Azure OpenAI config
/my-light-module/decorations/ai-accelerator-azureai/config.yamlapiKey: <your-azure-key>
azureDeployment: <your-azure-deployment>
azureResourceName: <your-azure-resource-name>
apiKey: /path/to/apikeyFor more, see Passwords app.
Using multiple Azure deployments
Azure AI allows you to use multiple model deployments in parallel, even with different endpoints, by creating custom REST client configurations. This is useful when you want to:
-
Use different Azure OpenAI deployments for different models (e.g., GPT-4 in one deployment, DALL-E in another)
-
Connect to different Azure regions or resource names
-
Use separate API keys for different deployments
Custom REST client configurations for model deployment
Create a custom REST client in your light module to connect to a specific Azure OpenAI deployment.
Here is an example for DALL-E-3:
/my-light-module/restClients/azureAI-dalle3.yamlbaseUrl: https://my-resource.cognitiveservices.azure.com/openai/deployments/dall-e-3 (1)
timeoutConfiguration:
readTimeout: 180
connectTimeout: 180
components:
audit:
class: info.magnolia.ai.rest.RestClientAudit
header-filter:
class: info.magnolia.ai.rest.ResponseHeaderFilter
restCalls:
imageGeneration:
method: POST
entityClass: com.fasterxml.jackson.databind.JsonNode
path: /images/generations
body: "{body}"
queryParameters:
api-version: 2024-08-01-preview
headers:
Content-Type: application/json
api-key: !env ${AZURE_AI_API_KEY} (2)| 1 | Full Azure OpenAI endpoint URL including the deployment name |
| 2 | API key using environment variable (can also use !secret for Passwords app) |
Use custom REST client in model definition
Reference your custom REST client in your model definition by setting the restClientId property.
Here, the custom DALL-E-3 REST client is referenced:
/my-light-module/aiModels/azure-dalle-3.yaml$type: azureImageModel
modelName: Azure OpenAI DALL-E 3
modelVersion: dall-e-3
restClientId: azureAI-dalle3 (1)
input:
prompt:
$type: prompt
required: true
description: A text description of the desired image(s).
size:
$type: enum
description: The size of the generated images.
enumValues: [1024x1024, 1792x1024, 1024x1792]
defaultValue: 1024x1024
quality:
$type: enum
enumValues: [standard, hd]
defaultValue: standard
style:
$type: enum
enumValues: [vivid, natural]
defaultValue: vivid
output:
data:
$type: list
itemType: imageUrl| 1 | Reference your custom REST client by id. |