# Mem0

[Mem0](https://www.mem0.ai/) provides a smart, self-improving memory layer for Large Language Models (LLMs), enabling developers to create personalized AI experiences that evolve with each user interaction. This page focuses on **Mem0 Open Source** as the primary integration for AG2; a hosted Mem0 Platform alternative is included below.

At a high level, Mem0 Platform offers comprehensive memory management, self-improving memory capabilities, cross-platform consistency, and centralized memory control for AI applications. For more info, check out the [Mem0 Platform Documentation](https://docs.mem0.ai/).

|  |  |
| --- | --- |
| 🧠 **Comprehensive Memory Management** | Manage long-term, short-term, semantic, and episodic memories |
| 🔄 **Self-Improving Memory** | Adaptive system that learns from user interactions |
| 🌐 **Cross-Platform Consistency** | Unified user experience across various AI platforms |
| 🎛️ **Centralized Memory Control** | Effortless storage, updating, and deletion of memories |
| 🚀 **Simplified Development** | API-first approach for streamlined integration |

## Installation

Mem0 Open Source works seamlessly with various AI applications.

1. **Install Mem0 SDK:**
   
   ```
   pip install mem0ai
   ```

2. **Configure Your Environment:** Mem0 Open Source uses your LLM provider by default. Add your OpenAI key (or configure another provider).
   
   ```
   OPENAI_API_KEY=<YOUR_OPENAI_API_KEY>
   ```

3. **Initialize Mem0 (Open Source):**
   
   ```
   from mem0 import Memory
   memory = Memory()
   ```

After initializing Mem0, you can start using its memory management features in your AI application.

> Note: `Memory` is the open-source client. The hosted Mem0 Platform uses `MemoryClient` (see below).

## Features

- **Long-term Memory**: Store and retrieve information persistently across sessions
- **Short-term Memory**: Manage temporary information within a single interaction
- **Semantic Memory**: Organize and retrieve conceptual knowledge
- **Episodic Memory**: Store and recall specific events or experiences
- **Self-Improving System**: Continuously refine understanding based on user interactions

## Common Use Cases

- Personalized Learning Assistants
- Customer Support AI Agents
- Healthcare Assistants
- Virtual Companions

## Mem0 Platform Examples

### AG2 with Mem0 (Open Source) Example

This example demonstrates how to use Mem0 with AG2 to create a conversational AI system with memory capabilities.

```python
import os
from autogen import ConversableAgent, LLMConfig
from mem0 import Memory

# Set up environment variables
os.environ["OPENAI_API_KEY"] = "your_openai_api_key"

# Initialize Agent and Memory
agent = ConversableAgent(
    "chatbot",
    llm_config=LLMConfig({"api_type": "openai", "model": "gpt-5-nano", "api_key": os.environ.get("OPENAI_API_KEY")}),
    code_execution_config=False,
    function_map=None,
    human_input_mode="NEVER",
)

memory = Memory()

# Insert a conversation into memory
conversation = [
   {
        "role": "assistant",
        "content": "Hi, I'm Best Buy's chatbot!\n\nThanks for being a My Best Buy TotalTM member.\n\nWhat can I help you with?"
    },
    {
        "role": "user",
        "content": "Seeing horizontal lines on our tv. TV model: Sony - 77\" Class BRAVIA XR A80K OLED 4K UHD Smart Google TV"
    },
]

memory.add(messages=conversation, user_id="customer_service_bot")

# Agent Inference
data = "Which TV am I using?"

relevant_memories = memory.search(data, user_id="customer_service_bot")
flatten_relevant_memories = "\n".join([m["memory"] for m in relevant_memories["results"])

prompt = f"""Answer the user question considering the memories.\nMemories:\n{flatten_relevant_memories}\n\n\nQuestion: {data}\n"""

reply = agent.generate_reply(messages=[{"content": prompt, "role": "user"}])
print("Reply :", reply)

# Multi Agent Conversation
manager = ConversableAgent(
    "manager",
    system_message="You are a manager who helps in resolving customer issues.",
    llm_config=LLMConfig({"api_type": "openai", "model": "gpt-5-nano", "temperature": 0, "api_key": os.environ.get("OPENAI_API_KEY")}),
    human_input_mode="NEVER"
)

customer_bot = ConversableAgent(
    "customer_bot",
    system_message="You are a customer service bot who gathers information on issues customers are facing.",
    llm_config=LLMConfig({"api_type": "openai", "model": "gpt-5-nano", "temperature": 0, "api_key": os.environ.get("OPENAI_API_KEY")}),
    human_input_mode="NEVER"
)

data = "What can the customer do to resolve the issue?"

prompt = f"""
Context:\n{flatten_relevant_memories}\n\n\nQuestion: {data}\n"""

result = manager.send(prompt, customer_bot, request_reply=True)

memory.vector_store.client.close()
memory._telemetry_vector_store.client.close()
```

Access the complete code from this notebook: [Mem0 with AG2](https://colab.research.google.com/drive/1NZEwC9w6V2S6hYmK7l2SQ9jhQrG1uKk8?usp=sharing)

This example showcases: 1. Setting up AG2 agents and Mem0 memory 2. Adding a conversation to Mem0 memory 3. Using Mem0 to retrieve relevant memories for agent inference 4. Implementing a multi-agent conversation with memory-augmented context

### Mem0 Platform (Hosted) Alternative

If you prefer the hosted Mem0 Platform, use the `MemoryClient` with an API key:

```python
import os
from mem0 import MemoryClient

client = MemoryClient(api_key=os.getenv("MEM0_API_KEY"))
results = client.search("Which TV am I using?", filters={"user_id": "customer_service_bot"})
```

For more Mem0 examples, visit our [documentation](https://docs.mem0.ai/examples).