Azure OpenAI

Use your own Azure OpenAI credits through Keywords AI

Azure OpenAI requires three credentials: API key, API base URL (also called base_url or endpoint), and API version. Make sure you have all three from your Azure Portal before proceeding.

Prerequisites

Before integrating Azure OpenAI with Keywords AI, ensure you have:

1. An active Azure OpenAI subscription
2. At least one deployed model in your Azure OpenAI resource
3. Your deployment name matches those listed on our Models page

How to get your Azure OpenAI credentials

1. Access Azure Portal

   Visit your Azure Portal OpenAI Service.

   You will see your deployments. Choose the deployment you want to use.

   
2. Get your API Base URL (endpoint)

   Click on the “Endpoints” link to view your API base URL.

   The API base URL is also called endpoint or base_url. For OpenAI, this would be https://api.openai.com, but for Azure it’s unique to your resource (e.g., https://your-resource.openai.azure.com/).

   
3. Get your API Key

   Click on the “Manage Keys” link to view your API keys. You can use either Key 1 or Key 2.

4. Choose API Version

   Pick an API version from the available Azure OpenAI versions.

   Common API versions

   • 2024-10-21 (recommended)
   • 2024-08-01-preview
   • 2024-06-01
   • 2024-02-01

Azure OpenAI models compatibility

✅ Supported Frameworks

• OpenAI SDK
• LangChain SDK
• Vercel/OpenAI
• Vercel/Google
• LlamaIndex SDK
• Google GenAI
• Keywords native (Otel)

❌ Unsupported Frameworks

• Anthropic SDK
• Vercel/Anthropic

Add Azure OpenAI credentials

There are 2 ways to add your Azure OpenAI credentials to your requests:

Option 1: Via UI (Recommended)

1. Navigate to Providers page

   Go to Providers page and select Azure OpenAI.

   
2. Add your credentials

   Fill in the three required Azure OpenAI credentials:

   api_key · string · required

   Your Azure OpenAI API key from the Azure Portal.

   api_base · string · required

   Your Azure OpenAI endpoint URL (also called base_url or endpoint).

   What is api_base?

   The api_base (synonyms: base_url, endpoint) is the base URL for your Azure OpenAI resource.

   Examples:

   • Azure OpenAI: https://your-resource.openai.azure.com/
   • OpenAI (for reference): https://api.openai.com

   You can find this in your Azure Portal under “Endpoints”.

   api_version · string · required

   Your Azure OpenAI API version (e.g., 2024-10-21). See available versions.

   
3. Map your deployment name

   Azure allows custom deployment names. You need to map your deployment name so Keywords AI knows which Azure deployment to call.

   Important: When testing the model, use the format azure/{your-deployment-name} in the model test input.

   For example:

   • If your Azure deployment name is gpt-4o, use: azure/gpt-4o
   • If your Azure deployment name is my-gpt4-deployment, use: azure/my-gpt4-deployment

   
4. (Optional) Specify models to use these credentials

   You can limit which models use these credentials. Leave empty to apply your credentials for all Azure OpenAI models.

   To specify models:

   1. Go to the Models page
   2. Copy the model ID (e.g., azure/gpt-4o)
   3. Paste it in the input field
   4. Press Enter to add the model

   

Option 2: Via Code

OpenAI SDK

1. Install OpenAI SDK

   Python

   pip install openai
   
2. Initialize client with Azure credentials

   Python

   from openai import OpenAI
   
   client = OpenAI(
       base_url="https://api.keywordsai.co/api/",
       api_key="YOUR_KEYWORDSAI_API_KEY",
   )
   
   response = client.chat.completions.create(
       model="azure/gpt-4o",  # Use azure/ prefix with your deployment name
       messages=[{"role": "user", "content": "Hello, how are you?"}],
       extra_body={
           "customer_credentials": {
               "azure_openai": {
                   "api_key": "YOUR_AZURE_OPENAI_API_KEY",
                   "api_base": "https://your-resource.openai.azure.com/",
                   "api_version": "2024-10-21",
               }
           }
       }
   )
   print(response)
   
3. (Optional) Map custom deployment names

   If your Azure deployment name differs from the standard model name, use model_name_map to specify the mapping.

   Example: If you deployed gpt-4o with a custom name like my-gpt4-prod, map it like this:

   Python

   response = client.chat.completions.create(
       model="azure/gpt-4o",
       messages=[{"role": "user", "content": "Hello"}],
       extra_body={
           "customer_credentials": {
               "azure_openai": {
                   "api_key": "YOUR_AZURE_OPENAI_API_KEY",
                   "api_base": "https://your-resource.openai.azure.com/",
                   "api_version": "2024-10-21",
               }
           },
           "model_name_map": {
               "azure/gpt-4o": "azure/my-gpt4-prod"  # Map to your custom deployment name
           }
       }
   )
   

   model_name_map only works when your credentials are valid.
4. (Optional) Override credentials for specific models

   You can override credentials for individual models without changing the default credentials.

