Inference

Reranking

Scores documents against a query using a reranking model.#

POST/lmkit/v1/rerank/score

Computes relevance scores for one or more documents against a query. Uses a cross-encoder or embedding-based reranking model to produce sigmoid-normalized similarity scores. Supports batch scoring. If processing exceeds the configured timeout, returns 202 Accepted with a job_id. Poll GET /lmkit/v1/jobs/ for status and results.

Request body

application/json ·

Property	Type	Description
`query`required	string	The query text to score documents against.
`documents`required	string[]	A list of text passages to score against the query.
`model`	string	The reranking model identifier. If not provided, the default reranking model is used.
`normalize`	boolean	If true, scores are normalized to [0,1] range using sigmoid. Defaults to true.
`top_k`	object (int32)	If specified, returns only the top K results sorted by score descending. If not specified, returns all results.

Responses

Status	Type	Description
200		OK
202		Accepted
400		Bad Request

curl -X POST "$LMKIT_ONE_URL/lmkit/v1/rerank/score" \
  -H "Authorization: Bearer $LMKIT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "query": "string",
  "documents": [
    "string"
  ],
  "model": "string",
  "normalize": true,
  "top_k": "string"
}'

Reranks search results with alpha-blended score fusion.#

POST/lmkit/v1/rerank/rerank

Takes caller-supplied search results (each carrying its raw similarity score) and applies a reranking model to produce blended scores. The final score is computed as: (alpha * raw_similarity) + ((1 - alpha) * reranker_score). Results are returned sorted by blended score descending. If processing exceeds the configured timeout, returns 202 Accepted with a job_id. Poll GET /lmkit/v1/jobs/ for status and results.

Request body

application/json ·

Property	Type	Description
`query`required	string	The query text to rerank results against.
`results`required	[]	The search results to rerank. Each must include content and raw_similarity.
`model`	string	The reranking model identifier. If not provided, the default reranking model is used.
`alpha`	object (float)	Blending factor between original similarity and reranker score. 0.0 = original similarity only, 0.5 = equal blend (default), 1.0 = reranker score only.
`top_k`	object (int32)	If specified, returns only the top K results after reranking. If not specified, returns all results.
`normalize`	boolean	If true, reranker scores are normalized using sigmoid before blending. Defaults to true.

Responses

Status	Type	Description
200		OK
202		Accepted
400		Bad Request

curl -X POST "$LMKIT_ONE_URL/lmkit/v1/rerank/rerank" \
  -H "Authorization: Bearer $LMKIT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "query": "string",
  "results": [
    {
      "content": "string",
      "raw_similarity": "string",
      "collection_identifier": "string",
      "section_identifier": "string",
      "metadata": {}
    }
  ],
  "model": "string",
  "alpha": 0.5,
  "top_k": "string",
  "normalize": true
}'