[ZedingZhang (ZedingZhang)]

Motivation & Context

An injected AsyncAzureOpenAI client already has a deployment, but the shared OpenAI settings resolver reads AZURE_OPENAI_* model variables first. The wrapper then calls a different deployment than the SDK client the caller supplied. Chat, chat-completions, and embedding clients all go through this path.

Description & Review Guide

• What are the major changes? In load_openai_service_settings, resolve the Azure model as explicit model, then the injected AsyncAzureOpenAI deployment, then the environment aliases. Document that order and add regression tests for the three public clients.
• What is the impact of these changes? Injected Azure clients keep the caller's deployment when model is omitted. An explicit model still wins. A plain AsyncOpenAI client still uses the OpenAI model variables. Creating a client without an injected client is unchanged.
• What do you want reviewers to focus on? The precedence change is only in the Azure branch of load_openai_service_settings. Hosts that previously let an ambient Azure model variable override an injected client's deployment will now follow the client.

Related Issue

Fixes #9949. The issue has the ready-for-implementation label; no competing open PR was found before submission.

AI Assistance

• No material AI assistance was used.
• This is an AI-assisted contribution.

AI assistance details: Cursor's coding agent assisted with the implementation, regression tests, docstring and README updates, and this PR description. I reviewed the change and accept responsibility for it.

Contribution Checklist

• The code builds clean without any errors or warnings
• All unit tests pass, and I have added new tests where possible — new injected-client tests plus the existing Azure unit tests: 49 passed, 37 integration tests deselected. Ruff and Pyright report no issues on the changed files.
• The PR follows the Contribution Guidelines
• This PR links to an agreed issue with no competing open PR, or the Related Issue section documents a trivial-change or repository-automation exception.
• This is not a breaking change. If it is a breaking change, add the breaking change label (or add "[BREAKING]" to the title prefix, before or after any language prefix) — a workflow keeps the label and title prefix in sync automatically.

[Copilot]

🟡 Changes recommended

The new inline SDK-client factory violates the required test-helper layout and must be moved to a callable factory fixture or an imported helper module.

1 open issue

• Move the SDK-client factory into a fixture or helper module · New

What changed in this PR

Fixes #9949 by making the shared Python OpenAI settings resolver prefer an injected Azure client's deployment over ambient model variables.

Changes:

• Sets Azure model precedence to explicit model, injected deployment, then environment aliases.
• Adds regression tests across chat, chat-completions, and embedding clients.
• Documents the precedence while preserving plain OpenAI and no-client behavior.

File	Description
python/​packages/​openai/​tests/​openai/​test_injected_client_model.py	Tests injected-client precedence, explicit overrides, and environment fallback.
python/​packages/​openai/​README.md	Explains injected deployment precedence and client ownership.
python/​packages/​openai/​AGENTS.md	Documents request-model resolution order.
python/​packages/​openai/​agent_framework_openai/​_shared.py	Resolves injected Azure deployments before environment aliases.
python/​packages/​openai/​agent_framework_openai/​_embedding_client.py	Updates raw and layered embedding constructor docstrings.
python/​packages/​openai/​agent_framework_openai/​_chat_completion_client.py	Updates raw and layered chat-completion constructor docstrings.
python/​packages/​openai/​agent_framework_openai/​_chat_client.py	Updates raw and layered chat constructor docstrings.

_{Review effort: Balanced}

[Copilot]

🟢 Approval recommended

No active review issues block the Azure model-precedence update.

1 open issue

• Move the SDK-client factory into a fixture or helper module

_{Review effort: Balanced}