Azure OpenAI

Overview

Arnica utilizes Azure OpenAI to provide mitigation code samples for code risks, such as SAST and IaC vulnerabilities.

The integration with Azure OpenAI provides the following benefits:

1. Enterprise-Grade Infrastructure: Azure OpenAI, being a part of Microsoft Azure, offers robust cloud infrastructure, ensuring high scalability, security, and compliance standards suitable for enterprise needs.

2. Extended Support and SLAs: Azure provides extended support and service level agreements (SLAs) that are crucial for businesses and large-scale applications.

3. Customization and Control: Azure OpenAI might offer more customization and control options tailored for business applications, including private deployments and specific compliance needs.

4. Pricing and Billing: With Azure, businesses can get consolidated billing for all Azure services, including Azure OpenAI, which simplifies financial management.

Prerequisites

Deploy service resource

Service resources are required in order to connect to the models they host. Follow Microsoft's guidelines to create and deploy Azure OpenAI service resource.

IP allowlist

In some cases, customers may want to restrict who can access the deployed resources. If needed, use Arnica's IP addresses, as documented in the Ingress traffic section of the On-Premises integrations page.

Installation process

Set up Azure AI Foundry Project

1. Login to Azure AI Foundry and click on + Create new.

2. Select Azure AI Foundry Resource.

Select a Resource Type

1. Name your project. (Note that the name must be globally unique. The default name is your username followed by a random number. We recommend you replace your username with something such as arnica-ai-models- followed by the random numbers to ensure uniqueness. E.g. if the suggested name was yourusername-1234 change it to arnica-ai-models-1234).

Name the Project

1. Review the project configuration settings. Microsoft configures new Foundry projects with defaults optimized for functionality. We recommend keeping these default settings, though you can modify them if your organization has specific requirements.

2. Click Create. Azure will take a few minutes to set up your project.

3. Navigate to the Azure OpenAI tab under Libraries, copy the Azure OpenAI endpoint and API Key. Store these securely. Make sure you've stored the correct endpoint. Using the default Azure AI Foundry endpoint will cause configuration issues.

Store the Endpoint and API Key

Deploy AI Models

1. Navigate to Models + Endpoints on the sidebar.

Navigate to Models + Endpoints

2. Click on `+ Deploy Model`

Deploy a Model

3. Deploy the models you wish to use

1. Search and select a model you wish to use

Deploy a Model

5. Click Deploy

Deploy a Model

6. Repeat this for each model you want to add. We recommend the following models:

Recommended Models

Integrate with Arnica

1. Navigate to the Integrations page in Arnica, scroll down and under Artificial Intelligence, locate Azure OpenAI and click Connect.

2. Fill in the endpoint URL and API key from step 6 above.

3. Under deployment names make sure that only the models you deployed are selected, e.g. if you deployed: gpt-4.1, gpt-5-mini, gpt-4.1-mini and gpt-5-nano then remove (click the x) near gpt-4o, then click Validate.

4. If the process was successful the validation will succeed. Click OK.

5. Select the primary AI model and click Save.

6. Scroll down and ensure the new integration appears under the Existing Integrations list.

User Experience

Arnica allows users to select when to trigger the Azure OpenAI recommendation request, so that the cost will remain relatively low compared to execution on every finding.

To see the recommendation, navigate to the Code Risks page and click on one of the SAST / IaC findings. Click on the AI icon on the top right corner of the details pane - it will spin while the recommendation is generated and validated by Arnica.

If you would like to generate an alternative recommendation, click on the AI icon again.

The code example recommendation will be dynamically generated in the details pane, followed by the explanation of the generated code to ensure the solution is clear as much as possible for the developer or Arnica operator.

Increase Azure OpenAI quota limits

Azure throttles requests once a deployment reaches its token rate limit. We recommend increasing the assigned quota from the Azure AI Foundry quota management page to the maximum allowed quota, subject to the limits available in the Azure subscription and region.

Quota changes must be completed in the customer's Azure environment. Arnica cannot change the customer's Azure OpenAI quota on their behalf. You must have Owner access to the Azure OpenAI deployment or sufficient Azure permissions to edit deployment quotas.

1. Open the Azure AI Foundry quota management page.

2. In the quota list, expand the model used by the Arnica deployment, such as gpt-5-mini.

3. Locate the deployment connected to Arnica.

4. Click the pencil icon for that deployment.

5. Set the quota to the highest value available for the customer's Azure subscription and region.

6. Click Confirm.

7. Refresh the quota management page in the browser.

8. Expand the same model again and confirm that the deployment quota changed.

Azure AI Foundry may not visually confirm the quota update immediately. In some cases, clicking Confirm appears to do nothing though the quota is updated. Refreshing the page will display updates in the quota table.