A deep dive for engineers. This is the long one: every design decision, every config line, and every gotcha involved in running an embedding model on one machine and the web app on another — with no service API in between
6 Likes
Interesting post!
One note, I think some sections are duplicated, like “Building block 1: partitioning dependencies with runtime: false”.
Interesting read. I was thinking about building something similar, but I am leaning towards using Oban to keep and manage the jobs across the two machines (both reading from the same database). What is your thought process on avoiding this route?
