Ask HN: Is there a need for entity-based RAG?
2 comments
Can you link to your solution?
https://tilores.io/ is our original solution
https://tilosearch.com/ is when prompted people to ask us about RAG
https://tilosearch.com/ is when prompted people to ask us about RAG
We recently had a bunch of people contact us asking if they could use our solution as a "source of truth" for LLM RAG. We don't know much about LLM or RAG so have been trying to get up to speed quickly, so wanted to ask the question here - if you work on RAG do you see a use case for a fuzzy search engine for structured data (which is effectively what our solution is), where the underlying data is considered a "source of truth"?
Probably should mention the underlying data is deduplicated and linked (and searched) using rules based on various phonetic, similarity and distance algorithms (including Cosine). We don't use vectors or embeddings in our matching, although we plan to later.
We are just now trying to evaluate whether we should double down on the LLM/RAG space and build a Langchain connector for our solution.