tech
Defaultdeesfrzh

LangChain

LLM 应用框架 - 链接调用、连接数据源、Agent、记忆管理

TL;DR

是什么：用于构建大语言模型应用的框架。

为什么用：链接 LLM 调用、连接数据源、Agent、记忆管理。

Quick Start

安装：

pip install langchain langchain-openai

Hello LangChain：

from langchain_openai import ChatOpenAI
from langchain_core.messages import HumanMessage

# 初始化模型
llm = ChatOpenAI(model="gpt-4o-mini")

# 简单调用
response = llm.invoke([HumanMessage(content="Hello!")])
print(response.content)

Cheatsheet

概念描述
LLM语言模型封装
Prompt输入模板
Chain链接操作
Agent动态决策器
Memory对话历史
Retriever文档检索

Gotchas

提示模板

from langchain_core.prompts import ChatPromptTemplate

# 简单模板
prompt = ChatPromptTemplate.from_messages([
    ("system", "You are a helpful assistant."),
    ("user", "{input}")
])

# 创建链
chain = prompt | llm

# 调用
response = chain.invoke({"input": "What is Python?"})
print(response.content)

输出解析器

from langchain_core.output_parsers import StrOutputParser, JsonOutputParser
from langchain_core.pydantic_v1 import BaseModel, Field

# 字符串解析器
chain = prompt | llm | StrOutputParser()
result = chain.invoke({"input": "Hello"})  # 返回字符串

# 带 schema 的 JSON 解析器
class Joke(BaseModel):
    setup: str = Field(description="The setup")
    punchline: str = Field(description="The punchline")

parser = JsonOutputParser(pydantic_object=Joke)
prompt = ChatPromptTemplate.from_template(
    "Tell me a joke.\n{format_instructions}"
)
chain = prompt | llm | parser

RAG（检索增强生成）

from langchain_community.vectorstores import FAISS
from langchain_openai import OpenAIEmbeddings
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.runnables import RunnablePassthrough

# 创建向量存储
embeddings = OpenAIEmbeddings()
texts = ["LangChain is a framework", "It helps build LLM apps"]
vectorstore = FAISS.from_texts(texts, embeddings)
retriever = vectorstore.as_retriever()

# RAG 链
template = """Answer based on context:
{context}

Question: {question}
"""
prompt = ChatPromptTemplate.from_template(template)

chain = (
    {"context": retriever, "question": RunnablePassthrough()}
    | prompt
    | llm
    | StrOutputParser()
)

response = chain.invoke("What is LangChain?")

Agent

from langchain.agents import create_tool_calling_agent, AgentExecutor
from langchain_core.tools import tool

@tool
def search(query: str) -> str:
    """Search for information."""
    return f"Results for: {query}"

@tool
def calculator(expression: str) -> str:
    """Calculate math expression."""
    return str(eval(expression))

tools = [search, calculator]

prompt = ChatPromptTemplate.from_messages([
    ("system", "You are a helpful assistant."),
    ("human", "{input}"),
    ("placeholder", "{agent_scratchpad}")
])

agent = create_tool_calling_agent(llm, tools, prompt)
executor = AgentExecutor(agent=agent, tools=tools)

result = executor.invoke({"input": "What is 25 * 4?"})

记忆

from langchain_core.chat_history import InMemoryChatMessageHistory
from langchain_core.runnables.history import RunnableWithMessageHistory

store = {}

def get_session_history(session_id: str):
    if session_id not in store:
        store[session_id] = InMemoryChatMessageHistory()
    return store[session_id]

chain_with_history = RunnableWithMessageHistory(
    chain,
    get_session_history,
    input_messages_key="input",
    history_messages_key="history"
)

response = chain_with_history.invoke(
    {"input": "My name is John"},
    config={"configurable": {"session_id": "user123"}}
)

Next Steps