WIP: Attempt to fix reasoning models support

[raghubetina]

Summary

• WIP PR to save current progress on fixing reasoning model support
• Always sends the reasoning parameter with summary=auto to OpenAI API for all calls
• Attempts to extract and store reasoning data from API responses

Current status

This is still not working correctly but saving the work-in-progress for reference.

[ellipsis-dev]

Caution

Changes requested ❌

Reviewed everything up to 60024a4 in 1 minute and 41 seconds. Click for details.

• Reviewed 90 lines of code in 2 files
• Skipped 0 files when reviewing.
• Skipped posting 6 draft comments. View those below.
• Modify your settings and rules to customize what types of comments Ellipsis leaves. And don't forget to react with 👍 or 👎 to teach Ellipsis.

1. CHANGELOG.md:3

• Draft comment:
  The changelog update describes the new reasoning parameter functionality. Ensure that the changelog reflects all changes consistently (e.g., see additional notes in version 0.0.7).
• Reason this comment was not posted:
  Comment did not seem useful. Confidence is useful = 0% <= threshold 50% This comment is asking the PR author to ensure that the changelog is updated, which falls under the rule of not asking the author to update the PR description or similar documentation. It doesn't provide a specific code suggestion or point out a potential issue in the code itself.

2. lib/ai/chat.rb:128

• Draft comment:
  Good refactoring of the reasoning parameters in 'assistant!'. Always passing 'summary=auto' and conditionally adding 'effort' correctly meets the new API requirements.
• Reason this comment was not posted:
  Confidence changes required: 0% <= threshold 50% None

3. lib/ai/chat.rb:215

• Draft comment:
  The 'reasoning' method correctly retrieves reasoning data from assistant messages. Verify its usage in the context of the overall API response processing.
• Reason this comment was not posted:
  Confidence changes required: 30% <= threshold 50% None

4. lib/ai/chat.rb:8

• Draft comment:
  The attr_reader for :reasoning_output is declared but never assigned. If you intend to expose separate reasoning output, consider setting @reasoning_output when extracting reasoning data, else remove the attribute.
• Reason this comment was not posted:
  Marked as duplicate.

5. lib/ai/chat.rb:195

• Draft comment:
  The extraction of the output text uses a safe navigation with a ternary operator which might be less clear. Consider refactoring for clarity, e.g. explicitly checking if output_text_item is not nil and has a 'text' key.
• Reason this comment was not posted:
  Comment was not on a location in the diff, so it can't be submitted as a review comment.

6. lib/ai/chat.rb:215

• Draft comment:
  The 'reasoning' method searches the messages array each time for the last assistant message with reasoning. If this becomes a performance issue, consider caching the reasoning output upon extraction.
• Reason this comment was not posted:
  Confidence changes required: 50% <= threshold 50% None

Workflow ID: wflow_IVPzTxwBXXnSN0Ay

^{You can customize by changing your verbosity settings, reacting with 👍 or 👎, replying to comments, or adding code review rules.}

[ellipsis-dev]

The new attr_reader 'reasoning_output' is declared but never used in the class. Consider removing or implementing its usage.

Suggested change

	attr_reader :reasoning_effort, :reasoning_output
	attr_reader :reasoning_effort