Azure OpenAI

Azure OpenAI Service provides powerful AI capabilities with access to GPT models through an OpenAI-compatible API.

Supported Models

• gpt-4o - Latest GPT-4 optimized version
• gpt-4 - Standard GPT-4
• gpt-4-32k - Long context version
• gpt-35-turbo - GPT-3.5 Turbo
• gpt-35-turbo-16k - Long context version

Configuration

Basic Configuration

Configure in config.yaml or ~/.bytebuddy/config.yaml:

models:
  - name: "azure-gpt-4"
    provider: "azure"
    model: "gpt-4"
    apiKey: "${AZURE_OPENAI_API_KEY}"
    apiBase: "https://your-resource.openai.azure.com/"
    roles: ["chat", "edit"]
    env:
      deploymentName: "gpt-4-deployment"
      apiVersion: "2024-02-15-preview"
    defaultCompletionOptions:
      temperature: 0.7
      maxTokens: 4000

Multi-Model Configuration

models:
  - name: "azure-gpt-4"
    provider: "azure"
    model: "gpt-4"
    apiKey: "${AZURE_OPENAI_API_KEY}"
    apiBase: "https://your-resource.openai.azure.com/"
    roles: ["chat", "edit"]
    env:
      deploymentName: "gpt4-deployment"
      apiVersion: "2024-02-15-preview"
    defaultCompletionOptions:
      temperature: 0.7
      maxTokens: 4000

  - name: "azure-gpt-35-turbo"
    provider: "azure"
    model: "gpt-35-turbo"
    apiKey: "${AZURE_OPENAI_API_KEY}"
    apiBase: "https://your-resource.openai.azure.com/"
    roles: ["chat"]
    env:
      deploymentName: "gpt35-turbo-deployment"
      apiVersion: "2024-02-15-preview"
    defaultCompletionOptions:
      temperature: 0.5
      maxTokens: 4000

Configuration Fields

Required Fields

• name: Unique identifier for the model configuration
• provider: Set to "azure"
• model: Model name (e.g., gpt-4, gpt-35-turbo)
• apiKey: Azure OpenAI API key
• apiBase: Azure OpenAI resource endpoint

Environment Configuration (env)

Azure requires specific parameters via the env field:

• deploymentName: Deployment name (created in Azure Portal)
• apiVersion: API version (recommended: 2024-02-15-preview)
• resourceName: Resource name (optional)

Optional Fields

• roles: Model roles [chat, edit, apply, autocomplete]
• defaultCompletionOptions:
  • temperature: Control randomness (0-2)
  • maxTokens: Maximum tokens
  • topP: Nucleus sampling parameter
  • frequencyPenalty: Frequency penalty
  • presencePenalty: Presence penalty
• requestOptions:
  • timeout: Request timeout (milliseconds)
  • verifySsl: Whether to verify SSL certificates

Environment Variables

# ~/.bashrc or ~/.zshrc
export AZURE_OPENAI_API_KEY="your-azure-api-key"

Getting API Key

1. Log in to Azure Portal
2. Create or locate your Azure OpenAI resource
3. Get API key and endpoint from "Keys and Endpoint" section
4. Create model deployment in "Deployments" section
5. Note the deployment name for configuration

Use Case Configurations

Code Development

models:
  - name: "code-assistant"
    provider: "azure"
    model: "gpt-4"
    apiKey: "${AZURE_OPENAI_API_KEY}"
    apiBase: "https://your-resource.openai.azure.com/"
    roles: ["chat", "edit"]
    env:
      deploymentName: "gpt4-deployment"
      apiVersion: "2024-02-15-preview"
    defaultCompletionOptions:
      temperature: 0.1
      maxTokens: 2000

Quick Chat

models:
  - name: "quick-chat"
    provider: "azure"
    model: "gpt-35-turbo"
    apiKey: "${AZURE_OPENAI_API_KEY}"
    apiBase: "https://your-resource.openai.azure.com/"
    roles: ["chat"]
    env:
      deploymentName: "gpt35-deployment"
      apiVersion: "2024-02-15-preview"
    defaultCompletionOptions:
      temperature: 0.7
      maxTokens: 2000

Troubleshooting

Common Errors

1. 401 Unauthorized: Check if API key is correct
2. 404 Not Found: Verify deployment name, resource endpoint, and API version
3. 429 Too Many Requests: Rate limit reached, wait and retry
4. DeploymentNotFound: Check if deploymentName is correct

Debugging Steps

1. Verify API key and endpoint URL
2. Confirm deployment name exists in Azure Portal
3. Check if API version is supported
4. Confirm quotas and rate limits

Best Practices

1. Security

• Use environment variables to store API keys
• Rotate API keys regularly
• Use Azure Key Vault to manage keys

2. Performance Optimization

• Choose appropriate model for the task
• Set reasonable maxTokens limits
• Use streaming responses for better UX

3. Cost Control

• Monitor usage and costs
• Configure different deployments for different purposes
• Use GPT-3.5 Turbo for simple tasks

4. Configuration Management

• Use different deployments for different environments
• Configure appropriate rate limits
• Implement retry logic for transient errors