Memory Graph

🚀 MCP Memory System Using Redis Graph Database

This project implements a memory storage system for LLM conversations, utilizing Redis Graph for long - term memory storage.

🚀 Quick Start

This guide will help you quickly understand how to install and use the MCP Memory System.

✨ Features

Core Functions

• Memory Storage: Persist key information from conversations into the Redis graph database.
• Retrieval Capability: Support quick memory lookup based on ID, type, and content.
• Associative Relationships: Allow the establishment of associations between memories for complex queries and analysis.

Supported Memory Types

• General Memory: A universal type suitable for various scenarios.
• Finance Type: Specifically designed for finance - related fields, such as policies and transaction records.
• Technical Issues: Handle discussions related to programming or systems.

📦 Installation

Prerequisites

• Docker and Docker Compose
• Node.js (v16 or higher)

Starting Redis with the RedisGraph Module

1. Use Docker Compose to start the Redis container and load the RedisGraph module:

docker-compose up -d

2. Verify that Redis is running properly and the RedisGraph module is loaded:

docker exec -it mcp-memory-redis-1 redis-cli

After entering the Redis CLI, you can view module information using the following command:

MODULE LIST

💻 Usage Examples

Memory Management Tools

• Create Memory: Use the createMemory method to store new memories in the database.
• Retrieve Memory: Get a specific memory by ID using the getMemory method.
• Search Memories: Use the searchMemories method to find relevant memories based on keywords or conditions.

Basic Usage

// Create a new Finance type memory
await memoryService.createMemory({
  type: 'Finance',
  title: 'Credit card policy update',
  content: 'The new credit card policy will take effect next month.',
  tags: ['Credit card', 'Financial policy']
});

// Retrieve memory by ID
const financeMemory = await memoryService.getMemory('12345');

// Search for all Finance type memories related to "debit card"
const debitCardMemories = await memoryService.searchMemories({
  type: 'Finance',
  contentContains: 'debit card'
});

🔧 Technical Details

System Architecture

Data Storage Structure

• Nodes: Each memory corresponds to a graph database node, containing attributes such as type, title, content, and tags.
• Relationships: Used to describe associations between different memories, such as "related", "referenced", or "influenced".

Example Query Commands

# List all graph databases
GRAPH.LIST

# Count the total number of nodes
GRAPH.QUERY memory "MATCH (n) RETURN count(n)"

# Query Finance type memories
GRAPH.QUERY memory "MATCH (n:Finance) RETURN n.id, n.title, n.content"

# Search for specific content
GRAPH.QUERY memory "MATCH (n) WHERE n.content CONTAINS 'debit card' RETURN n.id, n.type, n.content"

# Display node associations
GRAPH.QUERY memory "MATCH (a)-[r]->(b) RETURN a.id, type(r), b.id"

Docker Run Guide

Building and Running the Image

# Build the image locally
docker build -t mcp/memory .

# Start the container (assuming Redis is running in the same network)
docker run --rm -i --network=mcp-memory_default -e REDIS_URL=redis://redis:6379 mcp/memory

Testing Guide

Executing Test Scripts

# Test the Redis connection
npm run test:redis

# Test basic memory operations
npm run test:memory

# Test the full service functionality
npm run test:service

In summary, through the above content, you can comprehensively understand the functions, installation methods, and usage scenarios of the MCP Memory System. For further detailed technical information or actual deployment, please refer to the relevant documentation and technical support resources.