三大集成模式揭秘：解锁检索增强智能体的核心策略​

# RAG基础组件链
from langchain.document_loaders import WebBaseLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.embeddings import OpenAIEmbeddings
from langchain.vectorstores import FAISS

# 1. 文档加载与处理
loader = WebBaseLoader("https://example.com")
docs = loader.load()
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000)
documents = text_splitter.split_documents(docs)

# 2. 向量化存储
embeddings = OpenAIEmbeddings()
vectorstore = FAISS.from_documents(documents, embeddings)

# 3. 智能体集成
retriever = vectorstore.as_retriever(search_kwargs={"k": 5})

from langchain.retrievers import ContextualCompressionRetriever
from langchain.retrievers.document_compressors import LLMChainExtractor

compressor = LLMChainExtractor.from_llm(ChatOpenAI(temperature=0))
compression_retriever = ContextualCompressionRetriever(
    base_compressor=compressor,
    base_retriever=vectorstore.as_retriever()
)
# 压缩后文档体积减少40%-60%

# 结合语义与关键词检索
from langchain.retrievers import BM25Retriever, EnsembleRetriever

bm25_retriever = BM25Retriever.from_documents(documents)
ensemble_retriever = EnsembleRetriever(
    retrievers=[
        vectorstore.as_retriever(search_type="mmr"), 
        bm25_retriever
    ],
    weights=[0.6, 0.4]
)

def iterative_retrieval(query, max_depth=3):
    context = ""
    for i in range(max_depth):
        docs = retriever.get_relevant_documents(query)
        context += "\n".join(d.page_content for d in docs)
        
        if sufficient_information(context, query):
            break
            
        # 生成深化问题
        prompt = f"""基于当前信息:{context[:2000]}...
        生成更精准的子问题（需满足：{query}）"""
        query = llm.predict(prompt)  
    return generate_final_answer(context, query)

def knowledge_update_agent(new_data):
    # 信息价值评估
    evaluation_prompt = f"""评估信息价值（1-10分）：
    {new_data}
    评估维度：新颖性、准确性、完整性"""
    score = int(llm.predict(evaluation_prompt))
    
    if score >= 8:
        # 知识结构化处理
        structuring_prompt = f"""将信息转化为知识条目：
        - 保留核心事实
        - 删除冗余描述
        - 标准化术语
        原始信息：{new_data}"""
        structured_knowledge = llm.predict(structuring_prompt)
        vectorstore.add_documents([Document(structured_knowledge)])

# 查询重写技术
def query_rewriter(original_query):
    return llm.predict(f"将用户查询优化为检索友好格式：{original_query}")

# 结果过滤管道
from langchain.retrievers import TFIDFFilter
filtered_retriever = TFIDFFilter(retriever, threshold=0.7)