A repair-shop system that files itself.

Find the right customer, or create them

Copy the machine, date, price, and parts notes

Paste it into their history table

Delete the tracker card so it is not done twice

Problem 01 · Duplicate customers

Notion’s index lags its own writes

Create a customer, then look them up for their next repair in the same batch, and the lookup can miss the profile you just made. It creates a duplicate. I keep an in-memory cache of everyone created or matched during a run, checked before every API query, so a customer is never created twice in one pass.

// Check in-memory live cache first to avoid Notion's
// eventual consistency index race condition
if (liveCustomerCache.has(cacheKey)) {
  return liveCustomerCache.get(cacheKey);
}

Problem 02 · The same person, twice

Messy phone numbers and stray spaces

Phone numbers arrived in inconsistent formats, and one trailing space in a name silently split a customer into two. I normalize before comparing: digits only for phones, trimmed and lowercased for names, so the match key is stable no matter how the data was typed.

function standardizePhone(phone) {
  if (!phone) return '';
  return phone.replace(/\D/g, '');   // digits only
}

const cleanName = name.trim().toLowerCase();
const cacheKey = `${cleanName}_${standardizePhone(phone)}`;

Problem 03 · It runs on live shop data

A dry run before anything is touched

This writes to a real shop’s records, so a mistake is expensive. I built a read-only mode that simulates the entire flow, creating nothing and archiving nothing, using its own cache so multi-repair matching still behaves. Safe to run against live data before committing to changes. The daemon also isolates errors per record, so one bad row never stops the loop.

const isDryRun = process.argv.includes('--dry-run');

if (isDryRun) {
  console.log(`[Dry-Run] Would create profile for "${name}"`);
  return;   // nothing is written, edited, or archived
}