Jina Rerank

The Jina Rerank API reorders a list of documents based on their relevance to a given query. This is particularly useful for improving retrieval quality in RAG (Retrieval-Augmented Generation) pipelines.

Endpoint Info

Item	Value
URL	/v1/rerank
Method	POST
Auth	Bearer Token (API Key)
Content-Type	application/json

Request Parameters

Parameter	Type	Required	Description
model	string	Yes	Rerank model, e.g. jina-reranker-v2-base-multilingual
query	string	Yes	The search query
documents	array	Yes	List of document strings to rerank
top_n	integer	No	Number of top results to return. Default: all
return_documents	boolean	No	Whether to include document text in results. Default true

cURL Examples

Basic Reranking

curl https://api.quickapi.store/v1/rerank \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "jina-reranker-v2-base-multilingual",
    "query": "What is machine learning?",
    "documents": [
      "Machine learning is a subset of artificial intelligence.",
      "The weather is sunny today.",
      "Deep learning uses neural networks with multiple layers.",
      "I like to eat pizza on weekends."
    ],
    "top_n": 3,
    "return_documents": true
  }'

Without Document Text

curl https://api.quickapi.store/v1/rerank \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "jina-reranker-v2-base-multilingual",
    "query": "How to implement authentication?",
    "documents": [
      "Use JWT tokens for stateless authentication.",
      "The capital of France is Paris.",
      "OAuth 2.0 is a standard for access delegation.",
      "Python is a versatile programming language."
    ],
    "return_documents": false
  }'

💡 RAG Pipeline Integration

Use Jina Rerank after your initial retrieval step (e.g., vector search) to reorder the top-K results. This significantly improves the quality of context passed to the LLM.

Response Example

{
  "id": "rerank_abc123",
  "object": "rerank",
  "model": "jina-reranker-v2-base-multilingual",
  "results": [
    {
      "index": 0,
      "document": "Machine learning is a subset of artificial intelligence.",
      "relevance_score": 0.9234
    },
    {
      "index": 2,
      "document": "Deep learning uses neural networks with multiple layers.",
      "relevance_score": 0.8521
    },
    {
      "index": 1,
      "document": "The weather is sunny today.",
      "relevance_score": 0.0234
    }
  ],
  "usage": {
    "prompt_tokens": 45,
    "total_tokens": 45
  }
}

Python Example

from openai import OpenAI

client = OpenAI(
    base_url="https://api.quickapi.store/v1",
    api_key="YOUR_API_KEY"
)

# Basic reranking
response = client.post(
    "/rerank",
    json={
        "model": "jina-reranker-v2-base-multilingual",
        "query": "What is machine learning?",
        "documents": [
            "Machine learning is a subset of artificial intelligence.",
            "The weather is sunny today.",
            "Deep learning uses neural networks with multiple layers.",
            "I like to eat pizza on weekends."
        ],
        "top_n": 3
    }
)

results = response.json()["results"]
for r in results:
    print(f"Score: {r['relevance_score']:.4f} | {r['document']}")

# Using with requests directly
import requests

response = requests.post(
    "https://api.quickapi.store/v1/rerank",
    headers={
        "Authorization": "Bearer YOUR_API_KEY",
        "Content-Type": "application/json"
    },
    json={
        "model": "jina-reranker-v2-base-multilingual",
        "query": "How to implement authentication?",
        "documents": [
            "Use JWT tokens for stateless authentication.",
            "The capital of France is Paris.",
            "OAuth 2.0 is a standard for access delegation."
        ],
        "top_n": 2
    }
)

data = response.json()
for result in data["results"]:
    print(f"Rank {result['index']}: {result['relevance_score']:.4f}")

⚠️ Document Count Limit

The maximum number of documents per request depends on the model. For jina-reranker-v2-base-multilingual, the limit is 1000 documents. For large-scale reranking, process documents in batches.