knowledge-base/docs/examples-llama-index-mem0.md

LlamaIndex ReAct Agent with Mem0

Overview

A ReAct agent combines reasoning and action capabilities, making it versatile for tasks requiring both thought processes (reasoning) and interaction with tools or APIs (acting). Mem0 as memory enhances these capabilities by allowing the agent to store and retrieve contextual information from past interactions.

This guide demonstrates how to create a ReAct Agent with LlamaIndex that uses Mem0 as the memory store, showcasing the dramatic difference between agents with and without memory capabilities.

Setup

Installation

pip install llama-index-core llama-index-memory-mem0

Initialize the LLM

import os
from llama_index.llms.openai import OpenAI

os.environ["OPENAI_API_KEY"] = "<your-openai-api-key>"
llm = OpenAI(model="gpt-4o")

Initialize Mem0 Memory

You can find your API key here. Read about Mem0 Open Source.

os.environ["MEM0_API_KEY"] = "<your-mem0-api-key>"

from llama_index.memory.mem0 import Mem0Memory

context = {"user_id": "david"}
memory_from_client = Mem0Memory.from_client(
    context=context,
    api_key=os.environ["MEM0_API_KEY"],
    search_msg_limit=4,  # optional, default is 5
)

Create Agent Tools

These tools will be used by the agent to perform actions:

from llama_index.core.tools import FunctionTool

def call_fn(name: str):
    """Call the provided name.
    Args:
        name: str (Name of the person)
    """
    return f"Calling... {name}"

def email_fn(name: str):
    """Email the provided name.
    Args:
        name: str (Name of the person)
    """
    return f"Emailing... {name}"

def order_food(name: str, dish: str):
    """Order food for the provided name.
    Args:
        name: str (Name of the person)
        dish: str (Name of the dish)
    """
    return f"Ordering {dish} for {name}"

# Create tool instances
call_tool = FunctionTool.from_defaults(fn=call_fn)
email_tool = FunctionTool.from_defaults(fn=email_fn)
order_food_tool = FunctionTool.from_defaults(fn=order_food)

Initialize the Agent with Memory

from llama_index.core.agent import FunctionCallingAgent

agent = FunctionCallingAgent.from_tools(
    [call_tool, email_tool, order_food_tool],
    llm=llm,
    memory=memory_from_client,  # Mem0 memory integration
    verbose=True,
)

Building User Context

Let's start by having the agent learn about the user through conversation:

Introduction

Input:

response = agent.chat("Hi, My name is David")
print(response)

Output:

> Running step bf44a75a-a920-4cf3-944e-b6e6b5695043. Step input: Hi, My name is David
Added user message to memory: Hi, My name is David
=== LLM Response ===
Hello, David! How can I assist you today?

Learning Preferences

Input:

response = agent.chat("I love to eat pizza on weekends")
print(response)

Output:

> Running step 845783b0-b85b-487c-baee-8460ebe8b38d. Step input: I love to eat pizza on weekends
Added user message to memory: I love to eat pizza on weekends
=== LLM Response ===
Pizza is a great choice for the weekend! If you'd like, I can help you order some. Just let me know what kind of pizza you prefer!

Communication Preferences

Input:

response = agent.chat("My preferred way of communication is email")
print(response)

Output:

> Running step 345842f0-f8a0-42ea-a1b7-612265d72a92. Step input: My preferred way of communication is email
Added user message to memory: My preferred way of communication is email
=== LLM Response ===
Got it! If you need any assistance or have any requests, feel free to let me know, and I can communicate with you via email.

Comparing Agents: With vs Without Memory

Using the Agent WITHOUT Memory

Setup:

agent_no_memory = FunctionCallingAgent.from_tools(
    [call_tool, email_tool, order_food_tool],
    # memory is not provided
    llm=llm,
    verbose=True,
)

Input:

response = agent_no_memory.chat("I am feeling hungry, order me something and send me the bill")
print(response)

Output:

> Running step e89eb75d-75e1-4dea-a8c8-5c3d4b77882d. Step input: I am feeling hungry, order me something and send me the bill
Added user message to memory: I am feeling hungry, order me something and send me the bill
=== LLM Response ===
Please let me know your name and the dish you'd like to order, and I'll take care of it for you!

Result: The agent has no memory of previous conversations and cannot act on user preferences.

Using the Agent WITH Memory

Setup:

agent_with_memory = FunctionCallingAgent.from_tools(
    [call_tool, email_tool, order_food_tool],
    llm=llm,
    memory=memory_from_client,  # Mem0 memory integration
    verbose=True,
)

Input:

response = agent_with_memory.chat("I am feeling hungry, order me something and send me the bill")
print(response)

Output:

> Running step 5e473db9-3973-4cb1-a5fd-860be0ab0006. Step input: I am feeling hungry, order me something and send me the bill
Added user message to memory: I am feeling hungry, order me something and send me the bill
=== Calling Function ===
Calling function: order_food with args: {"name": "David", "dish": "pizza"}
=== Function Output ===
Ordering pizza for David
=== Calling Function ===
Calling function: email_fn with args: {"name": "David"}
=== Function Output ===
Emailing... David
> Running step 38080544-6b37-4bb2-aab2-7670100d926e. Step input: None
=== LLM Response ===
I've ordered a pizza for you, and the bill has been sent to your email. Enjoy your meal! If there's anything else you need, feel free to let me know.

Result: The agent remembers:

• User's name is David
• User loves pizza on weekends
• User prefers email communication
• Automatically orders pizza and sends bill via email

Key Benefits of Memory Integration

1. Personalized Responses

• Agents remember user preferences and act accordingly
• No need to repeat information in every conversation

2. Contextual Decision Making

• Agents can make informed decisions based on past interactions
• Improved user experience through continuity

3. Efficient Interactions

• Reduced friction in user-agent communication
• Faster task completion with fewer prompts needed

4. Learning and Adaptation

• Agents improve over time by learning from interactions
• Better understanding of user behavior patterns

Configuration Options

Memory Configuration

memory_from_client = Mem0Memory.from_client(
    context=context,
    api_key=os.environ["MEM0_API_KEY"],
    search_msg_limit=4,  # Controls how many past messages to retrieve
)

Context Parameters

• user_id: Unique identifier for memory isolation between users
• search_msg_limit: Number of relevant past messages to include in context

Use Cases

1. Personal Assistants

• Remember user preferences, schedules, and habits
• Provide personalized recommendations and actions

2. Customer Support Agents

• Maintain conversation history and user preferences
• Provide consistent support across multiple interactions

3. E-commerce Assistants

• Remember shopping preferences and past purchases
• Suggest relevant products and services

4. Educational Tutors

• Track learning progress and adapt teaching methods
• Remember student strengths and areas for improvement

Best Practices

1. Context Management

• Use meaningful user IDs for proper memory isolation
• Adjust search_msg_limit based on conversation complexity

2. Tool Design

• Create tools that can leverage memory context
• Design functions with clear parameter definitions

3. Memory Hygiene

• Regularly review and update memory contents
• Implement privacy controls for sensitive information

4. Testing

• Test agents both with and without memory to understand impact
• Validate memory persistence across sessions

Conclusion

Integrating Mem0 with LlamaIndex ReAct agents transforms static, forgetful assistants into intelligent, context-aware companions. The dramatic difference between agents with and without memory demonstrates the power of persistent context in creating truly helpful AI assistants.

The combination enables:

• Continuity across conversations
• Personalization based on user preferences
• Efficiency through reduced repetition
• Intelligence through accumulated context

This integration makes AI agents more human-like in their ability to remember and build upon past interactions, creating a foundation for truly intelligent and helpful AI systems.