上下文缓存（Context Cache）

import os
from openai import OpenAI


client = OpenAI(
    # 若没有配置环境变量，请用阿里云百炼API Key将下行替换为：api_key="sk-xxx",
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",
)


def get_completion(messages):
    completion = client.chat.completions.create(
        model="qwen3-coder-plus",
        messages=messages
    )
    return completion


# 确保输入文本超过1024Token被缓存
red_black_tree = "<完整的红黑树代码实现>"
# 将cache_control添加到messages的最后一个content
messages = [
    {"role": "system", "content": "你是一个有用的代码助手"},
    {
        "role": "user",
        "content": [
            {
                "type": "text",
                "text": red_black_tree,
            },
            {
                "type": "text",
                "text": "该代码主要实现了什么功能",
                "cache_control": {"type": "ephemeral"},
            }
        ],
    }
]

completion = get_completion(messages)
print("第1轮")
print("=" * 20 + "回复内容" + "=" * 20)
print(completion.choices[0].message.content)
print("=" * 20 + "Token消耗" + "=" * 20)
print(completion.usage)
print(f"命中缓存Token数：{completion.usage.prompt_tokens_details.cached_tokens}")
print(f"用于创建缓存Token数：{completion.usage.prompt_tokens_details.cache_creation_input_tokens}")
assistant_output = completion.choices[0].message.content
# 将cache_control添加到messages的最后一个content
messages = [
    {"role": "system", "content": "你是一个有用的代码助手"},
    {
        "role": "user",
        "content": [
            {
                "type": "text",
                "text": red_black_tree,
            },
            {
                "type": "text",
                "text": "该代码主要实现了什么功能",
                "cache_control": {"type": "ephemeral"},
            }
        ],
    }
]
completion = get_completion(messages)
print("=" * 20 + "回复内容" + "=" * 20)
print(completion.choices[0].message.content)
print("=" * 20 + "Token消耗" + "=" * 20)
print(completion.model_dump_json())
print(f"命中缓存Token数：{completion.usage.prompt_tokens_details.cached_tokens}")
print(f"用于创建缓存Token数：{completion.usage.prompt_tokens_details.cache_creation_input_tokens}")

第1轮
====================回复内容====================
这段 Python 代码实现了一个完整的 **红黑树（Red-Black Tree）** 数据结构。<省略后续回答>
====================Token消耗====================
CompletionUsage(completion_tokens=820, prompt_tokens=2069, total_tokens=2889, completion_tokens_details=None, prompt_tokens_details=PromptTokensDetails(audio_tokens=None, cached_tokens=0, cache_creation={'ephemeral_5m_input_tokens': 2065}, cache_creation_input_tokens=2065, cache_type='ephemeral'))
命中缓存Token数： 0
用于创建缓存Token数： 2065
第2轮
====================回复内容====================
对于这个红黑树实现，各个操作的时间复杂度如下：<省略后续回答>
====================Token消耗====================
CompletionUsage(completion_tokens=864, prompt_tokens=2905, total_tokens=3769, completion_tokens_details=None, prompt_tokens_details=PromptTokensDetails(audio_tokens=None, cached_tokens=2065, cache_creation={'ephemeral_5m_input_tokens': 836}, cache_creation_input_tokens=836, cache_type='ephemeral'))
命中缓存Token数： 2065
用于创建缓存Token数： 836

import os
import dashscope
import json


def get_response(messages):
    response = dashscope.Generation.call(
        # 若没有配置环境变量，请用阿里云百炼API Key将下行替换为：api_key="sk-xxx",
        api_key=os.getenv('DASHSCOPE_API_KEY'),
        model="qwen3-coder-plus",
        messages=messages,
        # result_format参数不可以设置为"text"。
        result_format='message'
    )
    return response


# 确保输入文本超过1024Token被缓存
red_black_tree = '<完整的红黑树代码实现>'
messages = [
    {"role": "system", "content": "你是一个有用的代码助手"},
    {
        "role": "user",
        "content": [
            {
                "type": "text",
                "text": red_black_tree,
            },
            {
                "type": "text",
                "text": "该代码主要实现了什么功能",
                "cache_control": {"type": "ephemeral"},
            }
        ],
    }
]

response = get_response(messages)
print("第1轮")
print("=" * 20 + "回复内容" + "=" * 20)
print(response.output.choices[0].message.content)
print("=" * 20 + "Token消耗" + "=" * 20)
print(json.dumps(response.usage))
print(f"命中缓存Token数：{response.usage.prompt_tokens_details['cached_tokens']}")
print(f"用于创建缓存Token数：{response.usage.prompt_tokens_details['cache_creation_input_tokens']}")
assistant_output = response.output.choices[0].message.content

messages = [
    {"role": "system", "content": "你是一个有用的代码助手"},
    {
        "role": "user",
        "content": [
            {
                "type": "text",
                "text": red_black_tree,
            },
            {
                "type": "text",
                "text": "该代码主要实现了什么功能",
            }
        ],
    },
    {"role": "assistant", "content": assistant_output},
    {
        "role": "user",
        "content": [
            {
                "type": "text",
                "text": "各个操作的时间复杂度是多少",
                "cache_control": {"type": "ephemeral"},
            }
        ]
    }
]

response= get_response(messages)
print(json.dumps(response))
print("第2轮")
print("=" * 20 + "回复内容" + "=" * 20)
print(response.output.choices[0].message.content)
print("=" * 20 + "Token消耗" + "=" * 20)
print(json.dumps(response.usage))
print(
f"命中缓存Token数：{response.usage.prompt_tokens_details['cached_tokens']}")
print(
f"用于创建缓存Token数：{response.usage.prompt_tokens_details['cache_creation_input_tokens']}")

第1轮
====================回复内容====================
这段 Python 代码完整地实现了一个 **红黑树（Red-Black Tree）**，这是一种自平衡的二叉搜索树。<省略后续回答>
====================Token消耗====================
{"input_tokens": 2069, "output_tokens": 757, "total_tokens": 2826, "prompt_tokens_details": {"cache_creation": {"ephemeral_5m_input_tokens": 2065}, "cache_creation_input_tokens": 2065, "cache_type": "ephemeral", "cached_tokens": 0}, "cached_tokens": 0}
命中缓存Token数：0
用于创建缓存Token数：2065
第2轮
====================回复内容====================
## 红黑树各操作的时间复杂度分析<省略后续回答>
====================Token消耗====================
{"input_tokens": 2842, "output_tokens": 1000, "total_tokens": 3842, "prompt_tokens_details": {"cache_creation": {"ephemeral_5m_input_tokens": 773}, "cache_creation_input_tokens": 773, "cache_type": "ephemeral", "cached_tokens": 2065}, "cached_tokens": 2065}
命中缓存Token数：2065
用于创建缓存Token数：773