Embeddings and reranking

Embeddings and reranking use different models. Copy the corresponding model name from the Pricing page.

Create embeddings

POST /v1/embeddings
Authorization: Bearer OMROUTER_API_KEY
Content-Type: application/json

curl --silent --show-error \
  'https://omrouter.com/v1/embeddings' \
  --header "Authorization: Bearer $OMROUTER_API_KEY" \
  --header 'Content-Type: application/json' \
  --data "{
    \"model\": \"$OMROUTER_MODEL_ID\",
    \"input\": [\"First document\", \"Second document\"]
  }"

The response contains one vector per input:

{
  "object": "list",
  "data": [
    {"object": "embedding", "index": 0, "embedding": [0.0]}
  ],
  "model": "model-id",
  "usage": {}
}

Optional fields include encoding_format and dimensions where the selected model supports them.

Rerank documents

POST /v1/rerank
Authorization: Bearer OMROUTER_API_KEY
Content-Type: application/json

curl --silent --show-error \
  'https://omrouter.com/v1/rerank' \
  --header "Authorization: Bearer $OMROUTER_API_KEY" \
  --header 'Content-Type: application/json' \
  --data "{
    \"model\": \"$OMROUTER_MODEL_ID\",
    \"query\": \"Which document explains authentication?\",
    \"documents\": [
      \"Use an API key in the Authorization header.\",
      \"Video jobs are asynchronous.\"
    ],
    \"top_n\": 1
  }"

{
  "results": [
    {"index": 0, "relevance_score": 0.0}
  ],
  "usage": {}
}

Set return_documents: true when the result should include the original text. Large document batches increase response time and cost, so split them according to the selected model’s limits.