Rerank

Reorder candidate documents by their relevance to a query through a supported reranking model.

Code example

curl https://api.zorveus.com/v1/rerank \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer zrv_your_api_key" \
  -d '{
    "model": "cohere/rerank-v3.5",
    "query": "What is write-ahead logging?",
    "documents": [
      "Write-ahead logging records transactions sequentially before disk writes.",
      "Redis is an in-memory key-value data store.",
      "PostgreSQL uses MVCC for concurrency control."
    ],
    "top_n": 2
  }'

Request and result

Send query, documents, and a routable model. Use top_n to limit returned results when the provider supports it. Each result identifies the original document position and includes a relevance score or provider-equivalent ranking value.

Reranking does not retrieve documents. Retrieve a candidate set first, then rerank only the number of documents that fits the model's limits and your latency budget.

Check GET /v1/models before using the example model. Provider model IDs and response fields can differ.