RexWzh
commented
1 month ago +1 似乎是最新版引入的问题，我之前使用是正常的
data: {"id":"","object":"chat.completion.chunk","created":0,"model":"gemini","choices":[{"index":0,"delta":{"content":"我是 Gemini，一个由 Google 开发的多模态 AI 模型。作为一个大型语言"},"finish_reason":"stop"}]}
data: {"id":"","object":"chat.completion.chunk","created":0,"model":"gemini","choices":[{"index":0,"delta":{"content":"模型，我接受了大量文本数据的训练，能够理解和生成人类语言、回答问题并提供信息。我不断学习和完善，致力于提供全面"},"finish_reason":"stop"}]}
data: {"id":"","object":"chat.completion.chunk","created":0,"model":"gemini","choices":[{"index":0,"delta":{"content":"且有帮助的响应。"},"finish_reason":"stop"}]}
data: [DONE]
SLF-Heisenberg
youjin-git
mxdlzg
manjieqi
例行检查
问题描述 Gemini仅返回几个字，因为它们把"finish_reason":"stop"作为回答停止的信号 如果openai的返回格式是标准的，那Gemini的流式还需要改一下
Gemini的回答，可以看见每一个部分都有stop
GPT的回答，仅最后的一个部分有stop