Asset Indexing Engine: a product image archive people can actually work from

The problem

The business kept its product imagery in a large, structured storage archive. It was well organised in principle, but nobody could actually work from it day to day. Reviewing product images meant navigating folders by hand, one level at a time, with no way to see across the archive as a whole. So there was no real way to see which of thousands of items had complete images, and gaps in the collection only surfaced when someone happened to go looking for them. Finding the right image was slow as a result, and that slowed down every day-to-day decision that depended on having it in front of you.

The approach

I left the underlying file system as it was and built something on top of the existing archive that people could work from. With the volume involved, indexing had to be automated, so the system needed to walk the archive on its own and nobody would have to keep it current by hand. I designed the ingestion schema (how incoming data would be structured) first, working backwards from the attachment limits of the interface it would feed. I also kept the scope narrow on purpose, to image review and completeness status and nothing beyond that.

What I built

I wrote an indexing engine that walks the full folder hierarchy of the archive itself, so nobody has to point it at specific folders. As it goes, it automatically normalises attachments into JSON arrays (a simple structured list) per asset type, so each product's images arrive in the same format however they were originally filed. That feeds a high-volume ingestion process into Airtable, which handles several attachment fields per record without falling over at that size. On top of that sits a single-table review interface, set up for fast visual scanning and filtering, so reviewing a product's images means glancing at a row.

The outcome

There's now instant visibility of the images across 1,000+ product records, where before there was none. Missing images are quick to spot, and nobody has to go through the archive folder by folder to find the gaps. Even with attachments this complex, ingestion now takes zero manual effort. It was a one-off build, and they're still getting use out of it well after handover.