Azure OpenAI Connection to SAS Retrieval Agent Manager

Introduction

Azure OpenAI is an LLM provider in Azure that you will need to allow connectivity from the SAS Retrieval Agent Manager cluster for it to work properly. Once connection is allowed, you can connect models to query from SAS Retrieval Agent Manager.

Allowing a Connection

To allow a connection between SAS Retrieval Agent Manager and your Azure OpenAI resource, the easiest way to do it, is by completing the following steps:

Retrieve egress IP from your API pod

You can exec into the API pod and perform a curl ip.me to retrieve the egress IP to whitelist on the Azure OpenAI resource. This can be done using the following commands:


kubectl exec -it <ram_api_pod> -n retagentmgr -- curl ip.me

# Example output: 15.124.136.98

Whitelist the egress IP on the Azure OpenAI Resource

After navigating to the homepage of your Azure OpenAI resource, go to the Networking tab and add the IP from the previous command under the Firewall Address Range section.

Connection

To connect SAS Retrieval Agent Manager to the Azure OpenAI Resource after you’ve added the IP address to the firewall whitelist, follow these steps:

  • Open the Settings panel in the top right of the UI
  • Select the Azure OpenAI model type
  • Insert the desired name and deployment that you have deployed in Azure OpenAI

  • Insert your Azure OpenAI endpoint into the URI field

    Example: https://my-azure-OpenAI-resource.azure.com/

  • Insert the API key in the API Key field
  • Select which groups will have access to this LLM

At this point it should be ready to be queried by the user via agents or chat.