eLADY — eDTL so CSV and Sheets become a partner taxonomy in 3-second cycles.

Tokyo, 1999, still a spreadsheet

eLADY runs eLADY GLOBAZONE (pre-owned luxury), UKEROO (genuine auto parts, with Ecoline / Toyota) and LISUTO! (list once, sell everywhere). Member of the International Anti-Counterfeiting Coalition. The catalogue is serious. The inbound data was not: CSV files and live Google Sheets from partners, each in its own shape.

eDTL, on a three-second clock

We built an Extract → Deduplicate → Transform → Load pipeline. Sources are CSV and Google Sheets. The loop is pseudo-realtime: poll every three seconds, apply only the incremental delta, write the partner taxonomy. Not a nightly job. Not a full reload every time a cell changes.

• Extract from CSV and Google Sheets
• Deduplicate before anything hits the partner map
• Transform into the partner taxonomy
• Load on a 3-second cycle, deltas only

Fastify first, then a Go kernel

Version one was Fastify / Node — enough to prove the loop. The compute kernel — taxonomy convert, mapping, pattern-match — moved to Go. About 60% faster on those steps. The Go core now eats tens of thousands of rows per iteration. Fastify stays in the picture; the heavy work does not.