AICentral.AzureAISearchVectorizer 0.21.0

AICentral.AzureAISearchVectorizer

A customer vectorizer endpoint enabling Azure AI Search to calculate embeddings using end to end private endpoints, in an architecture leveraging Azure API Management's AI Gateway.

Configuration

dotnet add package AICentral.AzureAISearchVectorizer

builder.Services.AddAICentral(
 builder.Configuration,
 additionalComponentAssemblies:
 [
 typeof(AzureAISearchVectorizerProxy).Assembly,
 ]);

{
 "AICentral": {
 "RouteProxies": [
 {
 "Type": "AzureAISearchVectorizerProxy",
 "Name": "azureAISearchVectorizer",
 "Properties": {
 "EmbeddingsDeploymentName": "embeddings",
 "ProxyPath": "/aisearchembeddings",
 "OpenAIApiVersion": "2024-04-01-preview"
 }
 }
 ],
 "Pipelines": [
 {
 "Name": "gpt4o",
 "Host": "*",
 "EndpointSelector": "endpoint",
 "AuthProvider": "anonymous",
 "RouteProxies": ["azureAISearchVectorizer"],
 "Steps": [
 ]
 }
 ]
 }
}

How does it work?

The middleware exposes a new AI Central Endpoint for a pipeline at the '/ProxyPath' URL, authorised the same was as the rest of the pipeline. This endpoint can be provided to an Azure AI Search instance as a custom vectorizer endpoint to calculate embeddings on AI Search requests.