Skip to main content

Azure AI Model Inference API

Azure AI uses the Azure AI Model Inference API exclusively. Legacy Azure OpenAI deployment endpoints (/openai/deployments/{name}?api-version=XXX) are no longer supported.
API VersionEndpoint FormatUsage
v1 (default){endpoint}/openai/v1Stable — model specified in the request body
preview{endpoint}/openai/v1?api-version=previewPreview features — model specified in the request body
Key differences from the legacy Azure OpenAI format:
  • Model in body — The model/deployment name is sent in the request body (model field), not in the URL path
  • No deployment URLs — No /openai/deployments/{name} paths
  • No dated versions — No 2024-10-21 or 2025-04-01-preview style version strings
  • Authentication — Uses api-key header (API key) or Authorization: Bearer (Managed Identity), same as before

Embeddings

Text embedding models use AzureOpenAIEmbeddings from @langchain/openai, which still requires a dated API version (e.g., 2024-06-01) for the Azure OpenAI Embeddings API. This is configured per embedding endpoint in Admin → AI Search Endpoints ([#/admin/search]).