It’s the business site: the AI Expert page, team training, the blog in Spanish and English, the newsletter and lead pages like the Harness Starter one, with double opt-in and its email sequence.
Every push to main runs the tests on GitHub, builds the image, pushes it to GHCR and tells Coolify, which only has to pull it and restart. About five minutes from start to finish, without loading the server.
Lessons learned
Building on the server itself took 8 to 10 minutes at 400% CPU. Now the image is built on GitHub and the server only pulls it. Along the way I found out the server was ARM and the image was being built for x86.