PlukoPluko

The problem

Products that need images, like catalogues, content tools and AI apps, usually end up with fragile scraping scripts, broken links and no record of where an image came from. We wanted one small, reliable API that returns clean results with their sources, so the people using it can check how an image may be reused.

The system

A Next.js workspace and a Node API share one repository and a SQLite database. Searches run as background jobs in a separate worker, so a large request never holds a connection open, and finished jobs can notify your server through signed webhooks. Pluko stores URLs and metadata, never the image files. It runs in Docker behind Caddy, and every push to main is tested, deployed and rolled back automatically if the new version isn't healthy.

What we learned

Designing an API that stays backwards compatible from the first version, running long work as jobs instead of long requests, and building a deployment pipeline that can undo itself. It's also our first open-source project, under the AGPL.