Embeddings

POST /v1/embeddings mirrors the OpenAI Embeddings API.

resp = client.embeddings.create(
  model="sable-embed-3-small",
  input="Confidential inference for the agent economy.",
)
vector = resp.data[0].embedding

Models

• sable-embed-3-small: 1536 dims, standard tier
• sable-embed-3-large: 3072 dims, standard tier (a closed model can't back a confidential tier)
• sable-embed-nomic: 768 dims, open-weights, standard tier

Input may be a single string or an array of strings. Same Sable privacy contract: the input is sealed on ingress, decrypted only at the egress shim, and never persisted.