Skip to content

2.3 Select An Azure Region

To ensure you can successfully deploy the Azure resources using the azd up command, you must choose a region that supports the required Azure OpenAI gpt-4o and text-embedding-ada-002 models, has at least 10K TPM of Standard capacity available for each of those models, and supports the Language service's abstractive summarization capability. On completing this step, you should have:

  • Selected an Azure region for workshop resources

Review regional availability and Azure OpenAI quotas

There are only a handful of Azure regions that support all of the required resources and capabilities needed to successfully complete this workshop.

Not selecting an appropriate region will result in a deployment failure!

Follow the instructions below to review regional availability of the required services, models, and capabilities, and then select one of those for your deployment.

  1. Review the regional availability guidance for the gpt-4o and text-embedding-ada-002 models in Azure OpenAI.

  2. Ensure you have a at least 10K TPMs of Standard capacity available in the region for both the gpt-4o and text-embedding-ada-002 models. Follow these instructions to check your available quota.

  3. Check the text abstractive summarization regional availability

Select an Azure region that supports workshop resources

  1. Based on the regions that meet the requirements from your review above, choose a region that supports both Azure OpenAI models, has adequate Standard TPM capacity, and supports Text Abstractive Summarization.

Select a region that supports both Azure OpenAI models!

  • Choosing a region that doesn't support both Azure OpenAI models will result in deployment failure when running azd up.

  • Selecting a region that does not have at least 10K TPM capacity for both the gpt-4o and text-embedding-ada-002 models will result in a deployment failure when running azd up.

  • Selecting a region that does not support abstractive summarization will not cause a deployment failure, but you will need to make code changes later in the workshop to use extractive summarization in its place.