Skip to content

Azure OpenAI Setup→

Azure has its own implementation of OpenAI which has stronger privacy guarantees and has been vetted by general counsels and CISOs. We prefer this implementation as well over OpenAI direct access.

  1. Create an Azure OpenAI GPT endpoint using the instructions on Azure

    • Resource Group: Choose a new resource group isolated from other resource groups for security, privacy, and billing

    • Region: Choose a region where Hasper server is deployed and/or where the models recommended below are present.

    • Network: Use the private endpoints network configuration if the usecase risk is high. Otherwise, you can use all networks.

    • Deploy Models: Select the following models - GPT 4o (2024-05-13 or 2024-08-06), GPT 4o-mini (2024-07-18 or 2024-08-06), text-embedding-ada-002 (2). 4o and 4o-mini will be the same end point and the text embedding will be another endpoint.

  2. Share the following with Scribble/Operations team along with chosen network cofiguration.

    export AZURE_OPENAI_KEY="<key>"
export AZURE_OPENAI_ENDPOINT="https://<endpoint>.openai.azure.com/"
export AZURE_OPENAI_DEPLOYMENT_NAME_GPT="<deployment-name>"
export AZURE_OPENAI_DEPLOYMENT_NAME_EMBEDDING="<embedding-endpoint-name>"