fix: add reasoning field fallback in ChatCompletionsTransport.parse (#1731)

[RemyLoveLogicAI]

Closes #1731

Summary

ChatCompletionsTransport.parse in helpers/litellm_transport.py only read reasoning from the legacy reasoning_content field. Newer vLLM / OpenAI-compatible backends (e.g. Qwen3.6 on vLLM) return reasoning under the newer reasoning field instead, causing reasoning to be silently dropped.

Fix

Added fallback checks for delta.reasoning and message.reasoning after the existing reasoning_content checks. The old field is checked first so existing providers are unaffected.

Verification

• Non-streaming message.reasoning_content preserves reasoning (unchanged)
• Non-streaming message.reasoning now preserves reasoning (new)
• Streaming delta.reasoning_content preserves reasoning (unchanged)
• Streaming delta.reasoning now preserves reasoning (new)
• Final-answer content handling unchanged
• Tool-call parsing unchanged
• Existing providers returning reasoning_content see no behavior change

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a0e62acb94

ℹ️ About Codex in GitHub

Your team has set up Codex to 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 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Close the new reasoning_delta expression

This assignment opens a parenthesized expression but never closes it before the following return, so importing this module now fails with SyntaxError: '(' was never closed (confirmed with python -m py_compile helpers/litellm_transport.py). Because helpers/litellm_transport.py is a core transport helper, this blocks any path that imports it rather than just the new reasoning fallback.

Useful? React with 👍 / 👎.