Context Optimizer MCP

🚀 Context Optimizer MCP

A Model Context Protocol (MCP) server that uses Redis and in-memory caching to optimize and scale large chat history context windows.

✨ Features

• Dual-layer Caching: Combines fast in-memory LRU caching with persistent Redis storage.
• Intelligent Context Management: Automatically summarizes old messages to maintain context within token limits.
• Rate Limiting: Redis-based rate limiting with burst protection.
• API Compatibility: A plug-and-play replacement for the Anthropic API with enhanced context handling capabilities.
• Metrics Collection: Built-in performance monitoring and logging.

📚 Documentation

This MCP server acts as middleware between an application and an LLM provider (currently supporting Anthropic's Claude model). It intelligently manages context through the following strategies:

• When the message length exceeds the token limit, the system automatically summarizes old messages.
• If the number of messages exceeds the threshold, the system retains the most recent N messages and discards earlier ones.

🚀 Quick Start

Prerequisites

• Operating System: Linux or macOS (Windows may require additional configuration).
• Memory Requirement: At least 4GB RAM (8GB or more recommended).
• Disk Space: At least 10GB of available space.
• Install Dependencies:
  • Python 3.8+
  • Redis 6.x
  • pip

📦 Installation

Install using npx

npx mcp install degenhero/context-optimizer-mcp

Manual Installation

1. Clone the repository:

   git clone https://github.com/yourusername/context-optimizer-mcp.git
   cd context-optimizer-mcp
   

2. Install dependencies:

   pip install -r requirements.txt
   

3. Start the service:

   python app.py
   

Docker Installation

1. Pull the image:

   docker pull yourusername/context-optimizer-mcp:latest
   

2. Run the container:

   docker run -d --name context-optimizer-mcp -p 5000:5000 yourusername/context-optimizer-mcp:latest
   

💻 Configuration

Create or edit the config.yml file with the following content:

# System configuration
system:
  port: 5000
  debug_mode: false
  log_level: INFO

# Cache configuration
cache:
  type: redis
  host: localhost
  port: 6379
  db: 0

# Context limits
context_limit:
  max_tokens: 1200
  max_messages: 20

💻 Usage Examples

Request Format

All requests are sent via HTTP with the following requirements:

• Supported MIME type: application/json
• Request body format:

  {
    "messages": [
      { "role": "system", "content": "You are an honest and friendly assistant." },
      { "role": "user", "content": "This is the user's query." }
    ],
    "model": "gpt-3.5-turbo",
    "temperature": 0.7,
    "conversation_id": "1234567890"
  }
  

Special Parameters

Add the following optional parameters to the request:

{
  "messages": [ ... ],
  "model": "...",
  "temperature": ...,
  "conversation_id": "...",
  "context_optimization": true
}

Testing

Run the test script:

python test.py

Or execute manually:

# After starting the service, open a browser and visit http://localhost:5000/test

🔧 Technical Details

Context Summarization

When the number of tokens in a request exceeds the max_tokens limit, the system automatically summarizes old messages.

Conversation Continuity

If the conversation_id is the same, the server retains the conversation history and merges it into the current context.

Performance Considerations

• In-memory caching provides the fastest access speed.
• Redis enables persistence and multi-instance sharing.
• Summarization operations add some latency to requests that exceed the threshold.

📚 Documentation

For more detailed documentation, refer to the docs/ directory:

• Architecture Overview
• Contribution Guide

📄 License

MIT