fix: OpenAI SDK compat (ResponseTextConfig)

[h-brenoskuk]

Purpose

Fix ReadTheDocs/MkDocs build failures caused by a recent OpenAI Python SDK type rename.
vllm/entrypoints/openai/protocol.py imported ResponseTextConfig, which is present in openai==1.99.x but renamed to ResponseFormatTextConfig in newer releases (e.g., >=1.100.0).

This PR adds a backward/forward compatible import shim so both SDK lines work.

No functional/runtime behavior change in vLLM.

No public API changes.

Keeps docs builds deterministic without requiring a docs-only dependency pin.

Test Plan

openai update resulted in failing https://app.readthedocs.org/projects/vllm/builds/29240829/

python -m mkdocs build --clean --site-dir $READTHEDOCS_OUTPUT/html --config-file mkdocs.yaml
--
  | WARNING 08-18 17:18:33 [__init__.py:26] The vLLM package was not found, so its version could not be inspected. This may cause platform detection to fail.
  | INFO 08-18 17:18:33 [__init__.py:245] No platform detected, vLLM is running on UnspecifiedPlatform
  | Traceback (most recent call last):
  | File "<frozen runpy>", line 198, in _run_module_as_main
  | File "<frozen runpy>", line 88, in _run_code
  | File "/home/docs/checkouts/readthedocs.org/user_builds/vllm/envs/23119/lib/python3.12/site-packages/mkdocs/__main__.py", line 370, in <module>
  | cli()
  | File "/home/docs/checkouts/readthedocs.org/user_builds/vllm/envs/23119/lib/python3.12/site-packages/click/core.py", line 1442, in __call__
  | return self.main(*args, **kwargs)
  | ^^^^^^^^^^^^^^^^^^^^^^^^^^
  | File "/home/docs/checkouts/readthedocs.org/user_builds/vllm/envs/23119/lib/python3.12/site-packages/click/core.py", line 1363, in main
  | rv = self.invoke(ctx)
  | ^^^^^^^^^^^^^^^^
  | File "/home/docs/checkouts/readthedocs.org/user_builds/vllm/envs/23119/lib/python3.12/site-packages/click/core.py", line 1830, in invoke
  | return _process_result(sub_ctx.command.invoke(sub_ctx))
  | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  | File "/home/docs/checkouts/readthedocs.org/user_builds/vllm/envs/23119/lib/python3.12/site-packages/click/core.py", line 1226, in invoke
  | return ctx.invoke(self.callback, **ctx.params)
  | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  | File "/home/docs/checkouts/readthedocs.org/user_builds/vllm/envs/23119/lib/python3.12/site-packages/click/core.py", line 794, in invoke
  | return callback(*args, **kwargs)
  | ^^^^^^^^^^^^^^^^^^^^^^^^^
  | File "/home/docs/checkouts/readthedocs.org/user_builds/vllm/envs/23119/lib/python3.12/site-packages/mkdocs/__main__.py", line 285, in build_command
  | cfg = config.load_config(**kwargs)
  | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  | File "/home/docs/checkouts/readthedocs.org/user_builds/vllm/envs/23119/lib/python3.12/site-packages/mkdocs/config/base.py", line 374, in load_config
  | errors, warnings = cfg.validate()
  | ^^^^^^^^^^^^^^
  | File "/home/docs/checkouts/readthedocs.org/user_builds/vllm/envs/23119/lib/python3.12/site-packages/mkdocs/config/base.py", line 231, in validate
  | run_failed, run_warnings = self._validate()
  | ^^^^^^^^^^^^^^^^
  | File "/home/docs/checkouts/readthedocs.org/user_builds/vllm/envs/23119/lib/python3.12/site-packages/mkdocs/config/base.py", line 188, in _validate
  | self[key] = config_option.validate(value)
  | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  | File "/home/docs/checkouts/readthedocs.org/user_builds/vllm/envs/23119/lib/python3.12/site-packages/mkdocs/config/base.py", line 55, in validate
  | return self.run_validation(value)
  | ^^^^^^^^^^^^^^^^^^^^^^^^^^
  | File "/home/docs/checkouts/readthedocs.org/user_builds/vllm/envs/23119/lib/python3.12/site-packages/mkdocs/config/config_options.py", line 1187, in run_validation
  | hooks[name] = self._load_hook(name, path)
  | ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  | File "/home/docs/checkouts/readthedocs.org/user_builds/vllm/envs/23119/lib/python3.12/site-packages/mkdocs/config/config_options.py", line 1205, in _load_hook
  | spec.loader.exec_module(module)
  | File "<frozen importlib._bootstrap_external>", line 999, in exec_module
  | File "<frozen importlib._bootstrap>", line 488, in _call_with_frames_removed
  | File "/home/docs/checkouts/readthedocs.org/user_builds/vllm/checkouts/23119/docs/mkdocs/hooks/generate_argparse.py", line 18, in <module>
  | from vllm.benchmarks import latency  # noqa: E402
  | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  | File "/home/docs/checkouts/readthedocs.org/user_builds/vllm/checkouts/23119/vllm/benchmarks/latency.py", line 18, in <module>
  | from vllm.engine.arg_utils import EngineArgs
  | File "/home/docs/checkouts/readthedocs.org/user_builds/vllm/checkouts/23119/vllm/engine/arg_utils.py", line 40, in <module>
  | from vllm.reasoning import ReasoningParserManager
  | File "/home/docs/checkouts/readthedocs.org/user_builds/vllm/checkouts/23119/vllm/reasoning/__init__.py", line 5, in <module>
  | from .deepseek_r1_reasoning_parser import DeepSeekR1ReasoningParser
  | File "/home/docs/checkouts/readthedocs.org/user_builds/vllm/checkouts/23119/vllm/reasoning/deepseek_r1_reasoning_parser.py", line 9, in <module>
  | from vllm.entrypoints.openai.protocol import (ChatCompletionRequest,
  | File "/home/docs/checkouts/readthedocs.org/user_builds/vllm/checkouts/23119/vllm/entrypoints/openai/protocol.py", line 20, in <module>
  | from openai.types.responses import (ResponseFunctionToolCall,
  | ImportError: cannot import name 'ResponseTextConfig' from 'openai.types.responses' (/home/docs/checkouts/readthedocs.org/user_builds/vllm/envs/23119/lib/python3.12/site-packages/openai/types/responses/__init__.py). Did you mean: 'ResponseFormatTextConfig'?

Test result

This PR should pass the CI test.

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can either: Add ready label to the PR or enable auto-merge.

🚀

[gemini-code-assist]

Code Review

This pull request introduces a compatibility shim to handle a type rename in the OpenAI Python SDK, ensuring that vLLM works with both older and newer versions of the SDK. The logic correctly handles the different import paths. However, the comments explaining the version compatibility are inverted, which could cause confusion for future maintenance. I've suggested corrections for the comments to accurately reflect the logic.

[aarnphm]

we have a PR that lock openai<1.100

#23118

but I think this is still worth doing?

cc @mgoin @DarkLight1337

[mgoin]

Okay, I think this fix is acceptable

[simon-mo]

Can you merge main, the revert the commit that pin the version bf75632 from @mgoin

[mgoin]

Thank you for helping!

[simon-mo]

This PR still didn't include the reversal of the pin. Can you do the revert again?

[simon-mo]

thx. Will force merge once a test involving this code path pass