Cohere Reranker

Overview

Cohere Reranker is an advanced result re-ranking service that improves knowledge retrieval quality by intelligently reordering search results based on semantic relevance to user queries. This service uses Cohere's specialized re-ranking models to evaluate and score knowledge segments, filtering large result sets down to the most pertinent information for RAG applications.

Service Information

Property	Value
Service Name	Cohere reranker
Status	Enabled
Compatible Nodes	Re-rank search

When to Use This Service

Cohere Reranker is ideal for:

RAG quality improvement where search result relevance is critical
Multi-source knowledge fusion combining results from different knowledge databases
Large result set filtering reducing overwhelming knowledge sets to manageable subsets
Production RAG applications requiring consistent, high-quality knowledge retrieval
Multilingual applications needing cross-language re-ranking capabilities

API Requirements

Cohere → API Key: Valid Cohere API key with re-ranking API access

Parameters

Parameter	Type	Default	Description
Reranking model	Choice	(varies)	Select Cohere re-ranking model for result evaluation
Top n	Number	5	Number of top-ranked results to return after re-ranking

Available models depend on your Cohere API access. Check the Reranking model dropdown for currently available options.

How It Works

Knowledge aggregation: Combines input knowledge from multiple sources
Relevance scoring: Evaluates each segment against user prompt using Cohere's re-ranking model
Result filtering: Returns top n most relevant segments based on re-ranking scores