Metis MCP Demo

🚀 Metis MCP Tools

This repository contains a set of Model Context Protocol (MCP) tools that enhance Metis RAG applications. These tools offer functionalities for database management, vector storage operations, document processing, and LLM interaction.

🚀 Quick Start

What is MCP?

The Model Context Protocol (MCP) is a standardized approach that allows language models to interact with external tools and resources. It enables models to access data, perform operations, and leverage specialized capabilities beyond the limitations of their training data.

✨ Features

Tool Overview

This repository includes the following MCP tools:

1. Database Management Tool

The database management tool provides functionality for interacting with PostgreSQL databases in Metis applications.

Features:

• Initialize database connections
• Execute SQL queries
• Retrieve table structures
• List database tables

💻 Usage Examples

Basic Usage

// Initialize a database connection
const dbResult = await useMcpTool({
  serverName: 'database-tool',
  toolName: 'init_database_connection',
  arguments: {
    host: 'localhost',
    port: 5432,
    database: 'metis_db',
    user: 'metis_user',
    password: 'password123'
  }
});

// Execute a query
const queryResult = await useMcpTool({
  serverName: 'database-tool',
  toolName: 'execute_query',
  arguments: {
    query: 'SELECT * FROM documents LIMIT 10'
  }
});

2. Vector Storage Tool

The vector storage tool provides functionality for interacting with vector databases like Chroma in Metis applications.

Features:

• Initialize vector storage connections
• Create and manage collections
• Add documents to collections
• Query collections for similar documents
• Delete documents from collections

💻 Usage Examples

Basic Usage

// Initialize a vector storage connection
const vsResult = await useMcpTool({
  serverName: 'vector-store-tool',
  toolName: 'init_vector_store',
  arguments: {
    host: 'localhost',
    port: 8000
  }
});

// Create or get a collection
const collectionResult = await useMcpTool({
  serverName: 'vector-store-tool',
  toolName: 'get_or_create_collection',
  arguments: {
    name: 'metis_documents'
  }
});

// Query a collection
const queryResult = await useMcpTool({
  serverName: 'vector-store-tool',
  toolName: 'query_collection',
  arguments: {
    queryTexts: ['What is the working principle of quantum computing?'],
    nResults: 5
  }
});

3. Document Processing Tool

The document processing tool provides functionality for loading, processing, and chunking documents in Metis applications.

Features:

• Load documents
• Process document content
• Split documents into chunks
• Convert formats
• Extract metadata

💻 Usage Examples

Basic Usage

// Load and process a PDF document
const pdfResult = await useMcpTool({
  serverName: 'document-processing-tool',
  toolName: 'process_pdf',
  arguments: {
    filePath: './docs/report.pdf'
  }
});

// Chunk text
const textChunks = await useMcpTool({
  serverName: 'document-processing-tool',
  // The original code seems incomplete here, keep it as it is
  //... original incomplete code
});