OpenAI embeddings via the REST API — embed

Convenience wrapper that posts to POST /v1/embeddings on the OpenAI API. Requires httr2 (Suggests). Batches requests at batch_size per call. This does NOT wrap ellmer; ellmer does not currently expose an embedding API (chat-only). Use embed_custom() if you prefer to route through a different SDK.

Usage

embed_openai(
  model = "text-embedding-3-small",
  dims = 1536L,
  api_key = Sys.getenv("OPENAI_API_KEY"),
  base_url = "https://api.openai.com/v1",
  batch_size = 64L,
  timeout = 30
)

Arguments

model: Model identifier. Defaults to "text-embedding-3-small" (1536 dims).
dims: Expected dimensionality. Defaults to 1536L which matches text-embedding-3-small. Pass the correct value for other models (e.g. 3072L for text-embedding-3-large).
api_key: OpenAI API key. Defaults to OPENAI_API_KEY env var.
base_url: API base URL (override for Azure OpenAI or proxies).
batch_size: Number of texts per POST.
timeout: Per-request timeout in seconds.

Value

A securecontext_embedder object.

Details

Privacy note: unlike embed_tfidf(), this sends text to OpenAI. Only enable it when that is acceptable for your data classification.

Examples

if (FALSE) { # \dontrun{
emb <- embed_openai()
embed_texts(emb, c("hello world"))
} # }