mcp_llm_inferencer

Introduction

The mcp_llm_inferencer is an open-source library designed to leverage the power of Large Language Models (LLMs) such as Claude and OpenAI's GPT to convert prompt-mapped inputs into concrete components for MCP servers. These components include tools, resource templates, and prompt handlers, making it a versatile tool for developers working with MCP server environments.

Features

• LLM Call Engine: Efficiently calls LLMs with built-in retry and fallback logic to ensure reliable responses.
• Interchangeable Claude & OpenAI Support: Seamlessly switch between Claude and OpenAI APIs based on your preference or availability.
• Streaming Support for Claude Desktop: Stream responses directly from Claude Desktop, providing real-time feedback.
• Tool and Resource Response Validation: Ensures that the generated tools and resources meet predefined criteria before deployment.
• Structured Output Bundling: Organizes output into structured bundles per component, simplifying integration and use.

Installation Instructions

Prerequisites

• Python 3.6 or higher
• An API key from Claude or OpenAI

Installing mcp_llm_inferencer

1. Clone the repository:

   git clone https://github.com/your-repo/mcp_llm_inferencer.git
   cd mcp_llm_inferencer
   

2. Install the package using pip:

   pip install .
   

3. Set up your API keys as environment variables:

   • For Claude:

     export CLAUDE_API_KEY='your-claude-api-key'
     

   • For OpenAI:

     export OPENAI_API_KEY='your-openai-api-key'
     

Usage Examples

Basic Example

Here is a simple example demonstrating how to use mcp_llm_inferencer with the OpenAI API:

from mcp_llm_inferencer import MCPInferencer

# Initialize the inferencer with OpenAI API
inferencer = MCPInferencer(api_type='openai')

# Define your prompt
prompt = "Generate a tool to extract emails from text."

# Generate components
components = inferencer.generate_components(prompt)

print(components)

Advanced Example with Claude

This example shows how to use the library with Claude API and handle streaming responses:

from mcp_llm_inferencer import MCPInferencer

# Initialize the inferencer with Claude API
inferencer = MCPInferencer(api_type='claude', stream=True)

# Define your prompt
prompt = "Create a resource template for an S3 bucket."

# Generate components with streaming support
for component in inferencer.generate_components(prompt):
    print(component)

API Documentation

Class: MCPInferencer

Initialization

MCPInferencer(api_type, api_key=None, stream=False)

• api_type (str): The type of LLM API to use ('claude' or 'openai').
• api_key (str, optional): The API key for the specified LLM. If not provided, it will attempt to read from environment variables.
• stream (bool, optional): Enable streaming support for Claude Desktop.

Methods

• generate_components(prompt)
  • Generates MCP server components based on the given prompt.
  • prompt (str): The input prompt to be sent to the LLM.
  • Returns: A dictionary or stream of dictionaries containing generated components.

Example Method Usage

components = inferencer.generate_components("Generate a tool for sentiment analysis.")

License

mcp_llm_inferencer is released under the MIT License. See the LICENSE file for more details.

Feel free to contribute to mcp_llm_inferencer by submitting issues or pull requests on our GitHub repository.

⚠️ Development Status

This library is currently in early development. Some tests may be failing with the following issues:

Contributions to fix these issues are welcome! Please submit a pull request if you have a solution.