Reranker NVIDIA icon

Reranker NVIDIA

Use Nvidia Reranker to reorder documents after retrieval from a vector store by relevance to the given query.

Reranker NVIDIA

AI-generated

Summary

Rerank a set of documents by relevance to a given query using a specified Nvidia model, returning the top N most relevant documents with added relevance scores.

Inputs

  • Model (required) — The Nvidia model to use for reranking the documents.
  • Top N — The maximum number of documents to return after reranking.

Output shape

a list of documents reordered by relevance, each with a metadata field 'relevanceScore' indicating its computed relevance score.

The output includes up to 'Top N' documents ranked by relevance to the input query. If fewer documents are returned by the service, only those are output. Each document includes a 'relevanceScore' in its metadata.

Examples

Example 1: Given an array of documents and a query, the node reorders documents based on their relevance to the query, attaching a relevance score to each.

Set 'Model' to the desired Nvidia reranker model (default is 'nv-rerank-qa-mistral-4b:1') and optionally adjust 'Top N' to limit the number of returned documents after reranking.

Discussion