To avoid incurring costs, we can simulate the outputs from the different LLM APIs
To avoid incurring costs, we can simulate the outputs from the different LLM APIs