LangChain 接入大模型 API

LangChain 是最流行的 LLM 应用框架。本指南演示用 LangChain 调用统一网关的多模型。

Python

安装

pip install langchain langchain-openai

调用代码

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    model="claude-opus-4-7",                  # 任意模型 ID
    openai_api_key="sk-xxx",
    openai_api_base="http://xdhdancer.top/v1",
    temperature=0.3,
)

# 直接调用
resp = llm.invoke("Explain quantum tunneling in one paragraph.")
print(resp.content)

# 链式
from langchain_core.prompts import ChatPromptTemplate
prompt = ChatPromptTemplate.from_messages([
    ("system", "You are a helpful assistant."),
    ("user", "{question}"),
])
chain = prompt | llm
print(chain.invoke({"question": "What is RAG?"}).content)

流式输出

for chunk in llm.stream("Write a short poem about Saturday"):
    print(chunk.content, end="", flush=True)

JavaScript / TypeScript

安装

npm install @langchain/openai @langchain/core

调用

import { ChatOpenAI } from "@langchain/openai";

const llm = new ChatOpenAI({
  model: "claude-opus-4-7",
  apiKey: "sk-xxx",
  configuration: {
    baseURL: "http://xdhdancer.top/v1",
  },
});

const resp = await llm.invoke("Hello, world!");
console.log(resp.content);

多模型场景：Router

from langchain_openai import ChatOpenAI

# 简单/便宜
fast = ChatOpenAI(model="gpt-5-mini", openai_api_base="http://xdhdancer.top/v1", openai_api_key="sk-xxx")
# 复杂/精准
strong = ChatOpenAI(model="claude-opus-4-7", openai_api_base="http://xdhdancer.top/v1", openai_api_key="sk-xxx")

def route(question: str):
    if len(question) > 200 or "code" in question.lower():
        return strong.invoke(question).content
    return fast.invoke(question).content

常见问题

Q: 用 from langchain.chat_models import ChatOpenAI（旧路径）

A: LangChain 0.1+ 把 OpenAI 拆到独立包。改用 from langchain_openai import ChatOpenAI。

Q: 流式回调没触发

A: 确认安装 langchain >= 0.1.0，且代码里用 llm.stream(...) 不是 llm.invoke(...)。