Principle:Cohere ai Cohere python Rerank Response Processing
| Field | Value |
|---|---|
| Type | Principle |
| Source | Cohere Python SDK |
| Domain | NLP Information Retrieval Response Parsing |
| Last Updated | 2026-02-15 |
| Implemented By | Implementation:Cohere_ai_Cohere_python_RerankResponse_Model |
Overview
A parsing pattern for extracting ranked document indices and relevance scores from rerank API responses.
Description
Rerank Response Processing extracts the ordered results from rerank API responses. Each result contains an index (mapping back to the original document list position) and a relevance_score (float 0-1, higher = more relevant). The results are pre-sorted by descending relevance. The response also includes id and meta fields for tracking and billing.
Usage
Iterate over response.results after calling rerank(). Use the index to map back to original documents. Use relevance_score for threshold-based filtering.
Theoretical Basis
The index-based reference pattern avoids retransmitting full document text in the response, reducing payload size. The relevance_score normalization to [0,1] enables consistent threshold-based filtering across different queries.