fix & refactor: add supportsTemperature and Responses API flags

[snipeship]

Extend ModelInfo schema with supportsTemperature and usesResponsesApi capabilities to control request param inclusion and API selection.

Refactor OpenAiNativeHandler to generically handle Responses API models instead of hardcoded families, normalizing IDs and gating temperature, verbosity, and max token params via getModelParams.

Update GlamaHandler, LiteLLMHandler, UnboundHandler, and XAIHandler to use getModelParams for capability-aware temperature/max token handling.

Enhance tests to cover Responses API flows, conversation continuity, and temperature stripping for unsupported models, replacing SSE mocks with SDK responses.create where applicable.

Related GitHub Issue

Status: Open.
#6965

Closes: #6965

Roo Code Task Context (Optional)

Description

Fixes some slop as well for better maintaineability

Test Procedure

Run GPT-5, o3, and other providers

Pre-Submission Checklist

• Issue Linked: This PR is linked to an approved GitHub Issue (see "Related GitHub Issue" above).
• Scope: My changes are focused on the linked issue (one major feature/fix per PR).
• Self-Review: I have performed a thorough self-review of my code.
• Testing: New and/or updated tests have been added to cover my changes (if applicable).
• Documentation Impact: I have considered if my changes require documentation updates (see "Documentation Updates" section below).
• Contribution Guidelines: I have read and agree to the Contributor Guidelines.

Documentation Updates

Additional Notes

Get in Touch

---

Important

Add supportsTemperature and usesResponsesApi flags to ModelInfo schema and refactor handlers to use these capabilities for parameter management.

• ModelInfo Schema:
  • Added supportsTemperature and usesResponsesApi flags to modelInfoSchema in model.ts.
• Handlers:
  • Refactored OpenAiNativeHandler to handle Responses API models generically, using getModelParams for parameter gating.
  • Updated GlamaHandler, LiteLLMHandler, UnboundHandler, and XAIHandler to use getModelParams for temperature and max token handling.
• Tests:
  • Enhanced tests in openai-native.spec.ts to cover Responses API flows and temperature stripping for unsupported models.
  • Updated unbound.spec.ts to test temperature handling for openai/o3-mini.
  • Added tests in model-params.spec.ts for verbosity and reasoning settings.

^{This description was created by for 301731a. You can customize this summary. It will automatically update as commits are pushed.}

[roomote]

Thank you for your contribution! I've reviewed the changes and have some suggestions for improvement. The implementation looks solid overall - good work on extending the ModelInfo schema with capability flags and refactoring the handlers to use these for parameter management.

Suggestions:

1. Temperature default handling - In openai-native.ts, the logic for defaultTemperature seems backwards. When supportsTemperature is true, you're passing undefined, but when false you're passing OPENAI_NATIVE_DEFAULT_TEMPERATURE. Should this be reversed?

2. Missing documentation - The new supportsTemperature and usesResponsesApi fields would benefit from more detailed JSDoc comments explaining their purpose and implications.

3. Unused import - GPT5_DEFAULT_TEMPERATURE is still imported in openai-native.ts but was removed from the exports.

4. Magic number - Consider extracting the 1.0 temperature value in model-params.ts as a named constant for hybrid reasoning models.

5. Comment style - The Q&A format comment about GPT-5 temperature could be more concise.

Overall, this is a well-structured refactor that improves the codebase's maintainability. The test coverage is comprehensive.

[daniel-lxs]

Hey @snipeship, Thank you for your contribution! I opened this PR which implements the responses API for all models. I'll close this PR in favor of this on #7067.

Again, we appreciate your contributions, and I look forward to seeing more of them!