Gemini's default safety thresholds are set too high, resulting in frequent censorship of generated text. I have lowered the thresholds for all four safety categories according to Google's documentation.
1.调用gemini api出现异常时没有向下游返回错误信息,后续处理流程可能要根据错误信息做相应补偿机制 2.修复特殊场景中出现索引越界导导致应用退出