langchain-tutorials

--- name: langchain-tutorials description: "Туториалы LangChain/LangGraph: RAG Agent, SQL Agent, Voice Agent, Semantic Search, Multi-Agent (субагенты, handoffs, router, skills), Custom RAG/SQL с LangGraph. Триггеры: 'tutorial langchain', 'туториал', 'RAG agent tutorial', 'SQL agent tutorial', 'voice agent', 'semantic search tutorial', 'multi-agent tutorial', 'пример агента', 'agent example', 'LangChain quickstart', 'быстрый старт langchain', 'как построить RAG', 'как построить SQL агента', 'custom RAG agent', 'custom SQL agent', 'пошаговый пример', 'step by step agent'. НЕ для API reference — используй langchain-core/langgraph-core." --- # LangChain / LangGraph Tutorials ## RAG Agent (LangChain) Агент с доступом к документации, самостоятельно решает когда искать. ```python from langchain.agents import create_agent from langchain.tools import tool from langchain_chroma import Chroma from langchain_openai import OpenAIEmbeddings # 1. Vector store + retriever vectorstore = Chroma( embedding_function=OpenAIEmbeddings(), persist_directory="./docs_db" ) retriever = vectorstore.as_retriever(search_kwargs={"k": 5}) # 2. Tool @tool def search_docs(query: str) -> str: """Search documentation for relevant information.""" docs = retriever.invoke(query) return "\n\n".join([d.page_content for d in docs]) # 3. Agent agent = create_agent( model="gpt-4.1", tools=[search_docs], system_prompt=( "You answer questions using documentation. " "Always search docs before answering." ) ) result = agent.invoke({ "messages": [{"role": "user", "content": "How to configure logging?"}] }) ``` --- ## SQL Agent (LangChain) Агент выполняет SQL-запросы с проверкой и обработкой ошибок. ```python from langchain.agents import create_agent from langchain_community.utilities import SQLDatabase from langchain_community.tools.sql_database.tool import ( QuerySQLDataBaseTool, InfoSQLDatabaseTool, ListSQLDatabaseTool, QuerySQLCheckerTool ) # 1. Database db = SQLDatabase.from_uri("sqlite:///chinook.db") # 2. Tools tools = [ ListSQLDatabaseTool(db=db), # Список таблиц InfoSQLDatabaseTool(db=db), # Схема таблиц QuerySQLCheckerTool(db=db, llm=model),# Проверка SQL QuerySQLDataBaseTool(db=db) # Выполнение SQL ] # 3. Agent agent = create_agent( model="gpt-4.1", tools=tools, system_prompt=( "You are a SQL expert. Workflow:\n" "1. List tables\n" "2. Get schema of relevant tables\n" "3. Write SQL query\n" "4. Check query with checker tool\n" "5. Execute query\n" "6. Present results" ) ) ``` ### SQL Agent + HITL ```python from langchain.agents.middleware import HumanInTheLoopMiddleware agent = create_agent( model="gpt-4.1", tools=tools, middleware=[ HumanInTheLoopMiddleware( interrupt_on={"sql_db_query": True} # Одобрение перед execute ) ], checkpointer=InMemorySaver() ) ``` ### Безопасность SQL - **READ-only permissions** на базу данных - **Checker tool** перед выполнением - **HITL** для опасных запросов (DELETE, UPDATE) --- ## Semantic Search (LangChain) Полный pipeline: загрузка → split → embed → search. ```python from langchain_community.document_loaders import WebBaseLoader from langchain_text_splitters import RecursiveCharacterTextSplitter from langchain_openai import OpenAIEmbeddings from langchain_chroma import Chroma # 1. Load loader = WebBaseLoader("https://docs.example.com/guide") docs = loader.load() # 2. Split splitter = RecursiveCharacterTextSplitter( chunk_size=1000, chunk_overlap=200 ) chunks = splitter.split_documents(docs) # 3. Embed + Store vectorstore = Chroma.from_documents( chunks, OpenAIEmbeddings(), persist_directory="./search_db" ) # 4. Search results = vectorstore.similarity_search("authentication setup", k=3) for doc in results: print(doc.page_content[:200]) ``` --- ## Custom RAG Agent (LangGraph) RAG с query rewriting и adaptive retrieval через StateGraph. ```python from langgraph.graph import StateGraph, START, END from typing import TypedDict, Literal class RAGState(TypedDict): query: str rewritten_query: str documents: list[str] answer: str needs_more: bool def rewrite_query(state: RAGState) -> dict: rewritten = model.invoke( f"Rewrite for better