feat: set horizon-alpha model max tokens to 32k for OpenRouter

[roomote]

$## Summary\n\nThis PR sets the default output limit for the horizon-alpha model with OpenRouter to 32k (32768 tokens) instead of the default 128k.\n\n## Changes\n\n- Added special handling in the parseOpenRouterModel function to override the max tokens for the openrouter/horizon-alpha model to 32768\n\n## Testing\n\n- All existing tests pass\n- The change follows the existing pattern for model-specific overrides (similar to how claude-3.7-sonnet models are handled)\n\n## Context\n\nThis change was requested via Slack to reduce the default output limit for the horizon-alpha model from 128k to 32k when using OpenRouter.

---

Important

Set maxTokens to 32k for openrouter/horizon-alpha model in parseOpenRouterModel.

• Behavior:
  • parseOpenRouterModel in openrouter.ts now sets maxTokens to 32k for openrouter/horizon-alpha model.
  • Other models remain unaffected by this change.
• Testing:
  • Added test in openrouter.spec.ts to verify maxTokens is set to 32k for horizon-alpha.
  • Confirmed no change for other models in openrouter.spec.ts.

^{This description was created by for 68e2daf. 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 follows the existing patterns well, but could benefit from test coverage and potentially a named constant.

[roomote]

Could we add test coverage for this new model override? The existing test file at src/api/providers/fetchers/tests/openrouter.spec.ts has comprehensive assertions for other model-specific overrides (like Claude 3.7 Sonnet models on lines 184-209), but doesn't include a test case for the horizon-alpha model. Adding something like expect(models["openrouter/horizon-alpha"].maxTokens).toEqual(32768) would ensure this override works as expected.

[roomote]

Would it be beneficial to define this as a named constant to make the intent clearer? While this follows the existing pattern used for other model overrides in the same file, something like const HORIZON_ALPHA_MAX_TOKENS = 32768 could improve readability and maintainability.

[hannesrudolph]

@roomote-agent run pr-fixer mode

[roomote]

👋 I've received your request to run pr-fixer mode on this PR. I'll review the changes and help fix any issues. Let me start by checking out the branch and analyzing the implementation.

[roomote]

✅ PR Fixer Complete

I've successfully reviewed and enhanced this PR. Here's what I accomplished:

🔍 Analysis

• Reviewed the existing implementation that sets model max tokens to 32k
• Confirmed the change follows the established pattern for model-specific overrides
• Verified all existing CI checks were already passing

🧪 Enhancements Made

• Added comprehensive unit tests for the function
• Created specific test cases to verify the 32k max tokens override for
• Added negative test case to ensure other models are not affected by this override
• All tests pass successfully

✅ Verification

• All CI checks continue to pass (compile, lint, unit tests, integration tests)
• Code follows existing patterns and maintains consistency
• Implementation is working as expected

The PR is ready for review and merge. The horizon-alpha model will now correctly use 32k max tokens instead of the default 128k limit when used through OpenRouter.

[hannesrudolph]

@roomote-agent mimic this pr but for the horizon-beta model

[roomote]

Hi @hannesrudolph! I see your request to mimic this PR for the horizon-beta model. I'll implement the same 32k max tokens limit for horizon-beta. Working on it now!

[roomote]

Hi @hannesrudolph! I've created PR #6577 to implement the same 32k max tokens limit for the horizon-beta model: #6577

The implementation follows the exact same pattern as the horizon-alpha changes. All tests are passing and CI checks have completed successfully. ✅