PythonSDKLangChainRAG

Python SDK 接入教程

用 Python openai SDK 接入 ChinaWHAPI，支持同步、流式、Function Calling、LangChain 和 RAG 全场景。

安装依赖

ChinaWHAPI 完全兼容 OpenAI Python SDK，直接安装 openai 包即可。

pip install openai tiktoken

基础调用

配置 base_url 指向 ChinaWHAPI 端点，使用方式和 OpenAI 完全一致。

from openai import OpenAI

client = OpenAI(
    api_key="your_chinawhapi_key",
    base_url="https://chinawhapi.com/v1"
)

# 同步调用
response = client.chat.completions.create(
    model="qwen3.6-plus",
    messages=[{"role": "user", "content": "解释什么是微服务架构"}]
)
print(response.choices[0].message.content)

流式输出

设置 stream=True 实现打字机效果，适合聊天界面和实时生成场景。

stream = client.chat.completions.create(
    model="qwen3.6-plus",
    messages=[{"role": "user", "content": "写一个 Python FastAPI 示例"}],
    stream=True,
)
for chunk in stream:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)

Function Calling / Tool Use

定义 tools 数组，模型会选择性调用指定工具，实现 Agent 工作流。

response = client.chat.completions.create(
    model="qwen3.6-plus",
    messages=[{"role": "user", "content": "北京现在多少度？"}],
    tools=[
        {
            "type": "function",
            "function": {
                "name": "get_weather",
                "description": "获取指定城市的天气",
                "parameters": {
                    "type": "object",
                    "properties": {"city": {"type": "string", "description": "城市名称"}},
                    "required": ["city"]
                }
            }
        }
    ]
)
print(response.choices[0].message.tool_calls)

LangChain 集成

在 LangChain 中使用 ChatOpenAI 并设置 openai_api_base，即可使用 ChinaWHAPI 所有模型。

from langchain.chat_models import ChatOpenAI
from langchain.schema import HumanMessage

llm = ChatOpenAI(
    model="qwen3.6-plus",
    openai_api_key="your_chinawhapi_key",
    openai_api_base="https://chinawhapi.com/v1",
    temperature=0.7,
)

response = llm([HumanMessage(content="用 Python 写一个快速排序")])
print(response.content)

LlamaIndex / RAG 集成

LlamaIndex 通过 OpenAILike 接口接入 ChinaWHAPI，适合构建知识库问答系统。

from llama_index.llms.openai_like import OpenAILike
from llama_index import VectorStoreIndex, SimpleDirectoryReader

llm = OpenAILike(
    model="qwen3.6-plus",
    api_key="your_chinawhapi_key",
    api_base="https://chinawhapi.com/v1",
    is_chat_model=True,
)

documents = SimpleDirectoryReader("./docs").load_data()
index = VectorStoreIndex.from_documents(documents)

query_engine = index.as_query_engine(llm=llm)
response = query_engine.query("总结这份文档的核心内容")
print(response)

Token 计数

使用 tiktoken 精确计算 token 量，便于成本估算和 Prompt 优化。

import tiktoken

enc = tiktoken.encoding_for_model("gpt-4o")
text = "这是一段中文文本，大约几十个字"
tokens = enc.encode(text)
print(f"Token 数: {len(tokens)}")

错误处理与重试

实现带指数退避的重试机制，保证生产环境的稳定性。

import time
from openai import RateLimitError, APIError

def call_with_retry(client, model, messages, max_retries=3):
    for attempt in range(max_retries):
        try:
            return client.chat.completions.create(model=model, messages=messages)
        except RateLimitError:
            wait_time = 2 ** attempt
            time.sleep(wait_time)
        except APIError as e:
            if e.status_code >= 500:
                continue  # 5xx 错误直接重试
            raise
    raise Exception("Max retries exceeded")