Rerank
Order a list of documents by relevance to a query with nebula-rerank, using a Cohere-compatible request and response.
On this page
POST
/api/v1/llm/v1/rerank Cohere-compatible reranking. Supports
Idempotency-Key.Scope: llm:access
Request
| Field | Type | Notes |
|---|---|---|
model |
string | nebula-rerank (default). |
query |
string, required | What you are looking for. |
documents |
array of strings, required | Up to 256 documents. |
top_n |
integer | Return only the best N. |
return_documents |
boolean | Include each document’s text in the results. Default false. |
curl https://api.breachline.io/api/v1/llm/v1/rerank \
-H "Authorization: Bearer $BREACHLINE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"query": "Which finding affects authentication?",
"documents": [
"Missing rate limit on the password reset endpoint",
"Verbose server banner on the marketing site",
"Session fixation after login"
],
"top_n": 2
}'
Response
{
"id": "rerank-...",
"model": "nebula-rerank",
"results": [
{"index": 2, "relevance_score": 0.91},
{"index": 0, "relevance_score": 0.84}
],
"usage": {"prompt_tokens": 38, "total_tokens": 38}
}
index is the document’s position in your request. Results are ordered from most to least relevant. Rerank is charged on input tokens at the Nebula input rate.