This documentation contains work-in-progress information for future Elastic Stack and Cloud releases. Use the version selector to view supported release docs. It also contains some Elastic Cloud serverless information. Check out our serverless docs for more details.
« Large language model performance matrix Connect to Amazon Bedrock »
Elastic Docs ›Elastic Security Solution [8.16] ›AI for security ›Set up connectors for large language models (LLM)

Connect to Azure OpenAI

edit

Connect to Azure OpenAI

edit

This page provides step-by-step instructions for setting up an Azure OpenAI connector for the first time. This connector type enables you to leverage large language models (LLMs) within Kibana. You’ll first need to configure Azure, then configure the connector in Kibana.

Configure Azure

edit
Configure a deployment
edit

First, set up an Azure OpenAI deployment:

  1. Log in to the Azure console and search for Azure OpenAI.
  2. In Azure AI services, select Create.
  3. For the Project Details, select your subscription and resource group. If you don’t have a resource group, select Create new to make one.
  4. For Instance Details, select the desired region and specify a name, such as example-deployment-openai.
  5. Select the Standard pricing tier, then click Next.
  6. Configure your network settings, click Next, optionally add tags, then click Next.
  7. Review your deployment settings, then click Create. When complete, select Go to resource.

The following video demonstrates these steps.


Configure keys
edit

Next, create access keys for the deployment:

  1. From within your Azure OpenAI deployment, select Click here to manage keys.
  2. Store your keys in a secure location.

The following video demonstrates these steps.


Configure a model
edit

Now, set up the Azure OpenAI model:

  1. From within your Azure OpenAI deployment, select Model deployments, then click Manage deployments.
  2. On the Deployments page, select Create new deployment.
  3. Under Select a model, choose gpt-4o or gpt-4 turbo.

  4. Set the model version to "Auto-update to default".

    The models available to you depend on region availability. For best results, use GPT-4o 2024-05-13 with the maximum Tokens-Per-Minute (TPM) capacity. For more information on how different models perform for different tasks, refer to the Large language model performance matrix.

  5. Under Deployment type, select Standard.
  6. Name your deployment.
  7. Slide the Tokens per Minute Rate Limit to the maximum. The following example supports 80,000 TPM, but other regions might support higher limits.
  8. Click Create.

The following video demonstrates these steps.


Configure Elastic AI Assistant

edit

Finally, configure the connector in Kibana:

  1. Log in to Kibana.
  2. Go to Stack Management → Connectors → Create connector → OpenAI.
  3. Give your connector a name to help you keep track of different models, such as Azure OpenAI (GPT-4 Turbo v. 0125).
  4. For Select an OpenAI provider, choose Azure OpenAI.

  5. Update the URL field. We recommend doing the following:

    1. Navigate to your deployment in Azure AI Studio and select Open in Playground. The Chat playground screen displays.
    2. Select View code, then from the drop-down, change the Sample code to Curl.
    3. Highlight and copy the URL without the quotes, then paste it into the URL field in Kibana.
    4. (Optional) Alternatively, refer to the API documentation to learn how to create the URL manually.
  6. Under API key, enter one of your API keys.
  7. Click Save & test, then click Run.

Your LLM connector is now configured. The following video demonstrates these steps.


« Large language model performance matrix Connect to Amazon Bedrock »