[Feature]: embedding/rerank module for rag

[feihongxu0824]

Problem / Motivation

Modern RAG pipelines typically use a two-stage approach: coarse retrieval via embeddings followed by fine-grained reranking. Users expect zvec not only to store and retrieve vectors efficiently but also to integrate smoothly with popular embedding and reranker models (e.g., BGE, Qwen-Reranker). A reference implementation will demonstrate best practices and accelerate adoption in production RAG systems.

Proposed Solution

Alternatives Considered

No response

Affected Area

No response

[tomaarsen]

I totally agree. I would love to see examples with embedding model -> CrossEncoder reranker. Currently, a lot of your examples/docs just use example vectors, but examples with real models would be quite valuable. Perhaps also with Sparse vectors, like https://huggingface.co/naver/splade-v3

Very nice work here, I'm going to experiment with this soon!

• Tom Aarsen