Surprise, your data warehouse can RAG (opens in new tab)

(rainforestqa.com)

18 pointsukd11y ago13 comments

13 comments

11 comments · 5 top-level

I don't know what a RAG is (and apparently it's forbidden to explain). And at this point, I'm afraid to ask.

Taken from the same blog:

"Roughly, RAG is runtime prompt engineering where you build a system to dynamically add relevant things to your prompt before you ask the agent for an answer."

fragmede1y ago

Have you tried asking Google LLM RAG or ChatGPT what RAG in the context of LLMs is?

maciejgryka1y ago

> “Retrieval-Augmented Generation” is nothing more than a fancy way of saying “including helpful information in your LLM prompt.”

29athrowaway1y ago· 2 in thread

RAG is as valuable as the data you can retrieve.

If the amount of data is small you don't need the flexibility of RAG. And if it is irrelevant it will stay irrelevant after found.

maciejgryka1y ago

For sure, it's only worth doing if you actually have so much relevant data that it doesn't fit in the context! This is definitely the case for us for this problem, but it's not universal.

29athrowaway1y ago

To me the RAG hype is just the sudden rediscovery of information retrieval by money hounds that did not care about AI/ML for decades and now are in panic mode due to FOMO.

1 more reply

maciejgryka1y ago· 1 in thread

This is one of the things we learned recently about building production workflows with LLMs. Happy to answer any questions/feedback here <3

ukd1OP1y ago

For me it's nice to see the re-use of existing infra - big query. Personally, I was rooting for Postgres, but your logic of why not makes sense! Great post.

djhn1y ago

Reading this was very valuable. I really appreciate the Vespa mention and introduction to their Multi-Vector HNSW Indexing - I’ve recently thought a lot about how difficult chunking is and this seems like a promising avenue.

rodrigovicuna1y ago

Here to learn more about this. Important for a startup I know.

j / k navigate · click thread line to collapse