Domino RAG reranking

Hi, I suggest add a configurable reranking stage to Domino IQ RAG. Today, Domino IQ sends top-N results from the local vector database directly into the LLM context, without an option to rerank.

Proposal: allow an optional reranker step between vector retrieval and context assembly. Configuration per command document: enable or disable, reranker endpoint, for example, OpenAI-compatible, model name, initial top-N and final top-K.

The system must enforce ACL and Readers fields before any content is sent to the reranker or LLM. Value: better precision for enterprise NSF data, less irrelevant context, lower token use, and more production-grade RAG, without customers having to build an external retrieval layer

Vlaad T