fix(compat): coerce by_alias=None to False in model_dump to fix TypeError with pydantic v2

[NIK-TIGER-BILL]

Summary

Fixes #2921

Problem

model_dump() in _compat.py declares by_alias: bool | None = None and was forwarding the value as-is to pydantic's model_dump(). pydantic-core's Rust-based serializer (pydantic v2) does not accept None for by_alias and raises:

TypeError: argument 'by_alias': 'NoneType' object cannot be converted to 'PyBool'

This error only manifests when DEBUG-level logging is enabled on the openai logger, because the affected call site omits by_alias (so it defaults to None) and is inside a if log.isEnabledFor(logging.DEBUG) block in _base_client.py.

Fix

Coerce None to False before passing to pydantic:

# Before
by_alias=by_alias

# After
by_alias=bool(by_alias) if by_alias is not None else False

The pydantic v1 code path already used bool(by_alias), so this aligns both branches.

Testing

import logging
logging.basicConfig(level=logging.DEBUG)
from openai import OpenAI
client = OpenAI(api_key="fake")
client.chat.completions.create(model="gpt-4.1", messages=[{"role": "user", "content": "hi"}])
# No longer raises TypeError — proceeds to the network request

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: bee7537f80

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

Preserve config-driven alias behavior in model_dump

When by_alias is omitted, supported Pydantic 2.11+ models can still serialize aliases via ConfigDict(serialize_by_alias=True). Forcing None to False here changes that contract: _compat.model_dump() no longer defers to the model config, and both _utils._transform._transform_recursive() and _async_transform_recursive() call this helper for arbitrary pydantic.BaseModel inputs when building request bodies. In that scenario, aliased third-party Pydantic models will now emit field names instead of aliases, which can send the wrong keys to the API.

Useful? React with 👍 / 👎.