Find answers from the community

Updated 2 months ago

I m using MIlvus Vector Store as my

I'm using MIlvus Vector Store as my backend, wondering if I'm using the vecstoreindex implementation by llama index, how can I better utilize the metadata I fetched from the nodes.
L
M
2 comments
Looks like the milvus vector store doesn't support anything too interesting (no hybrid, no metadata filtering) https://github.com/jerryjliu/llama_index/blob/9a374d3f0e742ececf32ee2a5ab14d1d5d50992f/llama_index/vector_stores/milvus.py#L423

Although more generally right now, your metadata is being used when calculating embeddings for a node. The LLM also sees the metadata by default

You can configure on each document/node which metadata is used for embeddings and for the LLM
Oh I figure. I think I can play with the prompt to put the spotlight to the metadata, hopefully.
Add a reply
Sign up and join the conversation on Discord