Skip to main content
The Embeddings API converts text into numerical vectors that preserve semantic meaning. Use it when you need retrieval, ranking, deduplication, clustering, or similarity search.

​
Best fit

Use embeddings for:
  • semantic search
  • retrieval pipelines
  • reranking and similarity matching
  • clustering and taxonomy work
  • dataset deduplication and nearest-neighbor lookup

​
Request shape

The direct API uses the standard OpenAI-compatible embeddings endpoint:
  • endpoint: POST /v1/embeddings
  • auth: Authorization: Bearer $INFERENCE_API_KEY
  • input: a string or an array of strings

​
Operational guidance

  • use realtime embeddings for synchronous application flows
  • use background or batch paths when you need to process very large corpora
  • keep the same model across index creation and query time unless you plan a full reindex

​
Common choices

  • Single text input for interactive similarity lookups
  • Array input when you want one request to generate multiple vectors
  • Batch API when you need to process a large offline corpus