Azure OpenAI Service on Bifrost

1) Chat Completions

Primary chat endpoint. Automatically routes to Azure deployment via alias mapping.

curl -X POST http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "api-key: YOUR_AZURE_API_KEY" \
  -d '{
    "model": "azure/gpt-4o",
    "messages": [{"role": "user", "content": "Hello"}],
    "deployment": "my-gpt4-deployment",
    "endpoint": "https://my-org.openai.azure.com"
  }'

2) Responses API

The Responses API is available for both OpenAI and Anthropic models on Azure. Bifrost routes to upstream /openai/v1/responses using the preview API version (distinct from Chat Completions API version).

OpenAI models (GPT-4, etc.): conversion follows OpenAI's Responses API format.

Anthropic models (Claude): instructions becomes a system message; reasoning maps to the thinking structure.

curl -X POST http://localhost:8080/v1/responses \
  -H "Content-Type: application/json" \
  -H "api-key: YOUR_AZURE_API_KEY" \
  -d '{
    "model": "azure/claude-3-sonnet",
    "input": "Hello, how are you?",
    "instructions": "You are a helpful assistant",
    "deployment": "my-claude-deployment",
    "endpoint": "https://my-org.openai.azure.com"
  }'

• Uses /openai/v1/responses with preview API version
• Request body conversions are handled automatically by Bifrost
• Supports raw request body passthrough for advanced use cases

See also OpenAI Responses API (gpt-oss reasoning) and Anthropic Responses API for model-family specifics.

3) Embeddings

Text embeddings via Azure OpenAI Service. Maps to configured deployment.

Embeddings are supported for OpenAI models only (not available for Anthropic models on Azure).

curl -X POST http://localhost:8080/v1/embeddings \
  -H "Content-Type: application/json" \
  -H "api-key: YOUR_AZURE_API_KEY" \
  -d '{
    "model": "text-embedding-3-small",
    "input": ["text to embed"],
    "deployment": "my-embedding-deployment"
  }'

4) Image Generation

Image Generation is supported for OpenAI models on Azure and uses the OpenAI-compatible format. Bifrost routes to upstream /openai/v1/images/generations. Conversion matches OpenAI Image Generation.

• bifrostReq.Model maps to deployment ID; bifrostReq.Prompt passes through
• Additional fields from the request are embedded into the upstream struct (OpenAI-compatible)

curl -X POST http://localhost:8080/v1/images/generations \
  -H "Content-Type: application/json" \
  -H "api-key: YOUR_AZURE_API_KEY" \
  -d '{
    "model": "azure/dall-e-3",
    "prompt": "A sunset over the mountains",
    "size": "1024x1024",
    "n": 1,
    "deployment": "my-image-gen-deployment"
  }'

Response conversion

• Non-streaming: Azure responses unmarshal into BifrostImageGenerationResponse (superset of OpenAI; fields pass through)
• Streaming: Server-Sent Events with the same event types as OpenAI image generation streaming

5) Video Generation

Azure routes video generation to OpenAI's Sora models via the Azure OpenAI-compatible endpoint. Parameters match OpenAI Video Generation. Bifrost upstream path: /openai/v1/videos.

curl -X POST http://localhost:8080/v1/videos \
  -H "Content-Type: application/json" \
  -H "api-key: YOUR_AZURE_API_KEY" \
  -d '{
    "model": "azure/sora",
    "prompt": "A calico cat playing piano on stage"
  }'

6) Files API

Files operations are supported for OpenAI models only (not Anthropic deployments on Azure). Bifrost routes file requests to upstream /openai/v1/files. Files are stored in Azure.

curl -X POST http://localhost:8080/v1/files \
  -H "api-key: YOUR_AZURE_API_KEY" \
  -F "file=@document.pdf" \
  -F "purpose=assistants"

Gateway endpoints:

• POST /v1/files (multipart upload)
• GET /v1/files
• GET /v1/files/{file_id}
• DELETE /v1/files/{file_id}
• GET /v1/files/{file_id}/content

7) List Models

Lists available models and deployments configured for your Azure key. No request parameters are required. Bifrost routes to upstream /openai/v1/models. The response includes model metadata, capabilities, and lifecycle status.

curl http://localhost:8080/v1/models

Example response shape

{
  "data": [
    {
      "id": "gpt-4",
      "object": "model",
      "created": 1687882411,
      "status": "active",
      "lifecycle_status": "stable",
      "capabilities": {
        "chat_completion": true,
        "embeddings": false
      }
    }
  ]
}