Use Self-RAG with vector databases
casper-hansen opened this issue · comments
The ability to retrieve from a vector database is often how RAG is used in applications. Could you please look into adding langchain or llamaindex support on top to leverage e.g. a Pinecone, Milvus, or Chroma vector store?
I am preparing an out of the box solution if you are interested
I am preparing an out of the box solution if you are interested
Definitely interested!
Awesome, I just put a comment here about what I am working on regarding this.
Would love to hear your thoughts in that thread
Sorry for my late response, I am currently not plannig to integrate vector DB such as Chroma -- I tried once and found the indexing process was quite slow for 30M passage scale as in the entire English Wikipedia but I might have tried some techniques to enhance the efficiency. The solution by @emrgnt-cmplxty looks really nice, and recommend to check it out!