Chain Of Draft

🚀 Chain of Draft (CoD) MCP Server

This MCP server implements the Chain of Draft (CoD) reasoning method, as described in the research paper "Chain of Draft: Reducing Thinking by Writing". CoD is a novel approach that allows LLMs to generate concise yet informative intermediate reasoning steps, significantly reducing token usage when solving tasks.

🚀 Quick Start

✨ Features

• AnalyticsService: Tracks performance metrics across different domains and reasoning methods.
• ComplexityEstimator: Analyzes problems to determine appropriate word limits.
• ExampleDatabase: Manages and retrieves examples, and converts CoT examples to CoD format.
• FormatEnforcer: Ensures reasoning steps adhere to word limits.
• ReasoningSelector: Intelligently selects CoD or CoT based on problem characteristics and historical performance.

📦 Installation

Python Implementation

1. Environment Setup

   • Download and install Python 3.8 or higher.
   • Install the required dependencies, such as anthropic, dotenv, etc.

2. Code Implementation

from client import ChainOfDraftClient

# Create a client
cod_client = ChainOfDraftClient()

# Use directly
result = await cod_client.solve_with_reasoning(
    problem="Solve: 247 + 394 = ?",
    domain="Mathematics"
)

print(f"Answer: {result['final_answer']}")
print(f"Reasoning steps: {result['reasoning_steps']}")
print(f"Tokens used: {result['token_count']}")

JavaScript Implementation

1. Environment Setup

   • Download and install Node.js 14 or higher.
   • Install the required dependencies, such as @anthropic-ai/sdk, dotenv, etc.

2. Code Implementation

import { Anthropic } from "@anthropic-ai/sdk";
import dotenv from "dotenv";

// Load environment variables
dotenv.config();

// Create an Anthropic client
const anthropic = new Anthropic({
  apiKey: process.env.ANTHROPIC_API_KEY,
});

// Import the Chain of Draft client
import chainOfDraftClient from './lib/chain-of-draft-client.js';

// Use the client
async function solveMathProblem() {
  const result = await chainOfDraftClient.solveWithReasoning({
    problem: "Solve: 247 + 394 = ?",
    domain: "Mathematics",
    max_words_per_step: 5
  });
  
  console.log(`Answer: ${result.final_answer}`);
  console.log(`Reasoning steps: ${result.reasoning_steps}`);
  console.log(`Tokens used: ${result.token_count}`);
}

solveMathProblem();

💻 Usage Examples

Basic Usage

The above installation examples also serve as basic usage examples. You can directly use the provided code snippets to solve problems with reasoning.

🔧 Technical Details

Python Implementation

1. AnalyticsService: Tracks performance metrics across different domains and reasoning methods.
2. ComplexityEstimator: Analyzes problems to determine appropriate word limits.
3. FormatEnforcer: Ensures reasoning steps adhere to word limits.
4. ReasoningSelector: Automatically selects CoD or CoT based on problem characteristics and historical performance.

JavaScript Implementation

1. analyticsDb: An in-memory database for tracking performance metrics.
2. complexityEstimator: Analyzes problems to determine complexity and appropriate word limits.
3. formatEnforcer: Ensures reasoning steps adhere to word limits.
4. reasoningSelector: Automatically selects CoD or CoT based on problem characteristics and historical performance.

Both implementations follow the same core principles and provide the same MCP tools, making them interchangeable in most use cases.

📄 License

This project is licensed under the MIT License. You are free to use, modify, and distribute the code.

💡 Contribution Guide

Contributions are welcome! Please refer to CONTRIBUTING.md to learn how to contribute to this project.

🐞 Issue Reporting

Encountered a problem during use? Please submit it to GitHub Issues, and we will resolve it as soon as possible.

📚 References

• Chain of Thought Paper
• Anthropic API Documentation
• dotenv Official Documentation