使用代理混合搜索让你的RAG应用程序成为智能化“助手” 原创

from langchain_core.runnables import (
   RunnablePassthrough,
   ConfigurableField,
)
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI, OpenAIEmbeddings
from langchain_astradb.graph_vectorstores import AstraDBVectorStore

llm = ChatOpenAI()
embeddings = OpenAIEmbeddings()
vectorstore = AstraDBVectorStore(
   collection_name="knowledge_store",
   embedding=embeddings,
)

ANSWER_PROMPT = (
   "Use the information in the results to provide a concise answer the original question.\n\n"
   "Original Question: {question}\n\n"
   "Vector Store Results:\n{'\n\n'.join(c.page_content for c in context)}\n\n"
)

retriever = vectorstore.as_retriever()

# Construct the LLM execution chain
chain = (
   {"context": retriever, "question": RunnablePassthrough()}
   | ChatPromptTemplate.from_messages([ANSWER_PROMPT])
   | llm
)

chain.invoke("How much is my annual membership fee?",)

# > Response: Your annual membership fee could be $250, $95, $695, or $325, depending on the specific plan or card you have chosen. Please refer to your specific card member agreement or plan details to confirm the exact amount of your annual membership fee.

from typing import List, Literal
from pydantic import BaseModel, Field
from langchain_core.documents.base import Document
from langchain_core.tools import StructuredTool
from langchain.agents import AgentExecutor, create_tool_calling_agent
from langchain_core.prompts import MessagesPlaceholder
prompt = ChatPromptTemplate.from_messages([
   ("system", "Concisely answer the following question, using information retrieved from tools and the provided information about the user."),
   ("system", "The following card types are associated with the user: {cards}"),
   ("system", "Always use the provided tools to retrieve information needed to answer policy-related questions."),
   ("human", "{question}"),
   MessagesPlaceholder("agent_scratchpad"),
])
# First we define the parameters to our search operation
class RetrieveInput(BaseModel):
   question: str = Field(description="Question to retrieve content for. Should be a simple question describing the starting point for retrieval likely to have content.")
   card_type: str = Field(description=f"Search for documents related to this card type. Value must be one of {pages.keys()}")

# Next, create a "tool" that implements the search logic
def retrieve_policy(question: str, card_type: str) -> List[Document]:
   print(f"retrieve_policy(card_type: {card_type}, question: {question})")
   retriever = graph_vectorstore.as_retriever(
     search_type = "similarity",
     search_kwargs = {"metadata_filter": {"card-type": card_type}},
   )
   return list(retriever.invoke(question))

policy_tool = StructuredTool.from_function(
   func=retrieve_policy,
   name="RetrievePolicy",
   description="Retrieve information about a specific card policy.",
   args_schema=RetrieveInput,
   return_direct=False,
)

# Finally, construct an agent to use the tool we created
agent = create_tool_calling_agent(llm, [policy_tool], prompt)
agent_executor = AgentExecutor(agent=agent, tools=[policy_tool], verbose=True)

agent_executor.invoke({
   "question": "How much is my annual membership fee?",
   "cards": ["gold"],
})

# > Agent: Invoking: `RetrievePolicy` with `{'question': 'annual membership fee', 'card_type': 'gold'}`

# > Response: Your annual membership fee could be $250, $95, $695, or $325, depending on the specific plan or card you have chosen. Please refer to your specific card member agreement or plan details to confirm the exact amount of your annual membership fee.

agent_executor.invoke({
   "question": "What's the annual membership fee for platinum cards?",
   "cards": ["gold"],
})

# > Agent: Invoking: `RetrievePolicy` with `{'question': 'annual membership fee for platinum cards', 'card_type': 'platinum'}`

# > Response: The annual membership fee for Platinum cards is $695. Additionally, each Additional Platinum Card has an annual fee of $195, but there is no annual fee for Companion Platinum Cards.

agent_executor.invoke({
   "question": "How much would my membership fee change if I upgraded to a platinum card?",
   "cards": ["gold"],
})

# > Agent: Invoking: `RetrievePolicy` with `{'question': 'membership fee for gold card', 'card_type': 'gold'}`
# > Agent: Invoking: `RetrievePolicy` with `{'question': 'membership fee for platinum card', 'card_type': 'platinum'}`

# > Response: The annual membership fee for your current American Express® Gold Card is $325. If you were to upgrade to a Platinum Card, the annual fee would be $695. Therefore, upgrading from a Gold Card to a Platinum Card would increase your annual membership fee by $370.