Skip to content

Embeddings

Convert text into vectors for semantic search, recommendations, clustering and RAG.

Updated on Aug 09, 2026

Embeddings are numerical representations of text. Semantically related inputs tend to produce nearby vectors, which enables semantic search, recommendations, clustering, duplicate detection and context retrieval for RAG.

Choose an embeddings model

Availability depends on your account. Check GET https://api.hinow.ai/v1/models and pick an item whose endpoint is /v1/embeddings. Use the returned id in the model field; do not automatically reuse a Chat Completions model.

POSThttps://api.hinow.ai/v1/embeddingsBearer

Converts text into vectors, for semantic search and clustering.

Parâmetros

  • modelstring· bodyobrigatório

    Id of a model whose endpoint is `/v1/embeddings`.

  • inputstring | string[]· bodyobrigatório

    Text or list of texts to be converted into vectors.

Respostas

200List of vectors in the same order as `input`
{
  "object": "list",
  "data": [
    {
      "object": "embedding",
      "index": 0,
      "embedding": [0.0123, -0.0042, 0.0311, "..."]
    }
  ],
  "usage": { "prompt_tokens": 8, "total_tokens": 8 }
}

Create an embedding

curl https://api.hinow.ai/v1/embeddings \
  -H "Authorization: Bearer $HINOW_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "SEU_MODELO_DE_EMBEDDINGS",
    "input": [
      "Como redefinir minha senha?",
      "Não consigo acessar minha conta"
    ]
  }'

Send batches when you can

input accepts a list and the response preserves the order through index. Batches reduce network requests; respect the input and size limits published for the model you chose.

Best practices

  • Use the same model to index documents and to query the index.
  • Store the embedding identifier and dimension alongside the index to avoid mixing incompatible vectors.
  • Split documents into chunks that preserve a unit of meaning and keep the source metadata.
  • Evaluate retrieval with real questions, measuring whether the relevant chunks appear among the first results.
Was this page helpful?