Reranking

POST /v1/rerank

Reranking is useful after vector search when you want a stronger relevance model to sort candidate documents.

Request

{
  "model": "bge-reranker-v2-m3",
  "query": "How does usage billing work?",
  "documents": [
    "Usage is deducted from prepaid balance.",
    "Image tasks can return a task ID.",
    "Deployments can be stopped and started."
  ],
  "top_n": 2
}

Response

{
  "results": [
    {"index": 0, "relevance_score": 0.92},
    {"index": 2, "relevance_score": 0.41}
  ]
}