\u672c\u6587\u6df1\u5165\u63a2\u8ba8 LangChain \u7684\u6838\u5fc3\u6982\u5ff5\uff0c\u7406\u89e3\u8fd9\u4e9b\u6982\u5ff5\u662f\u638c\u63e1 LangChain \u7684\u5173\u952e\u3002<\/p>\n
## 1. Runnable Interface
\nRunnable \u662f LangChain \u4e2d\u6700\u91cd\u8981\u7684\u63a5\u53e3\uff0c\u51e0\u4e4e\u6240\u6709\u7684\u7ec4\u4ef6\u90fd\u5b9e\u73b0\u4e86\u8fd9\u4e2a\u63a5\u53e3\u3002<\/p>\n
### \u7edf\u4e00\u8c03\u7528\u65b9\u5f0f
\n“`python
\nfrom langchain_openai import ChatOpenAI
\nfrom langchain_core.prompts import ChatPromptTemplate<\/p>\n
llm = ChatOpenAI(model=”gpt-4o-mini”)
\nprompt = ChatPromptTemplate.from_template(“\u56de\u7b54: {question}”)<\/p>\n
response = llm.invoke(“\u4f60\u597d”)
\nresult = prompt.invoke({“question”: “\u4ec0\u4e48\u662f AI\uff1f”})
\n“`<\/p>\n
### \u6d41\u5f0f\u8f93\u51fa
\n“`python
\nfor chunk in llm.stream(“\u8bb2\u4e2a\u7b11\u8bdd”):
\n print(chunk.content, end=””, flush=True)
\n“`<\/p>\n
### \u6279\u5904\u7406
\n“`python
\nquestions = [“\u4ec0\u4e48\u662f Python\uff1f”, “\u4ec0\u4e48\u662f JavaScript\uff1f”, “\u4ec0\u4e48\u662f AI\uff1f”]
\nresponses = llm.batch(questions)
\n“`<\/p>\n
## 2. \u94fe\u5f0f\u7ec4\u5408
\n\u4f7f\u7528 | \u64cd\u4f5c\u7b26\u53ef\u4ee5\u5c06\u591a\u4e2a\u7ec4\u4ef6\u4e32\u8054\u8d77\u6765\uff1a<\/p>\n
### \u7b80\u5355\u94fe
\n“`python
\nchain = (
\n ChatPromptTemplate.from_template(“\u7ffb\u8bd1\u6210\u4e2d\u6587: {text}”)
\n | ChatOpenAI(model=”gpt-4o-mini”)
\n | StrOutputParser()
\n)
\nresult = chain.invoke({“text”: “Hello, how are you?”})
\n“`<\/p>\n
### \u590d\u6742\u94fe
\n“`python
\nextractor = ChatPromptTemplate.from_template(“\u63d0\u53d6\u5173\u952e\u4fe1\u606f\uff1a{text}”) | llm | StrOutputParser()
\nsummarizer = ChatPromptTemplate.from_template(“\u603b\u7ed3\uff1a{content}”) | llm | StrOutputParser()
\nchain = extractor | summarizer
\n“`<\/p>\n
## 3. LangChain Expression Language (LCEL)
\nLCEL \u662f\u4e00\u79cd\u58f0\u660e\u5f0f\u65b9\u5f0f\u6765\u6784\u5efa\u94fe\uff0c\u66f4\u7075\u6d3b\u548c\u53ef\u8bfb\u3002<\/p>\n
### \u53ef\u7ec4\u5408\u6027
\n“`python
\ntranslator = ChatPromptTemplate.from_template(“\u7ffb\u8bd1\u6210{lang}: {text}”) | llm | StrOutputParser()
\nchinese = translator.invoke({“lang”: “\u4e2d\u6587”, “text”: “Hello”})
\nenglish = translator.invoke({“lang”: “\u82f1\u6587”, “text”: “\u4f60\u597d”})
\n“`<\/p>\n
### \u5e76\u884c\u5904\u7406
\n“`python
\nfrom langchain_core.runnables import RunnableParallel<\/p>\n
parallel_chain = RunnableParallel({
\n “summary”: summarizer,
\n “translation”: translator
\n})
\nresults = parallel_chain.invoke({“text”: “\u4eca\u5929\u5929\u6c14\u4e0d\u9519”, “lang”: “\u82f1\u6587”})
\n“`<\/p>\n
## 4. \u8bb0\u5fc6\u7ba1\u7406
\n– **\u5bf9\u8bdd\u7a97\u53e3\u8bb0\u5fc6\uff1a** \u4fdd\u5b58\u6700\u8fd1\u7684\u6d88\u606f\u5386\u53f2
