Agentic AI – Customer Service Agents
Table Of Contents:
- What Is Customer Service Agents ?
- What Tasks Customer Service Agents Can Perform ?
- Steps To Build Customer Service Agents.
- Python Implementation.
(1) What Is Customer Service Agents ?
(2) What Tasks Customer Service Agents Can Perform ?
(3) Steps To Implement Customer Service Agents.
(4) Python Implementation
(1) Set Up the Environment
pip install openai langchain pymilvus tiktoken
(2) Inject & Embed Support Document
from langchain.document_loaders import DirectoryLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.embeddings.openai import OpenAIEmbeddings
from pymilvus import Collection, connections
from langchain.vectorstores.milvus import Milvus
# 1. Connect to Milvus
connections.connect(alias="default", host="localhost", port="19530")
# 2. Load documents
loader = DirectoryLoader("./support_docs", glob="**/*.txt")
docs = loader.load()
# 3. Split text into chunks
splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=50)
chunks = splitter.split_documents(docs)
# 4. Embed chunks
embeddings = OpenAIEmbeddings()
# 5. Store in Milvus
vectorstore = Milvus.from_documents(
chunks,
embedding=embeddings,
collection_name="customer_support_docs"
)
(3) Create The RAG Chain With LangChain
from langchain.chains import RetrievalQA
from langchain.chat_models import ChatOpenAI
retriever = vectorstore.as_retriever()
llm = ChatOpenAI(model="gpt-3.5-turbo")
rag_chain = RetrievalQA.from_chain_type(
llm=llm,
retriever=retriever,
chain_type="stuff",
return_source_documents=True
)
(4) Query The Agent
query = "How can I return a damaged product?"
response = rag_chain(query)
print("Answer:", response["result"])
for doc in response["source_documents"]:
print("Source:", doc.metadata["source"])
(5) Define Tools
from langchain.agents import Tool
# Tool 1: Use your RetrievalQA chain from Step 4
retrieval_tool = Tool(
name="support_doc_lookup",
func=rag_chain.run,
description="Use this to answer customer support queries from FAQs, return policies, etc."
)
# Tool 2: Simulated product API call
def product_api_tool(query: str) -> str:
# You can replace this with a real API call
if "price" in query.lower():
return "The current price of the product is $199."
return "No product data found."
product_tool = Tool(
name="product_api",
func=product_api_tool,
description="Use this to get product info such as pricing, availability, or features."
)
# Tool 3: Escalate to human
def escalate_tool(query: str) -> str:
return f"This question may require human support. Escalating: {query}"
escalate = Tool(
name="escalate_human",
func=escalate_tool,
description="Use this when the question seems out of scope or needs human help."
)
(6) Initialize The Agent
from langchain.agents import initialize_agent
from langchain.agents import AgentType
from langchain.chat_models import ChatOpenAI
llm = ChatOpenAI(model="gpt-3.5-turbo")
# Create Memory
memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)
agent = initialize_agent(
tools=[retrieval_tool, product_tool, escalate],
llm=llm,
agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION,
memory=memory,
verbose=True
)
(7) Ask A Question
response = agent.run("What is the return policy for damaged items?")
print(response)
response2 = agent.run("How much does the X100 wireless speaker cost?")
print(response2)
response3 = agent.run("I want to file a legal complaint.")
print(response3)
