', 'data').split('\n').filter((f) => f.endsWith('.json'))) {
if (strip(git('show', `HEAD:${file}`)) === strip(readFileSync(file, 'utf8'))) {
git('checkout', 'HEAD', '--', file)
kept++
}
}
console.log(`keep-stamps: ${kept} unchanged snapshot(s) restored`)