Refer to Model Configs for how to set the environment variables for your particular deployment.

If you’re already familiar with Azure OpenAI, feel free to skip to the configuring Danswer section at the end

Why use Azure OpenAI

By far the most requested model integration has been Azure OpenAI. As of Aug 2023, GPT-4 is the most capable LLM available. When combined with the Enterprise grade security and geographical deployment options of Azure, this becomes a very attractive option for using with Danswer.

Set up Azure OpenAI Endpoint

As of Aug 2023, Azure OpenAI access is by application only. This may change moving forward, please check online for the latest information.

Start by creating the Azure OpenAI Resource

  • Search for Azure OpenAI service
  • Click on +Create to create a new Azure OpenAI resource

AzureResource

Getting the Model Endpoint Info

Once the resource is created, visit it in Azure OpenAI Studio Select Chat

AzurePlayground

Then click on View Code to see the connection info which we will configure for Danswer

AzureViewCode

Note the values provided in the sample code (also feel free to try out the code to verify it works)

AzureSample

Configuring Danswer to use Azure OpenAI Endpoint

Following the image above, you would configure Danswer with the following environment variables:

GEN_AI_MODEL_PROVIDER=azure
GEN_AI_MODEL_VERSION=Danswer  # "deployment_name", shown in image with the line # model = "deployment_name"
GEN_AI_API_ENDPOINT=https://danswer.openai.azure.com/
GEN_AI_API_VERSION=2023-02-15-preview
GEN_AI_API_KEY=<your-azure-key>  # found below the code section in the above image

You can also leave the GEN_AI_API_KEY unset and set it later via the UI