The logic that sets the variable is_openai_reasoning_model in line 92 in llm/openai.py does not take into account the existence of regional APIs. For instance, I have to set https://eu.api.openai.com/v1 as the api_base in the config yaml file – because of the data residency laws – and hence OpenEvolve thinks I'm not using an OpenAI and hence passes it a wrong set of parameters (max_tokens vs max_completion_tokens).
The logic that sets the variable
is_openai_reasoning_modelin line 92 in llm/openai.py does not take into account the existence of regional APIs. For instance, I have to set https://eu.api.openai.com/v1 as the api_base in the config yaml file – because of the data residency laws – and hence OpenEvolve thinks I'm not using an OpenAI and hence passes it a wrong set of parameters (max_tokens vs max_completion_tokens).