Interesting approach to RAG and vector db for long context

[hrstoyanov]

Please enable discussions for this GitHub project, the below is not really an issue:

https://www.zyphra.com/post/reaching-1b-context-length-with-rag

[jbellis]

sorry, what does this have to do with jvector?

[hrstoyanov]

Further, we go beyond prior applications of RAG to long-context tasks (e.g., [here](https://arxiv.org/abs/2409.01666)) by focusing on generating vector databases for retrieval quickly at test time, while simultaneously dealing with complex reasoning tasks, like Hash Hop, through the use of a novel, sparse graph-based retrieval method.