   Python

   response = client.chat.completions.create(
       model="azure/gpt-4o",
       messages=[{"role": "user", "content": "Hello"}],
       extra_body={
           "customer_credentials": {
               "azure_openai": {
                   "api_key": "DEFAULT_AZURE_API_KEY",
                   "api_base": "https://resource1.openai.azure.com/",
                   "api_version": "2024-10-21",
               }
           },
           "credential_override": {
               "azure/gpt-4o": {  # Override for this specific model only
                   "api_key": "ANOTHER_AZURE_API_KEY",
                   "api_base": "https://resource2.openai.azure.com/",
                   "api_version": "2024-10-21",
               }
           }
       }
   )
   
5. (Optional) Load balance between multiple deployments

   You can add multiple Azure deployments and load balance between them for better availability and cost optimization.

   Learn more: Load Balancing Guide

Standard API

1. Make API request with credentials

   Use the standard Keywords AI endpoint with Azure credentials:

   Python

   import requests
   
   url = "https://api.keywordsai.co/api/chat/completions"
   headers = {
       "Content-Type": "application/json",
       "Authorization": "Bearer YOUR_KEYWORDSAI_API_KEY"
   }
   
   payload = {
       "model": "azure/gpt-4o",
       "messages": [{"role": "user", "content": "Hello, how are you?"}],
       "customer_credentials": {
           "azure_openai": {
               "api_key": "YOUR_AZURE_OPENAI_API_KEY",
               "api_base": "https://your-resource.openai.azure.com/",
               "api_version": "2024-10-21"
           }
       }
   }
   
   response = requests.post(url, headers=headers, json=payload)
   print(response.json())
   
2. (Optional) Add model_name_map for custom deployment names

   If your Azure deployment has a custom name:

   {
     "model": "azure/gpt-4o",
     "messages": [{"role": "user", "content": "Hello"}],
     "customer_credentials": {
       "azure_openai": {
         "api_key": "YOUR_AZURE_OPENAI_API_KEY",
         "api_base": "https://your-resource.openai.azure.com/",
         "api_version": "2024-10-21"
       }
     },
     "model_name_map": {
       "azure/gpt-4o": "azure/my-custom-gpt4-deployment"
     }
   }
   

Testing your integration

Important: When testing in the model playground:

• The “Model test call” input field should contain the model identifier: azure/your-deployment-name (e.g., azure/gpt-4o)
• This is NOT for the message content
• The message content goes in the separate chat input field below
• The placeholder should show azure/gpt-4o as an example for Azure models

Requirements and troubleshooting

Requirements

1. Deployment name format: Use azure/{deployment-name} format (e.g., azure/gpt-4o)
2. Credentials match: Ensure your deployment name is correctly mapped if using custom names
3. Regional availability: Confirm your models are available in the region specified by your credentials
4. API version: Use a supported Azure OpenAI API version

Common issues

Error: Deployment not found

Cause: Your deployment name doesn’t match what’s configured in Azure.

Solution:

• Check your Azure deployment name in the Azure Portal
• Use model_name_map to map the standard model name to your custom deployment name
• Ensure you’re using the azure/ prefix (e.g., azure/gpt-4o)

Error: Invalid credentials

Cause: One or more of your Azure credentials are incorrect.

Solution:

• Verify api_key in Azure Portal → Manage Keys
• Verify api_base matches your resource endpoint exactly (including trailing slash)
• Verify api_version is a valid version

Model test call input is confusing

Issue: The “Model test call” input field is for specifying which model to test, not the message.

Correct usage:

• Model test call input: azure/gpt-4o (the model identifier with azure/ prefix)
• Message/chat input: Your actual message like “Hello, how are you?”

The placeholder should be dynamic and show azure/gpt-4o for Azure providers, not just gpt-4o.

Credential validation is not currently supported. Make sure your credentials are correct before making requests to avoid errors.

Supported models

DeepSeek models

• azure_deepseek/deepseek-v3
• azure_deepseek/deepseek-chat
• azure_deepseek/deepseek-reasoner
• azure_deepseek/deepseek-r1

O-series models

• azure/o1
• azure/o3
• azure/o3-mini
• azure/o4-mini

GPT-4 series

• azure/gpt-4
• azure/gpt-4-32k
• azure/gpt-4-1106-preview
• azure/gpt-4-vision-preview
• azure/gpt-4o
• azure/gpt-4o-mini
• azure/gpt-4o-2024-08-06
• azure/gpt-4.1
• azure/gpt-4.1-mini
• azure/gpt-4.1-nano

GPT-5 series

• azure/gpt-5
• azure/gpt-5-mini

GPT-3.5 series

• azure/gpt-35-turbo

You can find the complete and up-to-date list here.