mem0-integration

mem0-integration

Integrate Mem0 (formerly MemGPT) as a universal memory layer for AI agents. Enable persistent memory storage, semantic search across memories, and personalized context retrieval.

Overview

Mem0 provides intelligent memory management for AI applications:

• Persistent storage of conversation history and facts
• Semantic search across stored memories
• User-specific memory isolation
• Automatic memory extraction from conversations
• Support for local and cloud deployments

Capabilities

Memory Operations

• Add memories from text or conversations
• Search memories semantically
• Retrieve relevant context by user/agent
• Update and delete memories
• Get memory history with timestamps

Memory Types

• Conversation memories (dialogue history)
• Fact memories (extracted information)
• Preference memories (user preferences)
• Entity memories (people, places, things)

Storage Backends

• Local SQLite/JSON storage
• PostgreSQL for production
• Qdrant vector database integration
• Cloud-hosted Mem0 platform

Integration Patterns

• LangChain memory integration
• Direct API usage
• MCP server connectivity
• CrewAI and AutoGen compatibility

Usage

Basic Setup

from mem0 import Memory

# Initialize with default local storage
m = Memory()

# Or with custom configuration
config = {
    "vector_store": {
        "provider": "qdrant",
        "config": {
            "host": "localhost",
            "port": 6333,
        }
    },
    "llm": {
        "provider": "openai",
        "config": {
            "model": "gpt-4o-mini",
            "temperature": 0.1,
        }
    }
}
m = Memory.from_config(config)

Adding Memories

# Add memory from conversation
messages = [
    {"role": "user", "content": "I prefer dark mode for all my applications"},
    {"role": "assistant", "content": "I'll remember that you prefer dark mode."}
]
m.add(messages, user_id="user123")

# Add memory from plain text
m.add("User works at Acme Corp as a software engineer", user_id="user123")

# Add with metadata
m.add(
    "Prefers Python over JavaScript",
    user_id="user123",
    metadata={"category": "preferences", "confidence": 0.9}
)