fix(wren-ai-service): respect user response_format override in intent_classification pipeline

[jayvenn21]

Summary

This PR fixes an issue where the intent_classification pipeline always injects
OpenAI-only structured output defaults (json_schema), even when users
override response_format in config.

Details

Qwen-compatible API servers do not support OpenAI structured output and
disconnect when receiving unsupported parameters. The current merge logic
overwrites user-provided kwargs, making it impossible to disable structured
output.

This change ensures user configuration fully overrides defaults, allowing
non-OpenAI providers to function correctly.

Impact

• Fixes Qwen / LiteLLM compatibility
• Respects config precedence
• Preserves backward compatibility for OpenAI users

Closes #2113

Summary by CodeRabbit

Release Notes

• Refactor
  • Intent classification configuration now supports user and provider-supplied model parameters. Custom settings take priority over hard-coded defaults, enabling more granular control over intent classification behavior. Response format handling has been improved to prevent conflicts between custom and default configurations.

[coderabbitai]

Walkthrough

The intent classification generator now supports user-configurable model kwargs. A default configuration is defined, and any provider-supplied kwargs are overlaid on top, with special handling for response_format conflicts. The generation process now uses these computed kwargs instead of the static default constant.

Changes

Cohort / File(s)	Summary
Intent Classification Configuration wren-ai-service/src/pipelines/generation/intent_classification.py	Added logic to merge provider-supplied model kwargs with defaults, with conditional response_format adjustment and logging. Changed generator initialization to use computed model_kwargs instead of static constant.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~8 minutes

Poem

🐰 ✨ A rabbit hops with glee,
For defaults now bend willingly!
Override the old, let configs flow,
User choices steal the show! 🎯

🚥 Pre-merge checks | ✅ 3 | ❌ 2

❌ Failed checks (2 warnings)

Check name	Status	Explanation	Resolution
Linked Issues check	⚠️ Warning	The PR addresses issue #2113 by respecting user response_format overrides to fix Qwen compatibility, but does not comprehensively address all diagnostic and verification objectives outlined in the issue.	The code fix resolves the immediate response_format override issue, but the PR does not address network verification, timeout tuning, detailed logging, or reproducible connectivity tests mentioned in issue #2113.
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (3 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and specifically describes the main change: respecting user response_format override in the intent_classification pipeline, which directly matches the code modification.
Out of Scope Changes check	✅ Passed	The changes are narrowly focused on configurable model_kwargs and response_format override handling in the intent_classification pipeline, directly aligned with the stated objective of fixing user configuration precedence.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}