RAG as Similarity Engine

necromant2005.github.io

2 pointsrmykhajliw3 comments

This isn't RAG, it's using the embedding-based similarity search part of RAG for doing similarity search things.

The use-case presented, finding similar recipes based on ingredients, is interesting. I'm curious how well it works in practice, especially with a generic embedding model like the one used.

Small approach to use RAG for better result in similarity for related content. The best thing is the process of getting vectors are don't only once, then you can reuse them indefinitely.

As obvious it's possible to use local Ollama models instead of ChatGPT , it you have large amount of data. But I love using the same approach even for building linking inside internal documentation, I really helps to keep all the changes updated and related.

Reply on news.ycombinator.com