looks like gemini took 7s to send the end token, would receommend against using that llm, most people opt for openai gpt, when and why did you switch to gemini