AI Image Generation¶

Trigger: HTTP | State: stateless | Guarantee: request-response | Difficulty: intermediate | Showcase: Azure OpenAI image generation

Overview¶

This recipe exposes an HTTP endpoint that sends a prompt to Azure OpenAI image generation and returns the resulting image URL.

The sample uses the openai Python SDK against an Azure endpoint and keeps the Function app small enough to show the core pattern clearly. As with the other cookbook HTTP recipes, it calls out @with_context, @openapi, and @validate_http, and uses azure-functions-logging-python for structured telemetry.

When to Use¶

You want a serverless API that turns text prompts into generated images.
You need a simple backend wrapper around Azure OpenAI image generation.
You want request validation and OpenAPI documentation for a media-generation endpoint.

When NOT to Use¶

You need streaming text tokens rather than an image URL.
You need durable, multi-step asset workflows after generation completes.
You need advanced media pipelines such as moderation, resizing, or storage processing.

Architecture¶

flowchart LR
    A[Client] --> B[HTTP trigger\nPOST /api/images/generate]
    B --> C[@with_context + @openapi + @validate_http]
    C --> D[Azure OpenAI image generation]
    D --> E[Image URL + prompt metadata]
    E --> A

Behavior¶

The sequence below shows the runtime interaction between components.

sequenceDiagram
    participant Client
    participant Function as Azure Function
    participant AOAI as Azure OpenAI

    Client->>Function: POST /api/images/generate { prompt, size }
    Function->>AOAI: images.generate(...)
    AOAI-->>Function: image URL
    Function-->>Client: 200 JSON response

Prerequisites¶

Python 3.10+
Azure Functions Core Tools v4
openai SDK
Azure OpenAI resource with a DALL-E deployment

Project Structure¶

examples/ai-and-agents/ai_image_generation/
|- function_app.py
|- host.json
|- local.settings.json.example
|- pyproject.toml
`- README.md

Implementation¶

The example project is examples/ai-and-agents/ai_image_generation/.

function_app.py defines typed request and response models, configures azure-functions-logging-python, and creates an Azure OpenAI client with AZURE_OPENAI_ENDPOINT, AZURE_OPENAI_KEY, and AZURE_OPENAI_IMAGE_DEPLOYMENT.

The route uses the standard cookbook HTTP decorator order:

@app.route(route="images/generate", methods=["POST"])
@with_context
@openapi(summary="Generate an image", request_body=ImageRequest, response={200: ImageResponse}, tags=["ai"])
@validate_http(body=ImageRequest, response_model=ImageResponse)
def generate_image(req: func.HttpRequest, body: ImageRequest) -> func.HttpResponse:
    ...

Generation itself happens through the Azure OpenAI image API in the openai SDK:

result = client.images.generate(
    model=os.getenv("AZURE_OPENAI_IMAGE_DEPLOYMENT", "dall-e-3"),
    prompt=body.prompt,
    size=body.size,
)

The function returns the URL plus the revised prompt so callers can trace what the model actually used.

Run Locally¶

cd examples/ai-and-agents/ai_image_generation
pip install -e ".[dev]"
cp local.settings.json.example local.settings.json
func start

Expected Output¶

Functions:

    generate_image: [POST] http://localhost:7071/api/images/generate

Example request:

curl -X POST http://localhost:7071/api/images/generate \
  -H "Content-Type: application/json" \
  -d '{"prompt": "A futuristic serverless control room in watercolor style", "size": "1024x1024"}'

Example response:

{
  "image_url": "https://example.blob.core.windows.net/generated/serverless-control-room.png",
  "revised_prompt": "A futuristic serverless control room in watercolor style",
  "deployment": "dall-e-3"
}

Production Considerations¶

Add authentication, prompt moderation, and request quotas before exposing the API broadly.
Consider storing generated images in Blob Storage instead of returning provider URLs directly.
Track prompt size, image size, latency, and deployment name with azure-functions-logging-python.
Protect API keys with managed identity and Key Vault-backed configuration.