Agentic RAG：AI代理如何变革检索增强生成

用户查询 → 嵌入查询 → 向量搜索 → Top-K分块 → LLM → 回答

用户查询 → 代理规划 → 搜索工具 → 评估结果 →
  → 需要更多信息？ → 优化查询 → 再次搜索 →
  → 足够的上下文？ → 综合 → 回答

# 概念性Agentic RAG循环（伪代码）
def agentic_rag(query, tools, max_iterations=5):
    context = []
    plan = agent.plan(query)  # 分解为子问题
 
    for step in plan:
        # 代理决定使用哪个工具
        tool = agent.select_tool(step, tools)
        results = tool.execute(step.query)
        context.extend(results)
 
        # 代理评估是否有足够的信息
        if agent.has_sufficient_context(query, context):
            break
 
        # 代理根据学到的内容优化查询
        step.query = agent.refine_query(step, results)
 
    return agent.synthesize(query, context)

原始问题："比较Apple和Microsoft 2025年第三季度的收入增长，
           并解释哪家公司的营业利润率更高"

分解后：
  1. "Apple 2025年Q3收入增长"
  2. "Microsoft 2025年Q3收入增长"
  3. "Apple 2025年Q3营业利润率"
  4. "Microsoft 2025年Q3营业利润率"

from langchain.agents import create_react_agent
from langchain_openai import ChatOpenAI
from langchain.tools import Tool
from langchain_community.vectorstores import FAISS
from langchain_openai import OpenAIEmbeddings
 
# 创建检索工具
vectorstore = FAISS.from_texts(documents, OpenAIEmbeddings())
retriever = vectorstore.as_retriever(search_kwargs={"k": 5})
 
def search_docs(query: str) -> str:
    results = retriever.invoke(query)
    return "\n".join([doc.page_content for doc in results])
 
tools = [
    Tool(
        name="SearchDocuments",
        func=search_docs,
        description="Search the knowledge base for relevant information"
    ),
]
 
llm = ChatOpenAI(model="gpt-4o", temperature=0)
agent = create_react_agent(llm, tools, prompt_template)

from llama_index.core.agent import ReActAgent
from llama_index.core.tools import QueryEngineTool
from llama_index.core import VectorStoreIndex
 
# 创建查询引擎工具
index = VectorStoreIndex.from_documents(documents)
query_engine = index.as_query_engine(similarity_top_k=5)
 
tools = [
    QueryEngineTool.from_defaults(
        query_engine=query_engine,
        name="knowledge_base",
        description="Search the company knowledge base"
    ),
]
 
agent = ReActAgent.from_tools(tools, llm=llm, verbose=True)
response = agent.chat("What were our Q3 results?")