search: {state['query']}" ) return {"rewritten_query": rewritten.content} def retrieve(state: RAGState) -> dict: docs = retriever.invoke(state["rewritten_query"]) return {"documents": [d.page_content for d in docs]} def generate(state: RAGState) -> dict: context = "\n\n".join(state["documents"]) response = model.invoke( f"Context:\n{context}\n\nQuestion: {state['query']}" ) return {"answer": response.content} def check_quality(state: RAGState) -> Literal["retrieve_more", "done"]: if "I don't know" in state["answer"]: return "retrieve_more" return "done" builder = StateGraph(RAGState) builder.add_node("rewrite", rewrite_query) builder.add_node("retrieve", retrieve) builder.add_node("generate", generate) builder.add_edge(START, "rewrite") builder.add_edge("rewrite", "retrieve") builder.add_edge("retrieve", "generate") builder.add_conditional_edges("generate", check_quality, { "retrieve_more": "retrieve", "done": END }) rag_agent = builder.compile() result = rag_agent.invoke({"query": "How to set up auth?"}) ``` --- ## Custom SQL Agent (LangGraph) SQL агент с retry loop через граф. ```python class SQLState(TypedDict): query: str sql: str result: str error: str | None attempts: int def generate_sql(state: SQLState) -> dict: prompt = f"Write SQL for: {state['query']}" if state.get("error"): prompt += f"\nPrevious error: {state['error']}" sql = model.invoke(prompt) return {"sql": sql.content, "attempts": state.get("attempts", 0) + 1} def execute_sql(state: SQLState) -> dict: try: result = db.run(state["sql"]) return {"result": str(result), "error": None} except Exception as e: return {"result": "", "error": str(e)} def should_retry(state: SQLState) -> Literal["retry", "respond"]: if state.get("error") and state.get("attempts", 0) < 3: return "retry" return "respond" builder = StateGraph(SQLState) builder.add_node("generate", generate_sql) builder.add_node("execute", execute_sql) builder.add_node("respond", format_response) builder.add_edge(START, "generate") builder.add_edge("generate", "execute") builder.add_conditional_edges("execute", should_retry, { "retry": "generate", "respond": "respond" }) builder.add_edge("respond", END) sql_agent = builder.compile() ``` --- ## Multi-Agent Tutorials ### Subagents (supervisor) ```python # Координатор делегирует задачи @tool def research(query: str) -> str: return research_agent.invoke(...)["messages"][-1].content @tool def write(content: str) -> str: return writer_agent.invoke(...)["messages"][-1].content supervisor = create_agent(model="gpt-4.1", tools=[research, write]) ``` ### Handoffs ```python # Агент передаёт управление через state @tool def transfer_to_billing(runtime: ToolRuntime) -> Command: return Command(goto="billing_agent", update={"active": "billing"}, graph=Command.PARENT) ``` ### Router ```python # Классификация → параллельная обработка → синтез builder.add_edge(START, "classify") builder.add_conditional_edges("classify", route, ["tech", "billing"]) builder.add_edge("tech", "synthesize") builder.add_edge("billing", "synthesize") ``` ### Skills ```python # Динамическая загрузка специализированных промптов @tool def load_skill(name: str) -> str: return SKILLS[name]["prompt"] ``` --- ## Voice Agent (LangChain) Голосовой агент через speech-to-text → LLM → text-to-speech. ```python # Whisper (speech-to-text) → Agent → TTS from langchain_community.tools import WhisperTool, TTSTool agent = create_agent( model="gpt-4.1", tools=[whisper_tool, tts_tool, search_docs], system_prompt="Voice assistant. Listen → Process → Respond with speech." ) ``` --- ## Поддерживаемые модели (все tutorials) | Провайдер | Модель | Пакет | |-----------|--------|-------| | OpenAI | gpt-4.1, gpt-5 | `langchain-openai` | | Anthropic | claude-sonnet-4-5-20250929 | `langchain-anthropic` | | Google | Gemini | `langchain-google-genai` | | AWS | Bedrock models | `langchain-aws` | | HuggingFace | Open models | `langchain-huggingface` | | Azure | Azure OpenAI | `langchain-openai` | --- **Источники:** Учиться/ — 10+ файлов (RAG Agent, SQL Agent, Voice Agent, Semantic Search, Multi-Agent tutorials, Custom RAG/SQL, Концептуальные обзоры)