Introduction to RAG and Agents

Amazon handles 300 million active customer accounts searching through 350+ million products. How do you build a search experience that understands "I need something for my mom's birthday, she likes gardening" instead of just matching keywords?

That's the power of RAG + Agents for e-commerce.

What is RAG?

# Simple RAG flow for e-commerce
from langchain.embeddings import OpenAIEmbeddings
from langchain.vectorstores import Pinecone
from langchain.chat_models import ChatOpenAI

# 1. User asks a question
query = "waterproof hiking boots under $150"

# 2. Retrieve relevant products from vector database
retriever = vector_store.as_retriever(search_kwargs={"k": 5})
relevant_products = retriever.get_relevant_documents(query)

# 3. Generate response with context
llm = ChatOpenAI(model="gpt-4")
response = llm.predict(f"""
Based on these products: {relevant_products}
Answer the customer's question: {query}
Recommend the best options and explain why.
""")

RAG vs Fine-tuning

What are AI Agents?

An AI Agent is an LLM that can:

• Reason about what steps to take
• Use tools to gather information or take actions
• Iterate until the task is complete

from langchain.agents import AgentExecutor, create_openai_tools_agent
from langchain.tools import Tool

# Define tools the agent can use
tools = [
    Tool(
        name="product_search",
        func=search_products,
        description="Search for products by description, category, or attributes"
    ),
    Tool(
        name="check_inventory",
        func=check_inventory,
        description="Check if a product is in stock and estimated delivery"
    ),
    Tool(
        name="compare_products",
        func=compare_products,
        description="Compare features and prices of multiple products"
    ),
    Tool(
        name="get_reviews",
        func=get_reviews,
        description="Get customer reviews and ratings for a product"
    )
]

# Create an agent that can reason and use tools
agent = create_openai_tools_agent(llm, tools, prompt)
agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)

# Agent handles complex queries by deciding which tools to use
response = agent_executor.invoke({
    "input": "I'm looking for a laptop for video editing under $1500. "
             "What's the best option that's in stock?"
})

RAG + Agents: The Complete Picture

When you combine RAG with agents, you get a system that can:

E-commerce Use Cases We'll Build

Real-World Success Stories

Key Takeaways

Next up: Vector Databases - The foundation for semantic search over your product catalog.