Skip to content

Latest commit

 

History

History
132 lines (93 loc) · 3.77 KB

README.md

File metadata and controls

132 lines (93 loc) · 3.77 KB

🤖 Minions Tools Comparison App

A Streamlit application for comparing different approaches to using MCP (Model Context Protocol) tools with language models. This app allows you to test and compare:

  1. Direct Ollama + MCP: Using OllamaClient with MCP tools directly
  2. Minions MCP: Using the full SyncMinionsMCP framework with task decomposition

📋 Prerequisites

Before running the app, make sure you have:

  1. Ollama installed with llama3.2:1b model available

    ollama pull llama3.2:1b

  2. MCP configuration (mcp.json) in the root directory with filesystem server setup

  3. OpenAI API key configured for the remote client (gpt-4o-mini)

  4. Python dependencies installed (see requirements.txt)

🛠️ Installation

  1. Navigate to the app directory:

    cd apps/minions-tools

  2. Install dependencies:

    pip install -r requirements.txt

  3. Ensure MCP configuration exists: Make sure you have a mcp.json file in the project root with filesystem server configuration.

🎯 Usage

  1. Start the Streamlit app:

    streamlit run app.py

  2. Configure your comparison:

    • Enter a task in the sidebar (filesystem-related tasks work best)
    • Select which methods you want to compare
    • Click "🚀 Run Comparison"

  3. View results:

    • Each method's results appear in separate tabs
    • Performance metrics are displayed for each method
    • A summary table compares execution times

📊 Method Comparison

Direct Ollama + MCP

  • Best for: Simple, direct tool calling scenarios
  • Pros: Lightweight, fast, direct integration
  • Cons: Limited coordination capabilities

Minion Protocol

  • Best for: Tasks requiring coordination between models
  • Pros: Balance of performance and capability
  • Cons: More complex than direct approach

Minions MCP

  • Best for: Complex tasks requiring planning and decomposition
  • Pros: Most sophisticated approach with multi-round processing
  • Cons: Most resource-intensive, slower execution

🔧 Configuration

The app uses several configuration options:

  • Local Model: llama3.2:1b (configurable in the code)
  • Remote Model: gpt-4o-mini (requires OpenAI API key)
  • MCP Server: filesystem (configurable via mcp.json)
  • Max Rounds: 3 (configurable per method)

📝 Example Tasks

Here are some example tasks that work well with the filesystem MCP server:

  1. Directory Analysis:

    Can you show me the directory structure of the examples folder and summarize what you find?

  2. File Search:

    Find all Python files in the current directory and list their names.

  3. Content Analysis:

    Read the contents of README.md and provide a summary of the main points.

🐛 Troubleshooting

Common Issues

  1. "Failed to initialize clients"

    • Check that Ollama is running and the model is available
    • Verify MCP configuration exists and is valid
    • Ensure OpenAI API key is properly configured

  2. "Tool execution failed"

    • Verify MCP server is running
    • Check file paths and permissions
    • Review MCP server logs

  3. Import errors

    • Ensure you're running from the correct directory
    • Check that the minions package is properly installed
    • Verify Python path includes the project root

Performance Tips

  • Start with simpler tasks to verify setup
  • Use the Direct Ollama + MCP method for quick tests
  • Enable only necessary methods for faster comparisons
  • Monitor system resources when running multiple methods

🔗 Related Documentation