fix(genai): preserve thought_signature in validation retries

[drdraad]

Describe your changes
Preserve GenAI response metadata (e.g. thought_signature) across validation retries in GENAI_TOOLS by replaying the model Content and then adding a tool function_response with validation errors.
Add targeted tests covering the GENAI reask path (unit) and a minimal end-to-end validation retry check (LLM-gated).

Issue ticket number and link
N/A

Checklist before requesting a review
[x] I have performed a self-review of my code
[x] If it is a core feature, I have added thorough tests.
[x] If it is a core feature, I have added documentation.

Test plan:
test_genai_reask.py
test_reask.py

---

Important

Enhances reask_genai_tools to preserve thought_signature during validation retries and adds tests for comprehensive coverage.

• Behavior:
  • reask_genai_tools in utils.py now preserves thought_signature by replaying model content and adding tool function responses with validation errors.
  • Handles cases where function calls are absent or not the first part.
• Tests:
  • Adds test_reask.py to test reask_genai_tools for preserving thought_signature, handling absent function calls, and processing None responses.
  • Adds test_genai_reask.py for end-to-end validation retry checks in GENAI_TOOLS mode.
• Misc:
  • Removes unused imports from audit_patterns.py, check_links.py, and validate_meta_tags.py.

^{This description was created by for f10c1a1. You can customize this summary. It will automatically update as commits are pushed.}

[ellipsis-dev]

Important

Looks good to me! 👍

Reviewed everything up to f10c1a1 in 1 minute and 5 seconds. Click for details.

• Reviewed 271 lines of code in 7 files
• Skipped 0 files when reviewing.
• Skipped posting 10 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. instructor/providers/gemini/utils.py:681

• Draft comment:
  The shallow copying of the 'contents' list is clear, but consider if a deep copy might be needed for nested mutable objects in future modifications.
• Reason this comment was not posted:
  Confidence changes required: 50% <= threshold 85% None

2. instructor/providers/gemini/utils.py:693

• Draft comment:
  Inside the loop iterating over candidates, the selection of 'function_call' and 'function_call_content' is clear. Adding a brief inline comment to explain the selection logic could improve future maintainability.
• Reason this comment was not posted:
  Confidence changes required: 30% <= threshold 85% None

3. instructor/providers/gemini/utils.py:718

• Draft comment:
  In the branch where no function_call is found, the code appends the model content (if available) and a user message with the error. A comment explaining why this fallback is chosen would benefit future readers.
• Reason this comment was not posted:
  Confidence changes required: 30% <= threshold 85% None

4. instructor/providers/gemini/utils.py:730

• Draft comment:
  When a function_call is found, the tool response is appended while preserving the original thought_signature. This behavior is validated by tests; ensure that any future changes maintain this preservation.
• Reason this comment was not posted:
  Confidence changes required: 20% <= threshold 85% None

5. scripts/audit_patterns.py:15

• Draft comment:
  Removed the unused 'Set' import from typing. This cleanup improves code clarity.
• Reason this comment was not posted:
  Confidence changes required: 0% <= threshold 85% None

6. scripts/check_links.py:12

• Draft comment:
  The removal of 'defaultdict' from the imports is appropriate since it's not used in this file.
• Reason this comment was not posted:
  Confidence changes required: 0% <= threshold 85% None

7. scripts/fix_old_patterns.py:14

• Draft comment:
  Simplified type hints by removing unused types (e.g. Dict) from the imports. This reduces clutter in the code.
• Reason this comment was not posted:
  Confidence changes required: 0% <= threshold 85% None

8. scripts/validate_meta_tags.py:15

• Draft comment:
  Removed the unused 'Set' import from typing to keep the import list clean.
• Reason this comment was not posted:
  Confidence changes required: 0% <= threshold 85% None

9. tests/llm/test_genai/test_reask.py:7

• Draft comment:
  The test for GENAI_TOOLS validation retry is well-structured and clearly verifies that the retry mechanism triggers the expected InstructorRetryException with the correct number of attempts.
• Reason this comment was not posted:
  Confidence changes required: 0% <= threshold 85% None

10. tests/test_genai_reask.py:12

• Draft comment:
  The newly added tests effectively validate reask_genai_tools behavior, including preservation of thought_signature, proper fallback when response is None, and handling cases with no function call. This provides comprehensive coverage.
• Reason this comment was not posted:
  Confidence changes required: 0% <= threshold 85% None

Workflow ID: wflow_vSPZ3odDPSOVjFeD

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

[jxnl]

@cursoragent can you update the change log and review this