Work Software & web

Fixed catalog images and automated Kubernetes deploys

I matched missing catalog images to the shop's artwork, corrected product data and automated deploys from GitHub. Database scripts copy affected tables before changing them. The junior developer worked on the front end.

Client catalog fix and deploy pipeline Artwork files from the shop's storage go to a matcher that links them to products by ID or SKU in the file name, with a curated mapping when no match is found. Image checks re-extract bad photos from PDFs, database migrations back up their tables first, and a GitHub Actions workflow runs a dry run before deploying to Kubernetes. Artwork files shop file storage Matcher ID or SKU in name Image checks re-extract from PDF Migrations backup tables first Deploy dry run first no match: curated mapping GitHub Actions to Kubernetes 276 products matched to artwork
Match artwork, check images, back up tables, then deploy through GitHub Actions.

Results

276products matched to their artwork files
62new products added, with backups first

The problem

Hundreds of products showed no image, others had cut-off or blank photos, and many descriptions were thin. The artwork existed in the shop's file storage, but nothing linked a product to its file. Any fix also had to go through a manual deploy to a production Kubernetes cluster.

What I did

  • Matched missing product images to artwork by ID or SKU in the filename, then supplied mappings for the remaining products.
  • Image checks for blank, cut-off or oversized crops, and re-extraction of the flagged product photos from PDFs.
  • Database changes that copy the affected tables first, then add new products and rewrite descriptions in transactions.
  • A GitHub Actions workflow that deploys to Kubernetes after checking secrets and running a dry run, one deploy at a time.

What I took from it

I put the backup step inside each database-change script. If an update needs to be undone, I have the affected tables from immediately before it ran.

Privacy