How-to guide

Set up Azure OpenAI

Connect Bristlenose to an Azure OpenAI deployment.

Azure OpenAI runs the analysis pass against a model deployment in your own Azure subscription, so the analysis runs within your Azure resource rather than against a provider's shared endpoint.

Before you start

You need a deployed model in an Azure OpenAI resource, and three values from the Azure portal:

  • The resource endpoint, e.g. https://my-resource.openai.azure.com/
  • An API key for the resource
  • The deployment name you gave the model (this is used in place of a model name)

Add the configuration

macOS app
  1. Open Settings (⌘,) and go to the LLM tab.
  2. Select Azure OpenAI and paste your key into the API Key field.
  3. In the Azure Settings section, fill in the Endpoint, Deployment name, and API version.
Command line

Set the endpoint and deployment as environment variables, then store the key:

export BRISTLENOSE_AZURE_ENDPOINT=https://my-resource.openai.azure.com/
export BRISTLENOSE_AZURE_DEPLOYMENT=my-deployment-name
bristlenose configure azure

The optional BRISTLENOSE_AZURE_API_VERSION defaults to 2024-10-21.

Unlike the other providers, the Azure key is not verified when you store it, because validation also needs the endpoint and deployment. Verify the whole configuration with bristlenose doctor before a run.

Run an analysis with Azure

bristlenose run interviews --llm azure

Troubleshooting

Symptom Cause and fix
doctor reports a missing value All three of endpoint, key, and deployment are required. Check each environment variable is set.
404 or DeploymentNotFound The deployment name must match the name in the Azure portal, not the underlying model name.

For what is sent to the provider, see Cloud or local AI and Privacy and data.

Previous Set up Gemini