POST https://api.wylon.cn/v1/rerank

Rerank

Given a query and a list of candidate documents, the rerank model rescores and reorders them by relevance. Often used as the second-stage refiner in RAG retrieval to boost recall quality.

Authorization

Authorizationstring · headerrequired

Bearer token.

Request body

modelstringrequired

Rerank model ID, e.g. BAAI/bge-reranker-v2-m3.

querystringrequired

Query text.

documentsstring[]required

Candidate documents to rerank; up to 1000 per request is recommended.

top_nintegeroptional

Return only the top N results; if unset, returns all sorted by score descending.

return_documentsbooleanfalse

When true, each result includes the original document text for easy consumption.

Response

modelstring

The rerank model ID that actually served this request.

resultsarray

Results sorted by relevance, descending.

indexinteger

Index of the document in the original documents array.

relevance_scorenumber

Relevance score in [0, 1]; higher means more relevant.

documentstringoptional

Returned only when return_documents=true.

usageobject

Token usage for this request.

Example response

{
  "model": "BAAI/bge-reranker-v2-m3",
  "results": [
    { "index": 2, "relevance_score": 0.94 },
    { "index": 0, "relevance_score": 0.78 },
    { "index": 1, "relevance_score": 0.32 }
  ],
  "usage": { "total_tokens": 156 }
}

Example — 400

{
  "error": {
    "type": "invalid_request_error",
    "message": "`documents` must contain at least 1 item."
  }
}