Support Azure and Ollama APIs, let open for more APIs

[YunQiAI]

Overview

This PR extends the OpenAI Agents SDK to support multiple model providers, including Azure OpenAI and local Ollama models. It marks a key step toward making the platform truly "open" and adaptable for different environments.

What's New

Azure OpenAI Support

• Full integration with Azure API
• Most examples work as expected
• Limitation: Azure’s Responses API is not yet supported
• Configuration supports endpoint, API key, version, and deployment name

Example:

azure_settings = ModelSettings(
    provider="azure_openai",
    azure_endpoint="https://your-resource-name.openai.azure.com",
    azure_api_key="your-azure-api-key",
    azure_api_version="2024-02-15-preview",
    azure_deployment="your-deployment-name"
)

agent = Agent(
    name="AzureAssistant",
    instructions="You are a helpful assistant running on Azure.",
    model_settings=azure_settings
)

response = agent.run("What's the weather like in Tokyo today?")
print(response)

Ollama Local Model Support

• Basic support for local model inference
• Works with agent creation, simple tool calls, and streaming
• Includes example scripts and error handling

Framework Enhancements

• ModelSettings extended for provider-specific options
• ModelProviderFactory enables dynamic backend selection
• Adapter layer added for unified API response handling

Limitations

• Azure: No Responses API yet
• Ollama:
  • Handoff not fully implemented
  • Limited complex tool support
  • Streaming could be more stable

Future Plans

• Add support for AWS Bedrock, Anthropic, Google Vertex AI, etc.
• Improve cross-provider tool compatibility
• Extend streaming support and examples

Community Collaboration

This project openAPI-agents-python is also being developed independently. Community contributions are welcome—let’s work together to support more APIs and make the platform more versatile and robust.

[github-actions]

This PR is stale because it has been open for 10 days with no activity.