Help with Implementing Reranking Using Ollama (xitao/bge-reranker-v2-m3)

[AryanKarumuri]

Hi techies,

I'm exploring how to implement document reranking using Ollama, and came across the xitao/bge-reranker-v2-m3 model listed on Ollama's official website.

I’m currently hosting all my Ollama models behind a personal API endpoint (self-hosted). While the base embedding and generation models are working fine, I’m not sure how to properly integrate this reranker model into my retrieval pipeline.

Specifically, I'm looking for:

• Examples or documentation on how to query reranker models via the Ollama API

• Expected input/output format (e.g., pairs of query-documents?)

• Whether this reranker model follows a standard prompt or accepts structured input like [query, passage] pairs

If anyone has successfully used a reranker like this with Ollama (especially in combination with LangChain or other RAG tools), I’d really appreciate your guidance or code snippets.

Thanks in advance!