Skip to content

MCP_SETUP_GUIDE.md

Latest commit

 

History

History
executable file
·
159 lines (119 loc) · 4.18 KB

MCP_SETUP_GUIDE.md

File metadata and controls

executable file
·
159 lines (119 loc) · 4.18 KB

IPFS Accelerate MCP Setup Guide

This guide will help you set up and use the IPFS Accelerate MCP (Model Context Protocol) servers with VS Code and the web dashboard.

Quick Start

1. Install Dependencies

pip install fastapi uvicorn fastmcp psutil numpy

2. Test the Servers

JSON-RPC Server with Web Dashboard

cd /path/to/ipfs_accelerate_py
python mcp_jsonrpc_server.py --port 8003

Then open your browser to: http://localhost:8003

VS Code MCP Integration

cd /path/to/ipfs_accelerate_py
python vscode_mcp_server.py

CLI Tools

# Test comprehensive server
python tools/comprehensive_mcp_server.py --help

# Test standalone server
python -m ipfs_accelerate_py.mcp.standalone --help

VS Code Configuration

To use the MCP server with VS Code:

  1. Install the MCP extension for VS Code
  2. Add this configuration to your VS Code settings or MCP config file:
{
  "mcpServers": {
    "ipfs-accelerate": {
      "command": "python",
      "args": ["/path/to/ipfs_accelerate_py/vscode_mcp_server.py"],
      "cwd": "/path/to/ipfs_accelerate_py"
    }
  }
}

Replace /path/to/ipfs_accelerate_py with the actual path to your repository.

Web Dashboard Features

The web dashboard provides a comprehensive interface for testing AI models:

  • Text Generation: Test language model text generation
  • Text Classification: Classify text into categories
  • Text Embeddings: Generate vector embeddings for text
  • Audio Processing: Transcribe audio files
  • Vision Models: Image classification and generation
  • Multimodal: Visual question answering (coming soon)
  • Specialized Tools: Code generation and more
  • Model Recommendations: Get AI model suggestions
  • Model Manager: Search and manage AI models

Available CLI Commands

JSON-RPC Server

python mcp_jsonrpc_server.py --help
python mcp_jsonrpc_server.py --port 8003 --verbose

Comprehensive MCP Server

python tools/comprehensive_mcp_server.py --help
python tools/comprehensive_mcp_server.py --transport stdio  # For VS Code
python tools/comprehensive_mcp_server.py --transport sse --port 8004  # For HTTP

Standalone Server

python -m ipfs_accelerate_py.mcp.standalone --help
python -m ipfs_accelerate_py.mcp.standalone --fastapi --port 8005

API Testing

You can test the JSON-RPC API directly:

curl -X POST http://localhost:8003/jsonrpc \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc": "2.0", "method": "list_models", "id": 1}'

Troubleshooting

Common Issues

  1. Missing Dependencies: Install all required packages:

    pip install fastapi uvicorn fastmcp psutil numpy

  2. Port Already in Use: Change the port number:

    python mcp_jsonrpc_server.py --port 8004

  3. VS Code Connection Issues:

    • Make sure the MCP extension is installed
    • Check that the path in the configuration is correct
    • Try running the server manually first to check for errors

Verification

To verify everything is working:

  1. Start the JSON-RPC server: python mcp_jsonrpc_server.py --port 8003
  2. Open http://localhost:8003 in your browser
  3. Test the "Text Generation" tab with a sample prompt
  4. You should see a JSON response with generated text

Support

If you encounter issues:

  1. Check the console/terminal output for error messages
  2. Verify all dependencies are installed
  3. Make sure no other services are using the same ports
  4. Check file paths in configuration files

Features Status

✅ Working Features:

  • JSON-RPC server with web dashboard
  • All API endpoints (28 methods available)
  • Static file serving
  • CLI tools
  • VS Code MCP wrapper
  • Dashboard UI with multiple model categories

⚠️ Known Limitations:

  • Some optional dependencies may not be available (DuckDB, IPFS, etc.)
  • Models are currently mock/demo responses
  • Full model integration requires additional setup

🔧 In Progress:

  • Multimodal processing features
  • Real model loading and inference
  • Advanced hardware acceleration