CohereReranker

A reranker that uses Cohere's reranking API to score and reorder documents based on query relevance.

Installation

pip install datapizza-ai-rerankers-cohere

datapizza.modules.rerankers.cohere.CohereReranker

Bases: Reranker

A reranker that uses the Cohere API to rerank documents.

__init__

__init__(
    api_key,
    endpoint,
    top_n=10,
    threshold=None,
    model="model",
)

Parameters:

Name	Type	Description	Default
api_key	str	The API key for the Cohere API.	required
endpoint	str	The endpoint for the Cohere API.	required
top_n	int	The number of documents to return.	10
threshold	float | None	The threshold for the reranker.	None

a_rerank async

a_rerank(query, documents)

Rerank documents based on query.

Parameters:

Name	Type	Description	Default
query	str	The query to rerank documents by.	required
documents	list[Chunk]	The documents to rerank.	required

Returns:

Type	Description
list[Chunk]	The reranked documents.

rerank

rerank(query, documents)

Rerank documents based on query.

Parameters:

Name	Type	Description	Default
query	str	The query to rerank documents by.	required
documents	list[Chunk]	The documents to rerank.	required

Returns:

Type	Description
list[Chunk]	The reranked documents.

Usage

from datapizza.modules.rerankers.cohere import CohereReranker

reranker = CohereReranker(
    api_key="your-cohere-api-key",
    endpoint="https://api.cohere.ai/v1",
    top_n=10,
    threshold=0.5,
    model="rerank-v3.5",
)

# Rerank chunks based on query
query = "What are the benefits of machine learning?"
reranked_chunks = reranker.rerank(query, chunks)

Features

• High-quality semantic reranking using Cohere's models
• Configurable result count and score thresholds
• Support for both sync and async processing
• Automatic relevance scoring for retrieved content
• Integration with Cohere's latest reranking models
• Flexible endpoint configuration for different Cohere services

Examples

Basic Usage

import uuid

from datapizza.modules.rerankers.cohere import CohereReranker
from datapizza.type import Chunk

# Initialize reranker
reranker = CohereReranker(
    api_key="COHERE_API_KEY",
    endpoint="https://api.cohere.ai/v1",
    top_n=5,
    threshold=0.6,
    model="rerank-v3.5",
)

# Sample retrieved chunks
chunks = [
    Chunk(id=str(uuid.uuid4()), text="Machine learning enables computers to learn from data..."),
    Chunk(id=str(uuid.uuid4()), text="Deep learning is a subset of machine learning..."),
    Chunk(id=str(uuid.uuid4()), text="Neural networks consist of interconnected nodes..."),
    Chunk(id=str(uuid.uuid4()), text="Supervised learning uses labeled training data..."),
    Chunk(id=str(uuid.uuid4()), text="The weather forecast shows rain tomorrow...")
]

query = "What is deep learning and how does it work?"

# Rerank based on relevance to query
reranked_chunks = reranker.rerank(query, chunks)

# Display results with scores
for i, chunk in enumerate(reranked_chunks):
    score = chunk.metadata.get('relevance_score', 'N/A')
    print(f"Rank {i+1} (Score: {score}): {chunk.text[:80]}...")