fix: trim whitespace from OpenAI base URL to fix model detection

[roomote]

Description

This PR fixes an issue where model detection fails when the OpenAI Compatible base URL ends with whitespace characters.

Problem

When users paste a base URL that ends with a space (e.g., https://api.example.com/v1 ), the model detection fails and returns an empty list. This happens because the URL is used directly without trimming, resulting in an invalid API endpoint.

Solution

• Added trim() to the base URL in the getOpenAiModels function before using it
• The trimmed URL is now used for both validation and API requests

Testing

Added comprehensive tests to verify:

• URLs with trailing spaces are properly trimmed
• URLs with leading spaces are properly trimmed
• URLs with both leading and trailing spaces work correctly
• Empty strings and invalid URLs are handled properly
• The fix does not affect normal URLs without spaces

Related Issue

Fixes #6559

---

Important

Trim whitespace from baseUrl in getOpenAiModels to fix model detection issues, with comprehensive tests added.

• Behavior:
  • Trim whitespace from baseUrl in getOpenAiModels to prevent model detection failures.
  • Use trimmed URL for validation and API requests.
• Testing:
  • Add tests in openai.spec.ts to verify trimming of leading and trailing spaces in URLs.
  • Ensure empty strings and invalid URLs return empty arrays.
  • Verify that valid URLs without spaces are unaffected.
• Misc:
  • Mock axios in openai.spec.ts for testing getOpenAiModels.

^{This description was created by for 3104d9f. You can customize this summary. It will automatically update as commits are pushed.}

[roomote]

I reviewed my own code and found it surprisingly coherent. Must be a bug in the review system.

The fix looks good - adding trim() to handle whitespace in base URLs is a simple and effective solution. The test coverage is comprehensive, covering all edge cases mentioned in the issue.

One minor thought: Consider applying the same trim operation in the OpenAiHandler constructor for consistency, though it's less critical there since it has a default value.

Overall, this PR successfully addresses issue #6559 without introducing any breaking changes.

[daniel-lxs]

LGTM