Skip to main content

Quick start with Azure OpenAI

Availability

Business Plan for the Wren AI Self-Hosted Version.

Prerequisites​

  • You have access to Azure AI Foundry
  • You have setup the following models in Azure AI Foundry:
    • LLM models: gpt-4.1, gpt-4.1-mini
    • Embedding models: text-embedding-3-large
  • Prepare your Azure OpenAI API base and API key
    • You could find API base in Azure AI Foundry under your Azure OpenAI resource:
      • Azure OpenAI API base
    • Your Azure OpenAI API key is also available on the same Keys & Endpoint page in Azure AI Foundry.
  • Prepare your model deployment names, you can find the deployment name in Azure AI Foundry -> Model deployments
    • Model deployments

Steps (Quick setup)​

1. During onboarding, select "Quick setup" and then "Azure OpenAI".​

Azure OpenAI onboarding

2. Enter your Azure OpenAI API base and API key.​

Azure OpenAI API key

3. Add an embedding model.​

Under Review & configure → Embedding model, click Add embedding model. Add embedding model

Enter the deployment name exactly as defined in Azure and the following model configuration, for example:

{
  "kwargs": {
    "timeout": 120,
    "api_version": "2023-05-15"
  },
  "dimension": 3072
}

Enter deployment name

4. Add LLM models.​

Under Review & configure -> LLM models, click Add model. Add LLM model

Enter the deployment name and model configuration to match the settings defined in Azure AI Foundry.

Enter deployment name and model configuration

Model configuration for gpt-4.1:

{
    "kwargs": {
        "api_version": "2025-01-01-preview",
        "temperature": 0,
        "n": 1,
        "seed": 0,
        "max_tokens": 4096,
        "timeout": 120
    },
    "context_window_size": 1000000
}

Model configuration for gpt-4.1-mini:

{
    "kwargs": {
        "api_version": "2025-01-01-preview",
        "temperature": 0,
        "n": 1,
        "seed": 0,
        "max_tokens": 4096,
        "timeout": 120
    },
    "context_window_size": 1000000
}

5. Assign models to pipelines.​

Under Pipeline assignments, assign models to each pipeline based on their intended usage.

Hover over the ℹ️ icon to view the requirements and purpose of each pipeline.

  • For pipeline that requires "Advanced capabilities", we recommend using gpt-4.1 model.
  • For other pipelines, we could assign gpt-4.1-mini model.

Assign models to pipelines

After all pipelines are assigned, click Complete setup. Congratulations! Now WrenAI is fully configured to use Azure OpenAI.