Fix: resolve parameter error when using o3 models with official OpenAI API

[uchkw]

Overview

This pull request fixes an API parameter error that occurred when using o-series models (such as o3) with the official OpenAI API endpoint.

Background

Previously, when calling the OpenAI API with o-series models, the following error was returned:

- INFO - HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 400 Bad Request"
- WARNING - Error on attempt 1/4: Error code: 400 - {'error': {'message': "Unsupported parameter: 'max_tokens' is not supported with this model. Use 'max_completion_tokens' instead.", 'type': 'invalid_request_error', 'param': 'max_tokens', 'code': 'unsupported_parameter'}}. Retrying...

This was due to the use of the max_tokens parameter, which is not supported by these models. Instead, the API expects the max_completion_tokens parameter.

Fix Details

• The code now checks if both the model name starts with "o" and the API base is https://api.openai.com/v1.
• If so, it uses the max_completion_tokens parameter instead of max_tokens when making the API call.
• For o-series models, parameters such as temperature and top_p are omitted because the official OpenAI API does not support customizing these values for these models.
• For all other models or endpoints, the original parameters are used.

Impact

This change ensures compatibility with the latest o-series models on the official OpenAI API, preventing 400 Bad Request errors due to unsupported parameters.

---

If you have any questions or need further clarification, please let me know. Thank you for reviewing this pull request!

[codelion]

This looks good, can you just run black openevolve tests examples to fix linting errors as described in https://github.com/codelion/openevolve/blob/main/CONTRIBUTING.md#code-style

[uchkw]

I did. Thanks for your comment.