private preview

Lightweight relevance scoring for retrieval pipelines.

Rerank takes a query and a list of candidate passages and returns them reordered by relevance. Designed for RAG systems, semantic search, and document QA workflows.

POST /v1/rerank
{
  "query": "wireless mesh routing protocols",
  "documents": [
    "OSPF is a link-state routing protocol...",
    "B.A.T.M.A.N. is designed for mesh networks...",
    "BGP defines inter-domain routing..."
  ],
  "top_n": 3
}

• Single endpointOne POST, no SDK required. Pure JSON in, pure JSON out.
• Fast first tokenSub-100ms p50 on warm queries up to 1k candidates.
• Model agnosticWorks alongside any embedding model or vector store.
• StatelessNo corpus upload, no indexing step, no retained data.

Currently in closed preview. Reach out for API access.