deepseek-ai

/

DeepSeek-V3.1-Base

like 887

Follow

DeepSeek 90.5k

发现使用deepseek-reasoner并指定messages一条为assistant时再生成回复（也就是自己回复自己时）会导致生成混乱的reasoning内容：首先输出<｜begin▁of▁sentence｜>然后输出一段随机的代码或者教程

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["OPENAI_API_KEY"], base_url="https://api.deepseek.com"
)

response = client.chat.completions.create(
    model="deepseek-reasoner",
    messages=[
        {"role": "system", "content": "You are a helpful assistant"},
        {"role": "user", "content": "read http://example.com"},
        {"role": "assistant", "content": "<xxxx>"},
    ],
    stream=True,
)

for chunk in response:
    if chunk.choices and chunk.choices[0].delta:
        delta = chunk.choices[0].delta
        data = getattr(delta, "reasoning_content", None)
        if not data:
            data = delta.content
        print(data, end="", flush=True)

我在写一个Agent，这个agent在调用工具失败的时候会无法生成工具的消息，导致messages列表最后一条是assitant的消息，像上面的例子一样，实测deepseek-r1-0528不会出现这种状况。