我是一名Python老师，正在做一个示例项目，想给学生展示如何在项目中运用AI，而不只是让ChatGPT帮忙写代码。

下面是我写的代码（它不是最高效的版本，但因为是给低年级孩子看的，所以效率不是重点，主要是让他们接触更进阶的Python内容）：

from google import genai
from google.genai import types
import time

# AI Connection
client = genai.Client(api_key="[KEY_HERE]")

# My attempt to deal with a "rate" issue (which turned out to not do anything useful)
print("Please wait...")
for x in range (0, 60):
    percent = round(x/60 * 100)
    print(str(percent) + "%")
    time.sleep(1)

# Function to make calls to AI more "modular"
def askAIStream(personality, question, history):
    # Variable to track history within this function call
    newHistory = ""

    chat = client.chats.create(
        model="gemini-2.0-flash",
        config = types.GenerateContentConfig(
            system_instruction=personality,
        ),
        history=history
    )
    response = chat.send_message_stream(question)
    
    for chunk in response:
        newHistory += chunk.text
        print(chunk.text, end="")

    # Return the response if we need it for a later prompt
    return newHistory

personality = "No special instructions"

history = []

response = askAIStream(personality, "Give me a random 1-word response that can be used as a topic for a joke", history)
history.append(response)

response = askAIStream(personality, "Tell me a joke about " + response, history)
history.append(response)

response = askAIStream(personality, "I don't get it, can you explain this joke to me?", history)
history.append(response)

# Start the history variable over to keep track of what's going on
history.clear()

# Generate the personality of the AI
personality = """You are a classic storyteller who has a way with words.
 You are weaving a story that you will get responses to that you will then 
 use to continue the story. All your responses should include a couple of options that the user can choose and end with 'What do you do?'
 to help the story continue. You will find a way to end the story in 5 
 responses and then end the story with a classic 'The End'."""

# Start the story from the "askAIStream" function and attach the response to the history list
response = askAIStream(personality, "Start a random story", history)
history.append(response)

# Start a loop for the story to continue for 5 more responses
## This is the where my responses eventually break
for i in range(5):
    # Prompt the user to say something
    user = input("")

    # Call the "askAIStream" function and attach the response to the history list
    response = askAIStream(personality, user, history)
    history.append(response)

遇到的问题 #

在使用send_message_stream时，AI的响应文本会突然中断，程序直接等待用户的下一次输入（这种情况会在末尾循环的不同阶段发生，但最终一定会出现）——具体表现就是AI的句子还没说完，就停住等着用户输入了。

如果我改用send_message()就不会有这个问题，但我更喜欢stream的方式，因为它和学生们平时玩的ChatGPT聊天框一样，是流式输出的。

我排查过的方向 #

一开始我以为是速率限制的问题（我有Google One订阅，但没单独付费使用Gemini），所以加了60秒的等待倒计时，但现在想想这很蠢——如果真的是速率限制，应该会有明确的“速率超限”错误提示，而且send_message()也会受到影响，但实际上后者完全正常。

我知道我的代码存在一些效率问题（打算在给学生看之前整理好），但我觉得这些问题不是导致响应中断的原因。

有没有人遇到过类似的情况？能不能给我一些排查的方向或者解决方案？

提前谢谢大家了！

备注：内容来源于stack exchange，提问作者Willowvayle