I got a RAG app running on a 512 MB free server. The trick was not running the embedding model at all.

· Dev.to

Read full story at source