\n– **\u4f1a\u8bdd\u6458\u8981\u8bb0\u5fc6\uff1a** \u81ea\u52a8\u603b\u7ed3\u5386\u53f2\uff0c\u8282\u7701 token<\/p>\n
## 5. \u5de5\u5177\u8c03\u7528
\n“`python
\nfrom langchain_core.tools import tool<\/p>\n
@tool
\ndef calculator(expression: str) -> str:
\n “””\u8ba1\u7b97\u6570\u5b66\u8868\u8fbe\u5f0f”””
\n try:
\n result = eval(expression)
\n return f”\u7ed3\u679c: {result}”
\n except Exception as e:
\n return f”\u9519\u8bef: {str(e)}”<\/p>\n
llm = ChatOpenAI(model=”gpt-4o”).bind_tools([calculator])
\nresponse = llm.invoke(“\u8ba1\u7b97 123 + 456″)
\n“`<\/p>\n
## 6. \u68c0\u7d22\u589e\u5f3a\u751f\u6210\uff08RAG\uff09
\n“`python
\nfrom langchain_core.documents import Document
\nfrom langchain_chroma import Chroma
\nfrom langchain_openai import OpenAIEmbeddings<\/p>\n
documents = [
\n Document(page_content=”LangChain \u662f\u4e00\u4e2a LLM \u5e94\u7528\u6846\u67b6”),
\n Document(page_content=”Python \u662f\u4e00\u79cd\u7f16\u7a0b\u8bed\u8a00”),
\n]
\nembeddings = OpenAIEmbeddings()
\nvectorstore = Chroma.from_documents(documents, embeddings)
\nretriever = vectorstore.as_retriever(search_kwargs={“k”: 2})
\n“`<\/p>\n","protected":false},"excerpt":{"rendered":"
\u672c\u6587\u6df1\u5165\u63a2\u8ba8 LangChain \u7684\u6838\u5fc3\u6982\u5ff5\uff0c\u7406\u89e3\u8fd9\u4e9b\u6982\u5ff5\u662f\u638c\u63e1 LangChain \u7684\u5173\u952e\u3002 ## 1. R […]<\/p>\n","protected":false},"author":0,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"zakra_page_container_layout":"customizer","zakra_page_sidebar_layout":"customizer","zakra_remove_content_margin":false,"zakra_sidebar":"customizer","zakra_transparent_header":"customizer","zakra_logo":0,"zakra_main_header_style":"default","zakra_menu_item_color":"","zakra_menu_item_hover_color":"","zakra_menu_item_active_color":"","zakra_menu_active_style":"","zakra_page_header":true,"footnotes":""},"categories":[7],"tags":[],"class_list":["post-14","post","type-post","status-publish","format-standard","hentry","category-langchain"],"_links":{"self":[{"href":"https:\/\/gmrea.com\/index.php?rest_route=\/wp\/v2\/posts\/14","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/gmrea.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/gmrea.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"replies":[{"embeddable":true,"href":"https:\/\/gmrea.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=14"}],"version-history":[{"count":0,"href":"https:\/\/gmrea.com\/index.php?rest_route=\/wp\/v2\/posts\/14\/revisions"}],"wp:attachment":[{"href":"https:\/\/gmrea.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=14"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/gmrea.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=14"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/gmrea.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=14"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}